All right. Tha nks so much. Good morning, everyone, and thank you so much for joining our 2025 Investor Day. I'm Erica Gessert. I'm the CFO of Upwork. For those who don't know me, I think most of you do. We are absolutely delighted to have everyone here with us today. I know I speak for everyone in our company when I say we are incredibly excited about this time for Upwork. It's such a great time for our customers, for our employees, and of course, for our shareholders. Over the next three hours, we plan to give you an overview of the significant opportunity ahead for Upwork and our strategy to capture that opportunity to create long-term shareholder value. First, of course, before we get started, I want to call your attention to our safe harbor statement.
During today's event, we will be making statements related to our business that are forward-looking. Please do read the note up here at your leisure. These statements are not guarantees of future performance and are subject to a variety of risks and uncertainties. Our actual results could differ materially, and we will be sharing non-GAAP financial measures today, which have reconciliations in the back of the deck. All right, let's do it. We have a fantastic agenda. First, of course, our own Hayden Brown will be up here, yes, to talk about our journey and strategy. Next, we will have Dave Bottoms. Where is Dave? There he is. Our amazing GM of Marketplace. He is going to talk about the tremendous market opportunity both for AI native marketplace and for the SMB segment on our platform.
Peter Sanborn, our VP of Strategy, Corporate Development and Partnerships, will be here leading a customer panel. You'll get to meet some of our customers up here to really see the value that we bring on our platform to our customers every day. After that, we will have Andrew Rabinovich, our formidable CTO and head of AI and the previous CEO of Headroom up here to talk about the emerging A gentic AI opportunity on our platform. Of course, Ernesto, our GM of Enterprise, will be up here to talk about the market that we're unlocking with Lifted. I'll wrap up by sharing the financial targets for our business, which I'm sure everyone is eager to hear about. Then we'll have time for Q&A. We are incredibly excited. As you all know at earnings, this is an inflection point for Upwork.
We've returned to GSV growth two quarters early, and it's really up and to the right from here for our business. With that, with no further ado, I would love to invite Hayden Brown up to the stage.
We dream big. Looking into the future and always staying a step ahead through every chapter of work, from the rigid nine to five grind to working how and when we want, desks to wherever this is.
Hey.
Business as usual to business as anything but. The story of work is still unfolding, ushering in a new era where AI helps talent make more of their talent and businesses make more of their time, pushing past what was possible just an instant ago, embracing AI and agents to augment human talent and deliver better work outcomes. The story is far from over. AI can draft, build, code, and do whatever this is. The vision, the meaning, the judgment, the taste, that comes from us. We're leading the way, boldly creating opportunities and shaping the future of work because the story of work is ours.
Music here. I'll walk on music.
Good morning, everyone. I am so excited to be here today and to introduce all of you to the new Upwork. We have done a lot of work over the time that I've been at this company. I've been leading this company for six years, and the company today is completely different from when I stepped into the role. If you thought you knew Upwork, you are in for a surprise. No changes that we've made to this business are more profound than those that we made in the last three years. We have fundamentally transformed this business. We have rebuilt it for speed, for scale, and for the new era of human and AI collaboration. Let me be clear. This is not a refresh. This is a full reinvention of our business on every dimension.
It's how we're growing and how we've achieved our accelerated GSV growth that Erica mentioned two quarters ahead of schedule. It's how we're achieving record profitability. Today, you'll see what we've done and how we've done it. You'll see how we've reinvented our product, our customer experience, and our operations for this next era and this next chapter of our business. Look around. No one else is making these types of moves. We have taken bold steps to reinvent and reposition this business to win in this next era of work. It's been a lot of work for our team, but today we are so excited to share it with all of you. This work is underpinned by a new AI-native platform that we have launched across the business, and this is how we have achieved accelerated GSV growth.
It's how we're achieving record free cash flow, profitability, and revenue in Q3. It's how we've achieved margin expansion, including being on track for six points of margin expansion this year alone. It's how we are absorbing the cost of high-impact M&A while still being on track for our long-term adjusted EBITDA margin target of 35%. We have built a formidable growth engine with this business, and we have now got the latitude to continue to make high ROI investments that perform for us and pay off for shareholders. This morning, we'll talk you through what this reinvention has looked like, how we've gotten here, and why we're positioned to win in this next exciting chapter of work. This is the most exciting chapter in Upwork's history. Because we've done this fundamental reinvention, we are positioned for this next era.
Importantly, we have traction on three major growth opportunity areas. These are our building blocks for growth that we've seen already and for so much runway we have ahead. We have a huge opportunity with AI tailwinds, and we'll talk about that growth building block. We have another amazing opportunity with the SMB market, and we'll unpack that. Of course, we know we can expand in the enterprise, and we have a strategy and product to do that as well. This is how we're achieving our accelerated metrics already in Q3, as you've seen, and we are poised for more acceleration ahead. This morning will reveal all of this and where we are going next. Now, I want to start by talking about the elephant in the room.
Every day, we all read headlines about how AI is changing work, and it's confusing because we read on the one hand that AI is taking jobs. Then we read that AI is disappointing. MIT recently published a study that said 95% of AI pilots have failed. You are probably wondering, is AI good or bad for Upwork? We ourselves have faced a stock market headwind from those who are not close to our story speculating about what AI does mean for this business. I can tell you, AI is definitively positive for us. In fact, it is one of our most exciting and growing tailwinds. The signals are totally clear. How do I know this? Our platform operates with tremendous scale. I look at the data every day, and we have 60,000 jobs posted a week. That's more than three million a year.
We get to see in real time how businesses are changing their needs with AI and how talent are responding. We can measure precisely the value that's being created or destroyed by AI, and we see this in our data. Yes, this is a huge, huge opportunity for talent on our platform who are now able to win more work because of AI, and we'll talk all about that. Let me also be clear. AI substitution is happening. AI is replacing humans in some parts of the labor market as a whole and on some parts of Upwork. We see this on our platform in small jobs, in categories like writing and translation. This is why jobs that are $300 or smaller on Upwork have actually gone from 5% of our GSV five years ago to now just 3.5% of GSV.
There has been some erosion in that part of our business, but this is a tiny fraction of what we do, and this is a tiny fraction. Small jobs make up a tiny part of the huge contingent market that we are going to talk about today. There is a bigger trend going on, bigger than substitution, and that is the creation of new demand for talent that is expert and highly skilled specifically because of this exciting AI technology. That is incredible news for talent on our platform. That is great news for Upwork and our shareholders because our superpower is matching talent with exactly the job that needs to be done and matching clients with exactly the talent they need in real time. The demand for that is absolutely growing. I want to unpack these tailwinds a little bit more in terms of AI.
are two trends going on under the hood that are creating this tailwind and benefiting our platform. The first one is the net new creation of demand because of AI work. Let me explain this. Many businesses, when they see the power of AI technology, it causes them to embark on projects they were not doing before, whether it is building a new product using AI or whether it is trying to do something that they did not even think was possible, but now AI makes possible for them and they are going to try that thing. That is all net new work they were not doing before, and their existing team does not have the capacity nor the skills to do this work. They come on Upwork and they find the talent that has the skills and the ability to help make these new dreams come into reality. There is another trend going on.
Companies that are trying to reinvent their existing workflows with AI are experiencing a fractionalization of full-time jobs. That happens when they look at an existing workflow, they bring in AI automation. There is still a residual amount of work that needs to be done by humans, but they realize this does not have the shape of a full-time employee. This work does not really need a full-time employee. It looks different. That is where their demand for contingent and freelance workers is also increasing because they know they need people to complement AI, but those people need to be different and they need to be hired differently. A great example of this is Grammarly, now known as Superhuman. They have been a customer of ours for years, and with the AI revolution that is going on, they set out to build a new product.
They wanted to have an AI writing assistant for their customers. They launched this, and they realized they could only create and deliver value with this product if they also had humans complementing what the AI technology could do. Now they have hundreds of Upwork writers and editors working alongside their AI tool as part of how they programmatically create value and deliver that to their customers. Upwork as a platform is benefiting as a result of these trends. We have seen our project sizes increase significantly. They're up 12% year over year just last quarter. We're seeing spend per client also increase because they're doing bigger, harder, more complex projects on our platform because of AI. Probably most exciting is the fact that we are seeing 50% growth in the number of clients on our platform that are engaging in AI projects.
Clients that are doing AI work spend three times as much as clients that are not doing AI work. This is a huge benefit for us that is just early. The momentum is still just building. We are early in this AI era, and so many businesses are still just starting to figure out how they transform or invent new things with this powerful technology. All the evidence is we're on track to unlock something substantial. When we talk to existing customers and prospects in the market, this is reaffirmed. 77% of hiring decision-makers tell us that putting AI into their businesses is going to increase their demand for fractional and freelance workers. This is great news, and we see it in the data. We see it building, and we are, again, early in this extremely exciting revolution. Others have been debating, what does AI mean?
What will its impact be? We have been building, building an entirely new category, the human and AI work marketplace that brings together people and technology in new ways to create new value for customers. Our marketplace is predicated on this new transformation of becoming a human and AI work marketplace. This is the heart of our business. It is the core of what we do. It's the foundation of our success, and it's how we create sustaining, durable competitive advantage in the market. This is why we've been able to deliver our strong Q3 results and why we're set up for further acceleration ahead. Dave Bottoms, our GM of Marketplace, will unpack this a lot more, but I want to pause and tell you a little bit more about this powerful flywheel that is really at the center of our business and creating so much value for us.
Our flywheel starts in the first part with our work delivery platform. This is the technology layer that we have built that enables clients and talent to work together online and takes out the risk for them. It basically de-risks hiring for clients and de-risks work for freelancers. It's tremendously valuable, and it's a closed ecosystem that goes end to end from hire to management, payment, reputation signals. All of that powers what we can do with this technology. Furthermore, it has been tuned for complex work, long-term engagements, not throwaway projects. We do the big stuff, the serious stuff, and that is really how customers show up and do big, long engagements with us that are getting bigger and longer with now the AI trends we're seeing. We also rebuilt this product to be fully AI native.
That has been the transformation over the last few years, and we'll talk more about that today. Because we have this powerful platform, it in turn attracts an incredible stable of clients to Upwork. We have 800,000 clients annually working on our platform. Everyone from Microsoft and Cloudflare to content creators and your local business, they can all come and experience the value of Upwork. Over the last few quarters, we have chosen to focus on client quality over quantity. We want clients with bigger budgets, more spend, better retention, work across categories. As we've done that, we have been very successful increasing the client quantity of kind of bigger, more successful clients that have real budgets and doing less with kind of in and out folks who are just going to be tire kickers. Our metrics reflect this, this progress on getting up market in the SMB space.
We see client spend increasing. We see clients that are spending $1,000 or more right out of the gate, increasing rapidly. Because we have such an exceptional pool of clients, that attracts talent to our platform. It's a huge talent magnet. We get talent from all over the world, and we now today operate the largest and most liquid talent pool on the planet. This is a talent pool of highly skilled workers, people with advanced degrees, PhDs, master's degrees who are so relevant in the AI era, helping train models, review outputs. I mean, there is so much work for these folks to do. They're highly skilled. They're using AI themselves in their workflows at a very high rate. Our talent pool is expanding in a different way. We are now adding AI agents to both the talent supply and the demand side of our equation.
Let me explain this. On the demand side, we already see large language models and AI agents hitting our website. They've been asked to complete a task by a client. They can't do it on their own, so they come to Upwork and they're trying to hire someone to actually get the work done. This is a new and growing demand opportunity that's very nascent, but very exciting and a signal for how much human work is still required, even as AI is proliferating. AI agents will soon be a new demand channel that's much more relevant on our platform, and we've rebuilt Upwork to accommodate that. There's the supply side. Of course, our talent are already using agents to complete and execute work, but we're taking it a step further. We have rebuilt Upwork to include AI agents as part of the work delivery flow.
This year, we launched a test experience putting AI agents into experiences where they could help complete work with clients. We learned a lot from that. Andrew Rabinovich, our CTO and head of AI and ML, will unpack some of those fascinating insights, but that work this year is what is the precursor to a launch in 2026 of AI agents broadly available to all clients in our marketplace. This is huge. We rebuilt the platform to accommodate this, and we are on the tipping point of launching those exciting experiences next year. Now, let me finish talking about the flywheel. We've got our demand, we've got our supply. Because we have all of these incredibly skilled individuals, that fosters tremendous category breadth on our platform. More than 130 categories of work, more than 10,000 skill areas growing all the time. Our clients use us holistically across their business.
We become their operational backbone. We hear this time and again from them because they're able to not just do one use case, but to spread out across multiple parts of their business. This is great for businesses, but it's also great for Upwork. We have superior economics as a result. We can pay more for high-value clients, profitably acquiring them and delivering still against incredible ROI thresholds because we can expand spend and serve businesses across many more categories than any of our competitors. This gives us an advantage versus vertical players and point solutions. Because we're horizontal, we have better economics, better client stickiness, better retention. Finally, all of this powers an incredible data asset. We all know in the era of AI, data is king. At Upwork, we have a data asset no one else has. No one's got this.
We have data on every step of work across 130 categories of work, how it's conceived, how it's delivered, what good looks like, what bad looks like. We're using this proprietary data asset, which is so valuable and which is growing every single day to unlock the power of our platform with our AI tools like our work agent, Uma. We'll talk a lot more about Uma and what Uma is, but do not overlook the power of this data asset that we are putting to work, putting in the pockets of our customers with our powerful AI technology so they get the benefit of all of the insights from our platform every step of their journey. Now, our customers experience Upwork's value and the flywheel in a very particular way. We see time and again. We call it the aha moment.
Customers come into Upwork and they have a mindset of scarcity. They feel like, "I can't get it done. I don't have the resources. I can't compete. I can't get the talent." They feel limited. They come into Upwork. They get to a mindset of abundance. Abundance. They realize they can do it. They have the talent access. They can move at the pace of their ambitions. This is a huge unlock, and this is the beginning of how customers then start to use Upwork more and more because they realize there's so much that they could do they didn't even dream they had access to. No one says it better than a customer, so let's hear from one of our customers about the value of Upwork. It's making it easier for us to launch. I think it's more around timing. It gives us confidence as well.
If we have an idea, we don't go, "Oh, that's expensive. That's too risky to do," because we tend to associate innovative app development projects with high costs and high failure rates. Now Upwork is helping us experiment ideas faster and fail faster if we need to or find success faster. It's having the confidence to launch low-fidelity solutions quickly and innovative ones too because you're accessing all this amazing talent. There was one applicant, and his role is an LLM developer. I've played all these new job titles. I've never heard of an LLM developer before. I wouldn't even know that that's what I was meant to hire. Just reading the portfolio, it taught me this is actually the birth of a new talent. There's a new breed of talent out there. Kaneitha says it so well.
This is the type of power that we provide every day for our customers, large and small. We know we can reach more of them. There are so many more SMBs and enterprise businesses that have yet to discover the power of Upwork and have that aha moment. Our strategy is designed precisely to do that, to unlock more value with SMBs and to grow more in the enterprise. Ernesto, our GM of Enterprise, will be up later talking about the enterprise strategy specifically. One of the things I love the most about this business is that at Upwork, we make things possible that are not possible anywhere else. Customers cannot replicate their experience and the power of our platform anywhere. Duolingo is a great example of this. Duolingo has been a customer of ours since their early days as a startup, and we have all watched their rise.
They have expanded languages. They've added new courses like chess and math and music. This has required massive workforce adaptation for them. For each new course they launch, they need different people building and training and deploying those solutions. Upwork is a solution for that. We have given them that adaptability and flexibility that their business model fundamentally requires, and we know they can't get that anywhere else. Another great example of this is Atlas. Atlas is a small business customer in our marketplace. They're building a financial services product that is focused on helping consumers avoid high-interest debt and have financial health. Atlas has built a team of over 150 CX and operations specialists on our platform, spanning many different functions, and they've told us. They've said, "We could not build this solution anywhere else." We rebuilt our business to provide this type of value for these customers.
We have been laser-focused on amplifying these strengths, and I want to talk to you about what this rebuild has looked like because it's been major. We have done a major business transformation in the last three years, kind of revealing it for the first time today, so get ready. There have been three major pieces to this. We reinvented the business on our product layer, on our customer focus, and on our operations, and we did it all at the same time, which has been huge. We're already seeing this reinvention working. It shows up in our Q3 numbers, and it shows up in our outlook we're going to share more about today. We began by reinventing our product. Early in 2023, we knew AI game changer. We knew we had to reimagine what Upwork looked like on the foundation of this powerful technology.
As our AI vision crystallized, we executed two high-impact acquisitions that accelerated our roadmap, and that pushed us to having so many of the AI-powered and Uma features and functionality today. This is already benefiting us. $100 million of incremental GSV this year is coming from the work that we've done rebuilding with AI and improving our customer experiences specifically through this strategy. It was not enough, though. We needed to do more. At the same time, we set out to reinvent who we were serving and serve them better because we knew there were big markets out there that we could grow even faster in. Specifically, we knew SMB and enterprise were big opportunities for us. We set out to serve them with new high-value products specifically built for those customer types.
At the end of last year, we launched Business Plus, which is an SMB-focused product specifically engineered to deliver on what those businesses need. Dave will talk more about that shortly. We also realized in the enterprise we could go further and compete better and bigger. We made our biggest moves in the company's history on enterprise, buying two companies this past year, relaunching our enterprise business as Lifted in August. The reception has been amazing. We've already heard from so many customers who are excited to use us in new and different ways and inviting us for bigger and better RFPs than what we were able to compete with before. The strategy is clearly working. We're seeing Business Plus. It's grown 33% on GSV quarter on quarter in Q3, so it's already got tremendous traction. This was also not enough, though. We also wanted to go further.
We reinvented our operational layer, fully went deep. We reduced our team size in 2024 by 21%. We brought AI into our workflows across the business. We created small, nimble teams that could move fast towards our huge ambitions, and this is working for us as well. We're moving at the speed of a startup and the scale of a market leader. This is a formidable combination. Why did we do this? We did this in pursuit of our giant market opportunity. I'm so excited to share these numbers with you because our market is huge, and we have just scratched the surface. We're seeing a $1.3 trillion amount of spend in 2028 on global digital knowledge work. It's a massive number. Specifically, the SMB market is $530 billion of that, and we are poised to open up that market with our Business Plus offering.
Enterprise, another huge opportunity, $650 billion of spend. We have engineered our Lifted subsidiary and our product directly to go after that huge market. Of course, there's some new stuff happening with AI agents. Estimates are for $120 billion of spend with AI agents by 2028. We are positioning to lead here too. We know we have incredible assets, incredible insights based on the work we've already done testing agents in our ecosystem, and we're building on that with the rollout of agents in the marketplace in 2026. Andrew will talk more about that solution, but make no mistake, we are positioning to pioneer that specific part of the market. We're opening up this $1.3 trillion market with a very specific strategy that has three building blocks. The first building block is AI. The second is SMBs, and the third is enterprise.
I want to unpack these a little bit. I'm going to start with AI because it's one of our biggest and most exciting tailwinds. Specifically within AI, we have multiple opportunities, and there's three that we're focused on. Expanding our AI-powered workflows is making our product more effective and AI-powered to drive throughput and attract the right businesses into Upwork. Second is accelerating our AI categories of growth, where on our platform, customers are coming in and hiring more and more AI talent. And then, of course, there's AI agents. Let me unpack the first one here, the AI-powered workflows. We have come so far with our AI agent that we call Uma. Let me rewind the clock. In 2024, we launched Uma, and I'd say at that time, Uma improved our product and our service, kind of like moving from a manual to an automatic transmission in your car.
2025, this past year, we expanded Uma's capabilities more. At this point, I'd say it's like fully self-driving mode in your Tesla. Can do a lot, but you want to have your hands kind of close to the wheel. 2026, where is Uma going? Uma will be driving the experience while you relax in the passenger seat. That's our vision for Uma. Uma will take your goals as a client, turn them into job requirements, turn them into posting, recruiting plans, recruit on your behalf, and project manage so you get an amazing outcome. All of this is powered by our data asset. That's how we're able to build such an incredible and differentiated tool. Our data gives us the ability to train an agent on so many types of work and so many steps in the work process that no one else has visibility into.
This is a strategy that's already having that incremental $100 million of impact on our GSV this year, and that's only calculated based on features we launched in the first half. Features we launched in Q3 and those that are already coming out in Q4 are not even counted into this number. We know we have a ton of runway here. Andrew and Dave will both talk more about this in their presentations. Let's talk about the second AI opportunity: AI category growth. I'm excited to share that today on Upwork, we already see $300 million of GSV coming from AI categories. That's a big number, and it's growing at more than 50% year over year. This is evidence of how much businesses are relying on Upwork for the key talent they need in this AI era. We are leaning in hard here. We are amplifying this opportunity.
In 2026, you'll see us do more with talent curation, with product that is tuned for these AI categories, and with marketing. There is a lot we can do to even thrust the gas further on this tailwind we are benefiting from today. This is a sustaining and durable opportunity because we are still in the early innings. Let's talk about the third opportunity we have in AI. I've got to say, in my six years leading this business, I'm not sure there has been anything I have been so excited about as this strategy right here. We have built the human and AI agent marketplace in a way that no one else is doing. Frankly, no one else can do what we're doing here. It's a tremendously exciting opportunity. Others are building agent marketplaces at best. Candidly, those fall short.
Agent marketplaces lack the human involvement that both creates customer trust to use the service and creates quality in the outcomes. We know that because we've done a lot of testing and iteration here. We're taking a different approach. We are building the human and AI work marketplace where humans and agents work together, co-creating outcomes that are better, faster than anything else that they could do individually as agents or even that humans could do individually. This year, we did launch a test experience with clients working with agents in our ecosystem, and next year, we will be rolling that out broadly to all customers. No one else is doing this. Let's zoom out. That was our AI building block. There's a lot there. Our second huge building block as a business is SMB. $530 billion of spend.
Up until a year ago, we did not have a tailored product just for these customers. We launched the first version in Q4 of 2024, and this year have expanded Business Plus and its capabilities substantially. We have heard loud and clear from customers who are using this product that this is now an indispensable growth engine for them. They can't do without it. We see tons of traction. Actually, in Q3, 36% quarter on quarter increase in clients joining Business Plus. This is a product that we have only put into the experience without marketing it, just letting people self-discover. We are early. Customers are flocking into it, and that is before we even press gas on marketing.
We just started a new marketing campaign in October, and this is the first time we're really bringing new customers actively into this product, even though they've been discovering it and joining it on their own. We can see we have real traction here. It is paying off. The metrics support it. Dave will talk a lot more about Business Plus today. You'll get to see more of what that entails. Our third building block, enterprise. Massive opportunity that we have been serving for more than a decade. We know these customers. We also realized that our prior solution was coming up short, and we made significant changes. We knew we could retool and change our product to actually meet the customer needs, which we deeply understand after years of working with these folks.
We made this huge move with two acquisitions of Ascen and Buddy into our Lifted business, which is now fully dedicated to opening up this giant market opportunity. This solution has never existed before. What we brought to market is completely unique in the ecosystem, and we're hearing incredible feedback from customers as a result. They've essentially been asking us, begging us, "Please do something like this." Now that we've done it, they're like, "Yes, let's go." We know we have a lot of runway here. It's extremely exciting. We're seeing what the pipeline looks like, and we are excited about how this is really going to perform starting in the back half of 2026 and just continue to take off from there. These building blocks are specifically why we feel so confident in sharing new long-term targets with you all today.
We are expecting GSV 7%-9% CAGR over the next three years, which is awesome. We are seeing the acceleration now, and we know it can go further. On revenue, we have confidence in 13%-15% CAGR in the next three years. We know this is possible based on everything we're seeing and doing right now. Margin expansion. We are phenomenal at this. We operate a business with incredible discipline, incredible focus. We know we can achieve 20% margin expansion CAGR over the next three years as well. All of this will be detailed in more with Erica, our CFO, in a few minutes. She'll unpack these numbers, why we are so confident in our ability to reach them. It's clear. Upwork is a different company today than it was just a few years ago. We have rebuilt it to win in the AI era.
We are accelerating the business through amplifying our huge AI tailwinds, which are here today and are growing. We are also running down the runway of SMB and enterprise expansion. We see the traction. We know the market is there, and we are ready, and we're executing. We also have the financial prowess to continue to drive top and bottom line performance with our strategies. We know we can do this. It's already started in Q3, and we'll continue to accelerate. This is the most exciting chapter in Upwork's history. We've rebuilt for this moment with new capabilities and new products. We're positioned to take advantage of our incredible unique assets to continue to win and share in the market. We're poised to deliver incredible financial returns, and we know we can win. This is the time.
We are building a generation-defining business, and we are really excited that you're here joining us on that journey. Thank you. I now want to invite my amazing colleague, Dave Bottoms, up here who runs our marketplace. Come on up, Dave.
Thank you, Hayden, and good morning, everyone. I'm Dave Bottoms, the GM of the world's largest human and AI-powered marketplace. A little background on me. I've spent 25 years in Silicon Valley leading product and growth teams at some iconic companies behind major technology shifts: Netscape, Yahoo, Dropbox, and Meta. The next one's Upwork. I joined the company in 2022 because I deeply believe in what Upwork stands for: creating economic opportunities through technology, connecting businesses and people at global scale, and shaping the future of work. Today, I'll share some highlights from our journey and where we see Upwork going in the AI era. First, I'll take you on a tour of how we've transformed our platform into an AI-native experience, as Hayden teed up, how this is also transforming our core business.
I'll share a little bit more about why we believe we're uniquely positioned to grow reach among SMBs. First, let's get grounded in the numbers, the breadth and the depth of our marketplace. We're already the world's largest human and AI marketplace with 18 million active professionals from more than 180 countries. We offer 10,000 skills across 130 categories of work, and we have more than 250,000 AI experts on the platform with skills ranging from software development to data science to prompt engineering to traditional categories like design and marketing and customer support. The diversity of work that's happening on our platform has given us a front-row seat to every major technology shift, and right now, that shift is in demand for AI-savvy talent across these categories that I just mentioned. Professionals are not just learning new tools.
They're applying AI to real-world work that's happening across industry verticals and concrete use cases. This is helping businesses of all sizes scale in real time. Today, Upwork is a very vibrant platform that has weathered post-pandemic economic cycles. We've had the foresight to make critical product and technology investments to drive durable long-term growth, and we have essentially rebuilt, not retrofitted, almost every critical workflow and every meaningful customer touchpoint on the platform. This is from discovery at the top of our funnel through to work delivery. AI has changed how our customers post jobs, find talent, and collaborate together to deliver high-impact work. This reinvention of our core product experience cannot be overstated. I'll say that again. This cannot be overstated. This is a massive amount of work over the last three years, and I'm super proud of the teams that have built it.
By the end of the first half of 2026, virtually every step of the customer journey will be enhanced or orchestrated by Uma, Upwork's mindful AI. Let's take a look at what that transformation feels like for our customers.
Every small business owner dreams about their future. Upwork turns these dreams into reality. Getting started is fast, easy, and intuitive. Share what you need, and Uma, Upwork's mindful AI, helps you create the right job post. Find the best independent talent to get the job done. Soon, you'll be able to simply discuss your project needs and goals with Uma. Ask questions and get answers. Uma Recruiter puts together a talent shortlist and reviews proposals. As your full-service recruiter, Uma helps identify the right questions to ask. Once you've narrowed down your top candidates, Uma sets up AI video interviews and takes notes. Upwork video meetings make hiring a breeze. You connect with potential talent while Uma helps draft an offer. Uma stays with you from inception to delivery to provide the high-quality work outcome your business deserves. Humans and AI, it just works when you Upwork it.
I could watch that video another 20 times, and I still get excited because of all the things that we've put into the marketplace in this transformation. What you just saw really quickly highlights just a few of the features and capabilities that are transforming our marketplace into an AI-native experience. We are systematically removing friction between clients and talent so they can get on with the really important stuff, which is getting work done efficiently and easily. Today, we introduced Uma across every step of the end-to-end customer journey from the top of the funnel, starting on our homepage through to new client and talent onboarding. We've reimagined Job Post, and we have one of the, actually, one of the interesting things about Job Post that I wanted to mention. 70% of Job Post are now being powered by Uma on our platform. 70%.
Uma now guides talent through writing proposals and submitting proposals. Our Uma companion facilitates messaging between clients and talent and outlines project milestones. Uma can even draft and send job offers on a client's behalf. One notable release that I'm extremely proud of is AI-generated work summaries, which help clients compare freelancers based on their work history on the platform. We've introduced Uma-led videos, video interviews, and we're automating project milestone tracking in our work diary. We've introduced video collaboration, real-time video collaboration, powered by the tech from our Headroom acquisition. We've dramatically improved the relevance of our AI-powered search, modernized the entire tech stack, and accelerated through our acquisition of Objective.
As showcased in the video, we just started beta testing one of our most ambitious product upgrades to date, conversational search, powered by Uma, where clients simply express their desired outcome, and Uma bypasses the job flow to recommend a best match of candidates to compare. Again, reducing friction and accelerating time to hire. Today, direct hiring, such as through conversational search, happens through a combination of search and messaging. It makes up nearly 40% of hiring on the platform, and it yields a higher fill rate than the traditional Job Post flow. We expect the adoption of this beta of conversational search will actually increase as we roll out next-generation conversational search experience.
Looking a little further out into 2026, as Hayden alluded, we'll continue transforming the post-match experience, introducing the Uma Project Manager to orchestrate delivery of more high-value, complex work and keeping cross-functional teams aligned and on track. Today, our AI-native marketplace is faster, more intelligent, and resonates with both sides of the marketplace, clients and talent. Many of these capabilities are still early, and the combination is already starting to bend the curve of our business. I'm incredibly proud to say that these innovations are estimated to unlock over $100 million in incremental GSV in 2025, and we're really just getting started. We're investing in these AI-native experiences because the core product experience is one of the key drivers behind the new Upwork. All right, let's zoom out a little bit. Hayden introduced this concept that AI is a tailwind.
As she teed up in her part of the talk, I'm going to provide a few more details. First, there are net new categories of AI work that are emerging on the platform, creating a strong demand signal for our platform today. Second, full-time work is being fractionalized. As tools from AI workflows and agentic workflows are embraced, parts of people's jobs may be getting automated, but there's still a need for humans in the loop to guide delivery and ensure quality output. Demand for human expertise is fueling both of these scenarios and driving both of these tailwinds. I love this chart. We're going to unpack this. These trends are starting to unfold right before our eyes. Net new AI work is one of the fastest growing categories on Upwork.
We're seeing real sustained job creation happening across industries and job types. As Hayden mentioned, GSV from work on the marketplace grew more than 50% year over year in Q3. This work spans technical categories like software development, prompt engineering, data science, but also other traditional categories like marketing and design, as well as many subcategories like AI applications and machine learning. What you see in this chart is a long time horizon, 40 quarters, right? 40 quarters. We're seeing a traditional category like AI and machine learning, it's been around about 10 years, accelerating over the last 10 quarters. That's interesting. What's even more interesting is net new AI categories like AI apps and integration. Only about 8-10 quarters has existed on the platform, and we're already seeing it outpace some of our traditional categories of work that have emerged over time.
This chart also illustrates why we're uniquely positioned to capitalize on early trend spotting. We have more of this historical data on work demand than any other platform on the planet. All right, now that we've touched on net new AI work, let's dive into the second of the emerging categories, fractionalization. As companies integrate AI into their operations, entire workflows are being redefined. Some tasks can be automated, but new ones are emerging that require human oversight, expertise, and interpretation. In fact, to support AI implementations, there was that stat that you saw earlier, 77% of companies expect demand for fractional labor to increase, 77%. What happens when parts of full-time jobs get automated? The remaining human work becomes more specialized. Think designing prompts or training models or validating work output.
Full-time employees do not always have the requisite skills, but automation still requires skilled experts who can validate the work outputs. In the human and AI world, work is becoming more modular, more distributed, and outcome-driven, fully aligned with the fractional operating model that Upwork has pioneered. Our marketplace already supports flexible project-based collaboration at global scale. As AI accelerates the fractionalization of work, businesses will increasingly turn to Upwork to fill these specialized roles reliably and quickly. Unlike businesses that are large, that may have capacity to absorb and support AI-driven workflows, chances are SMBs do not have a lot of this expertise in-house. They may not have the specialized talent that they need for this era of AI work.
They increasingly are turning to Upwork to access on-demand fractional expertise from software engineers and data scientists to marketers and designers, allowing them to scale capabilities and stay competitive without adding full-time headcount. This is why demand for net new AI work, combined with the 250,000 AI talent that we have on the platform, makes Upwork the premier destination for AI talent and a catalyst for growth of our marketplace. Let's further unpack this SMB opportunity. Today, SMBs account for nearly half of global GDP and the majority of job creation. Over the next three years, the market opportunity for contingent digital knowledge work, that's a mouthful, contingent digital knowledge work, is expected to reach $530 billion by 2028. As the backbone of the global economy, SMBs already represent a major economic driver on Upwork today.
They come to our marketplace to find fractional expertise to do this high-value work. They spend more. The average GSV spend per client is $12,000. They hire more often and across more categories, the average of five contracts across multiple categories. They have higher net promoter scores. This is a trend that we have seen continue in Q3, which leads to stronger client retention and repeat spend on the platform. As Hayden shared, we believe that expanding our reach to serve larger, higher-spending SMBs is one of the key drivers to unlock growth. We are being very intentional about moving up the value chain to expand our SMB footprint, and our strategy starts with a very simple mantra, "Know thy customer." We have gotten to know SMB customers over many, many years. They have been on our platform.
Recently, we've done many, many rounds of qualitative and quantitative research and customer round tables, and we've built a very deep understanding of this segment and what their top pain points are. They want access to high-quality talent with reputation that is earned on the platform. They want to hire and centrally manage multiple people to create virtual teams. They want their hiring teams to be able to share pipelines and share tools so that the process becomes easier and faster. They want flexibility to manage their budgets to preserve working capital of their business. Our directive is very clear at Upwork. If we help small businesses solve their problems and stay competitive in today's market, we will grow in the segment. That's why we've developed Business Plus. It's a premium offering that was designed to super serve growing SMBs.
These are businesses that need advanced tools and capabilities to allow them to hire faster, collaborate more effectively, and scale with confidence. At the end of October, we introduced an exciting new bundle of features in Business Plus, each directly mapped to what SMBs need to be successful in today's environment. Business Plus gives SMBs access to the top 1% of talent, expert-vetted through human review and AI performance data to ensure it's the highest quality. Uma Recruiter, this is a new capability we built just for Business Plus. It's exclusively an AI-powered sourcing and shortlisting agent that analyzes millions of successful engagements and recommends the best match for each role. This reduces median shortlisting time by 5x and increases satisfaction with the match. Because, as I alluded, hiring is often a team sport, we've enabled team-based hiring.
This allows for shared access, role-based permissions, and team-oriented workflows to allow multiple stakeholders across teams and functions to evaluate candidates and make hiring decisions faster together. Another highlight is payment protection and net 30 payment terms. This reduces risk and offers flexible payment terms so businesses can manage their cash flow. That is not all. Business Plus has weekly work summaries, more advanced reporting, 24/7 customer support, even access to our APIs. This premium offering moves SMBs from searching for the right talent to actually delivering on their business goals faster. As these SMBs build confidence to execute higher-value, more complex work, Upwork allows them to efficiently and cost-effectively scale their business along with their ambitions. Let's throw some numbers.
Since the initial MVP, or minimum viable product, as we call it in Silicon Valley, was introduced in Q3 of last year, Business Plus has shown strong signs of product-market fit. This is largely driven by organic demand and an efficient product-led growth upgrade motion for existing SMBs on the platform. In Q3, we saw total active Business Plus clients rise 36% quarter over quarter. What's more, Business Plus clients hire 15% faster, Business Plus clients spend two and a half times more than the average marketplace client, and new Business Plus clients retain at a higher rate. The result is 33% quarter over quarter GSV growth of Business Plus. That is what durable, predictable growth looks like. We have a massive serviceable market opportunity. We have a discreet customer segment that we know and understand very well.
We have a differentiated product that will address their needs, and we have a compounding upsell and retain motion that scales efficiently over time. Now that we've demonstrated product-market fit with existing SMBs on the platform, we're dialing up our go-to-market plans for Business Plus, designed to amplify what's already working with our product-led growth motion. We've launched a new acquisition campaign at the end of October that calls on small business leaders to stop doing everything. We're acknowledging that many small business owners are wearing many hats, and they can come to Upwork to find the help and the talent that they need. We will help them grow. It's designed to meet our customers where they are. The campaign connects their existing pain points directly with the needs that we solve with Business Plus.
If marketing's job is really to drive top-of-funnel awareness and conversion, our revamped client onboarding experience is working harder to convert and register and activate and retain these users. In short, our marketing and our product are working together in concert, and we're really just getting started in marketing Business Plus to small business customers. Now, I'm excited to share one of our actual Business Plus customers, Brent Schuck, who's the Director of Marketing at Demco, on how we're supporting the growth of their business.
Demco is a fourth-generation owned family business. We produce products for the agriculture equipment space, construction equipment space, RV, trailer components, OEM provider to a lot of the brands people recognize. For our organization, Business Plus allows us to scale our teams up as we need, when we need. First thing that got my attention was that curated shortlist. That was a major pain point.
Secondly, the billing. With our accounting software processes, it works out much better to have a net 30 invoice versus having weekly credit card charges. As we looked at an organization where the scope of Upwork was expanding beyond a single department, just the organization and being able to reduce all of that internal overhead on processing the management side, that was a big plus. We're using it for some 3D animation and rendering support for a new product that we're launching. Maybe the last 10-12 months, we've started to lean more where it's maybe less predefined and tends to be more of an ongoing support effort. We've become more focused and effective in what we do.
As we have needs that arise, whether that's a skills gap, lack of bandwidth, where we need some additional help to get initiatives pushed across the finish line, our first thought is Upwork.
Thank you, Brent. As I said at the outset, 2025 has been a truly transformational year for the marketplace. Our efforts to reimagine our product experience have started to drive real business impact. Expected to unlock more than $100 million in incremental GSV in 2025. Next year, we'll continue the transformation with a focus on post-match work delivery.
With Uma continuing to evolve into project management and matching talented experts with agents, more to come on that from Andrew shortly, we will also continue trend spotting, looking for new or emerging categories of work on the platform, new and emerging AI categories, to ensure that when we see the demand, that we know how to capture it. Looking a little further out into 2026 and 2027, Uma will become even more advanced in orchestrating work delivery, moving from a companion to an orchestration agent. We'll continue packing value-added services into premium tiers like Business Plus, solving more complex work problems from larger customers seeking economies of scale. In short, there's a lot to be excited about about the marketplace business. I know I'm excited. I'm also excited about the panel that we have coming up. So I'll quickly summarize.
If you get just three main things out of my talk this morning, this is it. First, we have already reimagined the marketplace as AI native. We have rebuilt and relooked at the end-to-end customer journey, and it is now powered by Uma. Much of the experience has already been transformed, and we will finish the swing in the first half of next year. Second, we are uniquely positioned behind two AI tailwinds, the demand for net new work and fractionalization, as full-time work gives way to AI-powered automation with humans in the loop. Third, make no mistake, we are being explicit and intentional about super serving our small business customers. We will expand our footprint and meet the market with Business Plus, the product that has been purpose-built to help our customers transform and grow their business.
I trust this gives you a bit of a sense of where we are in the marketplace and where we're headed. No presentation and no video will beat hearing from our customers, the ones that we aim directly to serve. I'm very excited about this next segment. We've invited three Upwork customers to join us on stage in a panel that's moderated by Peter Sanborn, our VP of Corporate Development, Strategy, and Partnerships. Did I miss anything? No. Okay. Peter, please join us up here for the panel.
Hi. I'm Peter Sanborn, and I lead strategy, corporate development, and partnerships here at Upwork. You've just heard from Dave and Hayden how the new Upwork is rebuilt for the human and AI era and how SMBs are driving that transformation. There's nothing like hearing firsthand from real clients doing real work on the Upwork platform.
I'm joined here by three SMB leaders across the biotechnology, the financial services, and the tech sector. They're all using Upwork to accelerate innovation, tackle complex challenges, and unlock new capabilities for their businesses. These three leaders, they're also using AI to grow faster, to scale flexibly, and they're choosing Upwork as the critical infrastructure to do so. With that, let's get to know them. Gabe, there are two things that fascinate me about your business. First is just the problem that you're tackling. The second is how you've built a startup from the ground up with flexible talent. Tell us a little bit about OMIC.
Absolutely. I'm the founder and CEO of OMIC, and we're building biological superintelligence to end disease. Our platform completely reimagines how we think about diseases and how we engineer therapies to make them better.
We typically tackle problems that are too challenging or even impossible for traditional biotech and pharma. From the beginning, we knew that the traditional VC model, raising tens of millions and hiring a big in-house team, was not going to work for us. Without Upwork, OMIC would not be possible. It is how we have been able to attract world-class talent from genomic scientists to chemists to bioinformaticists to even our Chief Scientific Officer that we were able to attract from Upwork. Today, we have 15 contractors with Upwork, and they have become a core part of our research and development workflow.
Awesome. Phil, you operate in the highly regulated financial services industry. You have also been working with Upwork for over a decade across many areas of work. Tell us a little bit about MCT and what you are up to.
Yeah. MCT is not as noble as Gabe's business of ending the diseases.
Yeah, we operate in a niche of mortgage capital markets where we hedge interest rate risk for banks and mortgage companies. We started with Upwork in 2010, almost 16 years ago now. Back then, I had five or six employees. It was essential because we were running a business, growing a business, and I couldn't work with software developers during the day. I needed to kind of follow the sun. It was essential back then. It's been essential all the way up. Now we have about 115 employees, and we keep 25-30 Upwork contractors, including a core group of 15 developers out in Europe that are very important to our software development process.
Corey, you're a product leader at the forefront of helping companies modernize software with AI. Tell us a little bit about CoreStory and where Upwork fits in.
Yeah.
I'm Corey, the VP of Product and Research at CoreStory. What we do is we build large models that are able to understand huge legacy code bases that enterprises use to either modernize or maintain or migrate. For us, we are a smaller fish in a huge market, which AI is today. Being able to leverage Upwork for scale and bringing in really acute talent for certain problems has been huge.
Let's zoom out a little bit. Let's talk about flexible talent, flexible work. Is it really that important to you versus full-time hiring? Does it actually move the needle? Gabe, maybe let's start with you.
Sure. For us, it completely transformed the way that we do business. It removed the entire barrier to experimentation for us. I'll give you an example.
I had a crazy idea that we could re-engineer part of our platform where in traditional drug discovery, it takes hundreds of thousands of dollars and weeks running on supercomputers. I thought there was a chance we could do that in just a few minutes at near zero cost. I had two problems. I was not sure that it was going to work at all. It required a developer or developers with very specialized skill sets. I went on Upwork and I hired a PhD-level CUDA developer and just threw the problem at him and said, "I want you to tackle this." With a lot of work, he actually was able to solve it. Now this wild idea that I had that I was not sure was going to work has become a core part of our platform today.
As that process has repeated itself, I've found myself turning more and more to Upwork for really specialized, differentiated talent.
Are these short-term one-off projects? Are they longer-term? Tell me about that.
These are almost exclusively long-term projects. For our type of business, it takes a lot of time for people to get up to speed and understand what it is even we're doing. We almost exclusively work on long-term contracts with Upwork.
Phil, you've been working with Upwork across all these eras of work. I'm curious, AI, how's it transforming in your mind? How's that evolving this flexible talent need?
I feel pretty strongly about this, that this is the future for two reasons. I mean, number one, being a small business owner in California, I can tell you the return policy on employees, full-time employees is not great, right?
You need to be able to be flexible to, as Gabe and Corey deal with a lot, new experiments, start them, and turn off that resource if needed. Additionally, something I did not mention to you, Peter, was on a firm like ours, we do not capitalize a lot of projects, but now we are starting to capitalize more AI projects, right? Because these are new projects. If you think of my internal development staff, let's say I have 15 people, let's say it is about $2 million a year in cost. My CFO is not going to allow me to just take 40% of that $2 million and capitalize. It does not work like that, right? We have audited financials. It is a lot easier for me to capitalize that contract work and/or vendor work. We look at Upwork as a vendor. We can capitalize that direct spend.
It's a lot easier from an accounting standpoint.
We just heard Dave and Hayden reference the 77% of business leaders looking to increase flexible talent with the use of AI. Are you seeing that? Do you think kind of this fractionalization point is actually true?
Yeah. I think that's going to continue to grow. I think that you're probably going to see that in the future, as Dave was mentioning, that you're going to see that potential reduction in full-time staff.
Flexible talent, it's clearly important to all of you. I'd love to hear why Upwork. We know you have other options. There's other folks out there. Corey, what are your thoughts?
Yeah. For us, it kind of really has been three big things. First of all is the scale.
A lot of times when we're using Upwork, we're looking to run very high-volume studies quickly. Most recently, we ran a big study of around 300 people. We were able to bring all of them on in about a month or so, a month and a half, which in the academic community, that is unheard of speed. The second is really around the customer success because during that study, trying to manage 300 people, I mean, you think about it, we got 300 people, but that means that traditionally, you'd have to interview maybe 4X that to try to come down. The customer success team at Upwork helped us a lot through that.
Finally, the platform being kind of product-led, being really extremely easy to use for us has let us scale adoption of using flex talent, not just within my team, but kind of within the rest of the business, being able to know you can just kind of tap in, tap out, whatever you would like to do.
Gabe, what about you? Why Upwork?
Sure. For us, it's the breadth and depth of talent. As a small startup working across a number of different domains, we often need to hire people that have expertise in multiple areas. They need to understand biology and AI or chemistry and physics or deep computing infrastructure and data science. Upwork is the place where we've been able to consistently find that diverse level of expertise. That's why we keep coming back.
What about you, Phil?
Why do you keep coming back after decades here? I'm probably not the best person to ask because it's kind of all I've ever known, right? We've always used Upwork, been really valuable to us. I agree that the large talent pool and just what Gabe said about specialized skills, we're starting to see better and better talent on the platform.
Hayden and Dave also just talked about the new Upwork. What's your take? Is there anything to the platform, the technology, or is it really just about the talent for you? Gabe, what are your thoughts on this?
Upwork has actually become a core part of our business operations. We have certainly found that we're trying to recruit and manage this diverse set of talent as a small team.
We need to be able to attract, hire, retain, pay all of these people, and manage their time. For us, Upwork has been sort of the backbone for us. In addition, we've seen the platform grow and evolve over time. More recently, we've used Uma, Uma Recruiter, the Business Plus services that give us access to this whole new level of talent. That means that we can take that friction and the time and expense associated with recruiting and hiring and managing these people and reinvest it back into the core of our business, which is solving diseases like their engineering problems.
That's a great insight. Phil, you've been hiring and managing flexible talent internationally. You have some interesting use cases on how you've been leveraging the platform or bringing others onto the platform. Why don't you tell us about that?
Yeah.
In 2011, we hired a core, really good guy out in Croatia. He's been with us ever since. Actually, I've hired his brother. They were all in the good universities out there, and they had a lot of friends. We kept hiring their friends. Through this guy, we've hired about 20 contractors. Fourteen of them have been with us now almost a decade. It makes a lot of sense. We've never paid an employee overseas. I don't want to get into that, especially having audited financials. We just basically bring them onto the Upwork platform. I think that really shows the value of the platform to us. Obviously, there's an extra expense than directly paying the employee, but the ease of use is a no-brainer for us.
Finding employees on Upwork, also finding others, and then bringing them back onto the platform to manage everything holistically.
Yeah, literally. We have basically a sheet of how they start up, whatever you call it, a profile on Upwork. Then they'll shoot us a link, and we'll get them signed up.
Corey, let's bring it to you. 300 people, this research project. How did you get that done in the platform? How did that kind of enable you to move so quickly?
Yeah, of course. Like I said before, bringing on 300 people for an academic study and research study like we did is unheard of. We were under a lot of time pressure because this was a partnership with Microsoft and GitHub to present the paper at a conference.
Basically reaching out and saying, "Here's the kind of talent levels that we're looking for because we're looking to measure the impact of AI coding assistance on engineers." We needed a wide range of everyone that was kind of novice to beginner engineers all the way up to super advanced. We kind of submitted the profile, submitted what we were looking for. The customer success team at Upwork helped us kind of organize and manage and source and onboard and offboard all of them.
Post the hire, I think I've heard you talk about this a little bit. How long does that actually take?
To find them? Yeah. I mean, the big thing that we've seen a huge benefit in is that whenever we go to look for participants within the Upwork network, it's usually between 15-20 minutes.
We'll know this is probably a good candidate worth moving forward with. Through traditional hiring means this would be substantially longer. That velocity out of the gate is something that lets us be more ambitious with what we look to research and what we look to build.
Yeah, I think we've all been there trying to hire that one person. It can take months. The fact that you're able to do 300 in such a short time is pretty incredible. Let's zoom out again. Let's talk a little bit about AI. We've all seen the studies about questions on the return on investment, especially for SMBs. Each of you is powering the economy in a different way. I'm curious, is there real tangible value in AI that you're actually seeing? Is it transforming work? Corey, maybe start with you. You're leading product and research.
How do you see product, research, engineering kind of coming together?
Yeah, of course. I mean, our big philosophy and hypothesis that we operate on is that every company is moving towards being a research-led company, especially in the age and time of AI, meaning that you need to learn when and where to double down on. Having a facilitator and a means that if someone within your organization has an idea that you can quickly validate and get data that that idea has legs, the more competitive that you'll be in the market. For us, leveraging Upwork has let us kind of really be that strong research-led example. We think that's going to continue to kind of proliferate through the rest of the market.
Do you think that means you're kind of commoditizing the traditional engineering-led advantages?
Yeah.
I mean, the other big thing that we operate under is that the hypothesis that the cost of code is going to zero, that with AI, that the typical barriers of larger businesses or others of high capital expenditure to write code was their moat is kind of now disintegrating beneath their feet. Really what that means next is that the big differentiator is going to be who has really good ideas and who can execute and gain market share the fastest will lead you to be market leaders. For us, being able to use Upwork to, again, to find an idea, double down, get data that that idea has legs, and then move forward. Or if it does not, then to not move forward with it is extremely important.
Really leveling the playing field and kind of having the research piece be a driver of kind of novel and new ideas.
Exactly. Exactly.
Very cool. Gabe, you talked about AI creating a new breed of founders, and you're kind of walking this walk. Tell us a little bit about that.
Yeah. For us, we believe that the idea that only great technical innovations can come from a small handful of zip codes is just evaporating. We've certainly found that brilliant people exist everywhere. They can do great things everywhere, and it's not location-dependent. For our team, we've also seen the power of a great platform combined with AI. We've seen one PhD scientist on the OMIC platform be able to accomplish things that used to take entire teams or departments of people and huge budgets.
Upwork is enabling just this type of founder to exist. I'm sure there are millions of others that are just like me with small teams and big ideas taking on really challenging problems.
Phil, you just mentioned AI is increasing the need for flexible talent. Is this just a phase in your mind, or are you actually seeing kind of this work unlocking new opportunities for your business?
Yeah. Going back to what Corey was saying, Corey and I were talking earlier that he's on a little bit more of the frontier of AI development. Down in the trenches of getting the financial software developers into the new generation of AI development is really like pulling teeth.
With the Upwork staff, it has been helpful bringing on new developers quickly that are using new tools like Cursor that are just changing the game of software development. It really kind of enables—you use the term commoditize. It does kind of commoditize that skill set where we can then kind of force our internal developers to say, "Hey, you guys have to step up here. Otherwise, we're going to move on."
You were just mentioning to me in the hallway, actually, AI is also creating kind of new challenges or new opportunities for work. I think you had a problem that came up on the math side, and you had to bring someone in. Tell us a little bit about that.
Yeah. We had a constrained optimization problem for pooling, and we brought in a PhD in mathematics on Upwork.
That's what I mentioned earlier. I do see the talent. I feel like the talent's getting better on the platform. I think just what we're all seeing in the news, that as larger firms are shedding full-time staff, I would imagine, just my opinion, that Upwork's going to see better talent on the
platform. With that, let's wrap. This is what the future looks like. Bold ideas, human expertise, and AI all coming together to unlock new possibilities. There's a through line to these stories: scale, speed, flexibility, access to expert talent. Whether it's reimagining drug discovery, transforming legacy software, or transforming the financial services industry, all of these leaders are using Upwork not just to hire and manage, but to build the future of their businesses. Thanks so much for an incredible discussion. Really enjoyed having you on stage here.
We're going to take a quick break, 15 minutes, and then we'll return back to hear from Andrew Rabinovich, our CTO and head of AI.
We will now. My name is David Wrench. I'm a co-founder and CEO at Data Joy. Data Joy tells data stories on top of your data using AI. What that means for customers is anybody in the SMB or individual professional space dealing with data, reporting, and analytics, they're able to use AI to tell stories on their data rather than building out entire data systems, dashboarding, or analytical tools. It was a COVID startup, really. It was me against the world. I was burnt out. I got out of my six-figure job, and I was like, "No.
I deserve better, and I'm going to build a team and give them better as well. I knew I needed to use Upwork when I realized I was struggling and I was going to fail without help. When I found it, I definitely realized I found it because it was that much of a difference in the day-to-day of success and progress that we were able to make that it was undeniable. Demetrio is serving as a co-founder and CTO. Demetrio is the ice to my fire. He saw that I was struggling, and he's like, "I see what you're trying to do, and I can help you do that." A lot of what Demetrio has been able to do for us is our tempo of deliverables has gone from months in between to weeks in between to now we have daily, weekly bills or whatever we need.
He's capable of doing that. Demetrio is not the first best person I've met on Upwork. He's just the most recent best person that's still going and I hope to continue to work with. We were hiring software engineers, some web developers, AI-forward developers. Upwork is just a platform full of talented individuals. There's a set of talent that's deep and broad that I can pull from almost instantly. I can't do that at all the tech events and meetups and networking around town. I can make friends and make relationships, make introductions, but on Upwork, I can make things happen. Upwork has helped me deliver on all my technical gaps.
If it's a marketing, if it's a research thing, if it's a design thing, documentation, automation, I know I can find somebody that is compliant, skilled, and available as well so that I know I can hit my timelines and communicate with them as needed. I don't need people to just sit down and do the work. I need people that want it. I need somebody that gets it, wants it, and commits to it. That is what our team does. Half of that team is from Upwork.
What our company does is what we call physical marketing or branding and swagging. Branding is when somebody comes to us and says, "I want to put my logo on this item specifically." Swagging is more on the marketing side of promotional products. Let's talk about what your goals are.
Is it to get more people to your website? Is it name recognition? Here are two or three other ideas that maybe achieve your marketing goal. There is lots of work that you do in an average day that you are like, "Oh, if I did not have to do that or if my staff did not have to do this, we could increase our productivity. We could add a new feature. We could drop a new launch." We went on Upwork. I posted a job, and we found Jerry, who is in the Philippines, and he has been with us ever since. He is fantastic. He does such great work. That kind of led us down the path to hiring in inside sales and production support and stuff like that because once you see the value and you see the working relationship you can have with somebody that is global, there is no stopping you.
Every problem I come up with, I go, "Well, could we just hire somebody and hire somebody from Upwork that knows what they're doing to do that?" The answer is usually yes. Why hire the best graphic designer that can drive to my house in Minneapolis when I can have the best graphic designer in the world? Upwork takes care of all the messy stuff, painting, whatever the taxes situation is. You do not have to feel like there is all this paperwork and legal obligations because Upwork is taking care of all that stuff for you. The hiring process is just super easy. There is a ton of candidates. Whenever we post a job, we are getting a dozen, two dozen qualified candidates. To me, I am not even looking at other options because why would I when Upwork just works very smoothly?
I posted these three jobs on Friday and have the people hired starting to work already this week. The work that we're doing with Upwork, if I'm a small business owner, it's a no-brainer. It's a 100% no-brainer.
Hi, my name is Jacqueline Tangora. I'm the CEO and founder of Omni Business Intelligence Solutions, and I'm also a co-founder of Data Ops. I had looked to the platform as more of a supplementary income stream, and I had absolutely no idea what it would evolve into. The day I published my profile, I got my first offer, my first contract. I was just enamored with the ability to take matters into my own hands and generate income on my own time. Leaving PwC was probably one of the hardest decisions I made in my professional career.
Upwork really provided me with confidence where I knew that I wasn't just leaping blindly into the unknown. It's just been incredible ever since. We've built some extremely wonderful, trusted relationships with businesses on the platform, both at the marketplace level and at the enterprise level. It's always fun to say we work with some pretty big brands on the platform as well. What gives me confidence operating on the Upwork platform really is, I mean, I see how serious they are about security, about protocols, about making sure there's no scams, about reinvesting into the governance of the platform itself that allows us to feel like we're being protected. I started using Uma recently, and for me, it was just really awesome to understand how Upwork is using AI to connect us to the right opportunities.
I look at the product itself, and I don't just see new features. I actually see something deeper than that. It's really the beginning of an intelligence layer that's being laid within the company itself. We are using AI in various different ways. We're using it to help us with cutting down our lead time for proposal writing. I mean, I use it for competitive research. I use it for benchmarking. I have a team that has built a SDR, so Sales Development Representative Agent, that's helped us capture leads in the market as well. To be able to be sleeping and to find out that someone's interested in talking to us in the middle of the night is pretty incredible to wake up to a qualified lead and an opportunity.
We've built a proprietary AI solution for a franchise, a national healthcare franchise, and now we've worked with so many different territories in that particular space. It's incredible because the time to market to get that solution available and the cost to do it is ridiculous. To be able to have the ability to get a prototype and get something developed in four to eight weeks and to show them what we're doing and how we have the vision for them. I wouldn't have been able to do that had I not been able to find people on the Upwork platform that understand AI, that can use AI to code and to assist. I believe that AI is going to make us appreciate human connection more, and I believe it's going to make us desire depth and domain expertise even more.
I love Upwork because it is the platform, the place, the company that allowed me to build my businesses, and it allowed me to have financial freedom. It allowed me to be a creative, to be a professional, and a technical, a business-oriented AI professional, to be able to be exposed to so many businesses in such a short period of time. We got started with this idea of being able to kind of improve the software development speed and cost by using AI to generate code. What we work on is building models that are able to ingest and understand code repos that are on the size of 9 to 10 million lines of COBOL code, for instance, and out of that be able to build natural language requirements. We run a lot of research studies. Most recently, we ran a big one with Microsoft and GitHub.
In order to do that, we needed to source around 300 developers pretty quickly in order to have them participate in the study for us to get understanding on how these tools are kind of impacting the way they work. We saw Upwork as kind of the best in class to be able to go quickly find that number of different individuals that fit kind of a specific skill set that we needed for the study. Trying to just go out to LinkedIn and post an ad saying, "We're looking for these people. You might get two or three." When you're trying to find 300 over a month or a month and a half, to run a study of that size could take a year. If you're not publishing and you're not coming out with stuff quickly, you can become irrelevant really fast.
Knowing that we've kind of always got this pool of people to go tap into through Upwork kind of helps us de-stress to a degree of knowing that we don't need to be doing this kind of crazy rat race hiring thing. There are a few things that make me feel safe using Upwork. We had a lot of success in getting those people through the door, making sure that they were vetted, they had some background for it. Knowing that there are some protections and downfalls there from a budget and company perspective is really good. Being able to see individuals' feedback as well as comments from how other people have worked with them and that transparency is super, super important.
As software speeds up, as customer expectations speed up, as the need to differentiate and personalize tools and experiences for individuals starts to ramp up, having an option like Upwork to go and find and kind of bring people in that have these different pieces will be extremely, extremely important. That's how I at least plan to look to integrate Upwork into our growth strategy and really excited to do it. Atlas is a rewards credit card that seamlessly blends unsecured credit with banking features. That is something that we are incredibly proud to be able to deliver to a subprime market. We have people who are trusting us with their money, and it is up to us to be good stewards of their data, of their finances, of their money.
Ensuring that we are hiring the right people who are trustworthy is pretty paramount to what I'm doing. We have to scale responsibly. We have to scale in a healthy way, and we need to do it at lightning speed. As far as the type of roles I'm hiring for on Upwork, anything that is customer-facing, I really do want to run through Upwork, whether that is front-end support, whether it's fraud analysts, dispute analysts. I think that I find two really good talent pools. The first talent pool is people who interact with customers. And then the second talent pool is more of the analyst role where they're able to just take rather complex pieces of data and simplify it for us. My Chief of Staff is from Upwork. Her name is Elle. She is fantastic. When I came in, I said, "You're underutilized talent. You're phenomenal.
Will you please be my very first manager on the team?" She accepted, and she did an absolutely incredible job. I cannot express how I would be lost without her, both personally and professionally. Do you want to find talent quickly? The answer is typically yes. Do you want to be able to hire them without having to worry about having an HR team, a finance team, a talent team to go find them? If the answer is also yes, I think that Upwork is the natural place that someone should go, especially as they're growing. I actually think that leveraging Upwork is a huge secret advantage that we have, and not everyone else is doing it. I don't know why, but I don't want them to know. My name is Saswara Basu. You can call me Sass.
I'm the CEO and co-founder of Zeus, which is a low-cost, secure product storage platform with built-in privacy. Privacy, obviously, is everybody's right. As a human being, you want something private. We wanted to create a breach-proof product, and we took on that challenge. We have created such a product now that's available for consumers. Welcome. At this time, we ask everyone to please take a seat. Our program will begin shortly. It was a challenge to find a platform, and Upwork provided such capability. Talent was the main draw. We looked at several platforms. There wasn't enough talent to go through, whereas Upwork provided that ability to search to great talent from good universities. I don't think any other platform gives you that level of quality. Over the years, I have hired over 300 freelancers. We found some CTO-level expertise.
We're willing to come and work on an open-source project like ours that can really add value to them as well as value to us. It is kind of a win-win situation. AI has changed our business significantly. The workforce needs to adapt to be a manager of those agents and be able to configure the agents, be able to train the agent, or give proper instructions to the agent. I talk about more soft skills now because that is more important in terms of what direction you are giving to the agent, but somebody has to give them the right direction. I think the workforce will change and adapt, and I think there will be more jobs because of AI, not less. We are always going to stay on Upwork because the platform provides us with a lot of capabilities, searching talent, but also managing talent.
I love Upwork because it gives me the ability to hire quality talent and get my product to my customers.
Welcome. At this time, we ask everyone to please take a seat. Our program will begin shortly. I can't explain. A feeling that I've got. I can't erase the feeling that I want. Won't leave behind because it's something that is on. It's on my mind. And when I finish the ocean, I need to do walking music for anybody. Our program is now beginning. Please welcome on stage Andrew Rabinovich.
Hi, everybody. My name is Andrew Rabinovich, and I'm the CTO, and I lead AI at Upwork.
Throughout my career, I've been building AI teams and products at Google, Magic Leap, and as a co-founder and CEO of Headroom, a video collaboration company that was acquired by Upwork two years ago. Today, I lead an exceptional team of world-class scientists and engineers, including 20 PhDs. When I joined Upwork, we formed a new organization called Umami, which is a foundational layer of new Upwork. It combines product, engineering, and cutting-edge research to reinvent how humans and machines collaborate. We came for the largest data-rich ecosystem of real work in the world, a place where AI can fundamentally transform the future of digital work. I want to start with what we've accomplished to date. Search and recommendation is the foundation of Upwork. We've entirely rebuilt it to be AI-native and multimodal, which was underpinned by Objective and Headroom acquisitions.
We optimized it for query understanding, retrieval, ranking, and ad blending. Search and recommendation is the basis for compounding benefits. As Upwork reaccelerates its growth, the new SNR engine drives better matching with 5% more clients converting and approximately 20% talent finding new opportunities in just last year. Upwork's technology is much more than just intelligent AI-powered search. In partnership with product, we reinvented the entire tech stack behind Upwork's next generation of experiences that spans end-to-end customer journeys and powers Uma. Instead of generic, out-of-the-box LLMs that are trained on the entire internet of data, Uma is custom, fine-tuned on platform's data and optimized for efficiency. Active learning facilitates continuous refinement from the experience itself. Uma's specialized custom models significantly outperform general-purpose LLMs across quality with 50% better content accuracy, cost nearly 70% less expensive than off-the-shelf foundation models, better user experience, and nearly 20% more proposals.
Critically, the velocity of model iterations, what used to take three months of work, now can be done in less than one week. This is just a start. As Hayden mentioned, our infrastructure powers each part of our flywheel. Our platform is a closed-loop ecosystem driven by data. Workflows become richer. Demand gets stronger, which turns to turn higher-quality talent, and it expands categories of expertise, which in turn enriches our data yet again. This continuous loop makes Upwork's ecosystem faster, smarter, and more trusted with every project completed. Again, this is all data-driven. As I mentioned earlier, we attract top AI talent because we operate in one of the most data-rich environments. We have over 10 terabytes of data of how jobs are scoped, priced, evaluated, and delivered. Data is the moat for AGI today.
All of AI companies are either trying to buy it or synthesize it. At Upwork, we already have it. It's scaled, structured, and growing exponentially. At Upwork, data is the connective tissue that powers personalization, matching, and learning, yet it is only valuable as a structure of it. We've built a knowledge graph connecting clients and tasks, skills, and talent, all optimized for successful outcomes. This novel work on knowledge graph has been accepted to NeurIPS, a world's premier AI conference, and we presented there in December. Our data advantage is now becoming an execution advantage, powering a generation of agents to deliver work faster and higher-value outcomes. Dave spoke a lot about the first two components of our AI strategy: AI-native marketplace and the AI categories themselves. Now, I want to speak, I want to address how agents will augment human talent to deliver work outcomes.
Agents are a new addition to Upwork, both at supply and demand, as Hayden mentioned. Upwork is uniquely positioned to combine critical human creativity, insight, and judgment with machines' scale and speed to enhance work delivery. We've done extensive research to validate how and where we play in the agentic space on the talent and client side. We know that talent already uses AI tools and agents off the platform, and to make their work more efficient, we want to bring those tools inside. Our findings indicate that freelancers use AI tools 50% more than full-time employees. By working with talent to bring agents into the marketplace, we believe that we can offer higher-quality outcomes at a fraction of time, ultimately delivering much more value to our clients. Speaking of which, on the client side, we're seeing that human talent will continue to be critical in how businesses adopt AI.
90% of leaders prefer human involvement, either as human in the loop or as expert human evaluation to deliver outcomes. Enabling agents on the Upwork platform is natural because both sides want it. Upwork is uniquely positioned to help navigate agentic adoption like no one else, and this is a huge opportunity. We estimate that agents will unlock over $120 billion in digital knowledge work by 2028, growing 50% per year. Agents are the fastest-growing opportunity in AI. The trend of narrow AI is expanding to agents. Rather than having a few general-purpose agents, hyper-specialized agents will be abundant on Upwork. There aren't any clear leaders in this space yet, but because of our data, talent, and AI infrastructure, Upwork is positioned to win. Let me tell you how Upwork is actually implementing all of this in practice.
Over the last two years, we've been on a purposeful journey to reinvent Upwork with AI. We created Uma, as they've discussed. It started as an assistant powered by generic LLMs. It has evolved into an end-to-end companion powered by custom models built in-house. Next, Uma will become a strategic partner for clients and talent with deep context and personalization. Also, over the past year, we've started experimenting with third-party agents and testing their ability to collaborate with humans and complete tasks on the marketplace. This new collaboration space inside Upwork, our agentic playground, is the testing environment where clients, freelancers, and third-party agents collaborate to deliver work outcomes. From the first pilots, we've validated our core thesis that agents alone are not enough. They're extremely fast but unreliable. The best results from the collaboration emerge where human creativity, insights, and judgment are paired with machines' speed and scale.
Through reinforcement learning from experience, Upwork is transforming its playground into a live training space where agents can learn by doing, explore complex tasks, receive feedback from humans, and discover novel solutions, just like AlphaGo's Move 37, where AI discovers strategies that no humans have conceived. This collaboration model isn't theoretical. It is live in Alpha with customers today, with repeat usage across multiple categories, built on Upwork's next-generation infrastructure and design for rapid integration. Third-party agents will be fully available in Upwork's marketplace next year. Upwork's reputation demands that only high-quality agents that meet the quality standards are invited to our ecosystem. Everyone is trying to build economically viable agents today, but no one knows how good they are in practice. Last week, we introduced HAPI, Human and Agent Productivity Index, a new evaluation system aimed to measure performance of agents with humans at the center.
HAPI differentiates itself in two ways. It is not a static and academic data set, but a dynamic set of real-world jobs with economic value associated with them. It captures how agents are actually used with humans in the loop, reaffirming augmentation and not automation narrative. Early results validate Upwork's core thesis that the future of collaboration is between humans and AI agents. Our findings show that agents, even powered by the best AI models, struggle to complete real jobs on their own. This is consistent with findings from other benchmarks, but we have also found that when paired with human guidance, the performance of agents increases by 70% and can achieve results in just a fraction of the time compared to humans operating by themselves. This work has also been accepted to the NeurIPS conference, highlighting Upwork's leadership in the future of work and real AI applications.
Over the next several years, we're building an orchestration layer of work delivery between humans and agents. Uma will evolve from a companion to a partner. It will connect clients to talents and agents. It will be able to qualify, orchestrate, and evaluate work, and coordinate and optimize workflows. As Hayden suggested, as agents become clients on the platform, Uma will also facilitate agent-to-agent interactions. Let me show you how this comes to life.
Today, Uma can help build your next mobile app. Tomorrow, it can help orchestrate the critical work to drive the next big breakthroughs. Led by Uma, innovating beyond staffing, delivering outcomes, and unlocking the future of work. I need help developing an Alzheimer's prediction software for my telehealth company. Perfect. Let's create a project timeline and team composition for developing Alzheimer's prediction software, from prototype to pilot to regulatory readiness. Sounds great.
Communicate with Uma in more ways than one. Work flows seamlessly. Uma pairs the right talent and AI agents for you. Talent focuses on the big picture. AI agents save time, iterating with talent to improve the work. Uma guides the project and verifies the results. Outcomes become breakthroughs. When the work is complete, you can share your life-changing product with the world, from gaming to robotics to healthcare to all facets of your imagination. Smarter, trusted AI agents, greater opportunities for talent, higher-quality outcomes for clients, only at Upwork.
While this video may seem far into the future, in the customer panel led by Peter, Gabe from OMIC shared that they already use talent from Upwork with AI tools and agents to revolutionize drug discovery. In general, deep tech development is moving to independent talent, and Upwork is pioneering this shift.
Ultimately, integrating agents drives value for every constituent of the ecosystem. Clients receive high-quality outcomes trusted by the humans. Human talent, with the help of agents, will improve productivity, tackle more complex tasks, and ultimately earn more. Agents, our new addition, will unlock distribution channels both on the supply and the demand side, learn from real jobs on the benchmark, and benefit from the interaction with human experts. Lastly, Upwork will have new revenues from agent evaluation, have higher throughput, more demand, and higher-value work. This can be only done on our platform, the world's leading human and AI-powered marketplace. Over the last two years, we have had tremendous success. We reinvented the technical foundations and developed next-generation AI infrastructure at Upwork. Going forward, Upwork's competitive moat lies in our closed-loop, data-rich ecosystem.
Integrating third-party agents verified by the benchmark will unlock value for all participants in a large and growing agentic market opportunity. All of this is driven by the insight that the future of AI is human. Now, I'd like to invite my friend Ernesto to talk about Lifted. Thank you.
Thank you, Andrew. Good morning, everybody. The first time that I visited the staffing agency branch was in 2015, and that was back home in Milan, Italy, as you can guess from my accent. What I learned during that branch visit are two things. The first is that companies need contingent talent. They just need it. Why? Because it enables growth and efficiency that cannot be otherwise achieved by employees only. The second thing that I learned during that branch visit was that legacy players, so staffing agencies, have not kept the pace with technology innovation.
Because of that, they provide their customers with suboptimal user experiences. All of that led me to start Adya, which is a platform where we connect people with temporary jobs. That was a co-creation effort between one of the largest staffing agencies in the world and a very large BPO. I then ran the business as a CEO for about six years. Fast forward to 2024, that is when I started to learn about Upwork Enterprise and Upwork in general. I really understood that Upwork is unique in this space. It is unique because it is the only company that has been able to successfully scale a global multi-category, multi-industry marketplace that runs via a platform. Those are the exact same ingredients on which we actually built Lifted. The question you are all asking, what is Lifted?
Lifted is one solution that is specifically built for large enterprise companies so that they can source, contract, manage, and pay any kind of contingent talent. Essentially, it is how we can unify and then elevate an industry which is incredibly fragmented. What we're going to cover today, a few things. We're going to start with the market. We're going to uncover what are all the needs of those large enterprise companies, and then how Lifted actually meets all of those needs. After that, we're going to discuss why Lifted is actually very different from our legacy enterprise offering, and also why it is different from anything else that exists in the market, and how all of this is actually going to unlock a lot of growth for Upwork. Let's start with the market. The bottom line here is that the market for enterprise contingent talent is just massive.
We're looking at $650 billion by 2028. The question you might be thinking is, why is this market so big? What are the needs that those companies have? With contingent, they get access to talent that they cannot otherwise get access to. Second, they get flexibility. And lastly, efficiency and cost savings. Let me unpack this a little bit with an example. Just a few weeks ago, one of our large clients came to us with a very hard-to-fill role, a role they actually tried to fill for multiple weeks on their own unsuccessfully. We actually filled that job within a day. Now, think about the impact. Massive. It unlocks growth because now they have someone in seat. They can actually get that job done.
Because they decided to engage this person with a contingent labor assignment, this means that they retain a lot of flexibility, and they also know exactly how much this project is going to cost them, meaning they're not going to have any unnecessary cost on the side of this. Not only is this market huge, but there is AI as a major tailwind. Aiden and then Dave already mentioned that 77% of companies expect to increase the utilization of fractional labor. Fractional labor means contingent labor. The question is, why is that? It is because AI is doing two things. First, it's creating net new jobs. Second, it's actually fractionalizing jobs that were full-time before. Again, let's dive into an example. Let's say you are a very large enterprise company, and you implement a new AI billing and collection infrastructure. That activity is really mission-critical.
You really wanted to get it right. And because of that, you still need humans to verify the job done by AI. Those jobs are way more likely to be fractional jobs and therefore contingent versus being FTs. What we are hearing from our customers more and more are quotes like the one that you see on the screen. AI is pushing functional leaders to leverage contingent more than FTs. All of our expected outcome growth is happening in contingent. Now, let's zoom out for a second. What is it really telling us? It's telling us that the need for flexibility has gone up a lot because of AI. And they understand that contingent is the way to go. Lastly, how does enterprise actually access contingent talent? The first thing to notice is that those companies have established workflows, policies, and compliance requirements. That's how they work.
It is very important for anyone that wants to work with those companies to actually meet those requirements. Second, what we see here on the slide are the five main contract types that exist within contingent labor. You have independent contractors, agent of record, employer of record, staff augmentation, and outcomes. Now, do not worry. We are not going to go into each of the contract types. There is no legalese here. What you need to remember, though, is that there are those five main contract types, and companies either want or need to use one or more of those in order to get the job done. How has Upwork been playing in this market until now? We have 300-plus enterprise customers from very large technology companies like Microsoft to many tech-enabled businesses to a lot of industrial players. What have those customers been buying from us?
Three things: access to talent, speed of delivery, and an offering that essentially meets their enterprise-grade requirements. Let's unpack this a little bit. Think of any job that you can get done via laptop. You can essentially get the job filled via Upwork. That's how big this talent pool is. There are 10,000-plus skill sets, including 250,000-plus AI experts. The one data point I'm most proud of is that 4.92 out of 5. That is the average rating that our clients are giving to our talents. The talent quality on Upwork is high. Not only do we provide this talent high quality, but we provide it incredibly fast. What does fast mean? Fast means a couple of things. First, you can get shortlisted as fast as within the same day. That's massive. Think of the last time you needed to hire somebody.
Getting someone on seat fast is essential. Being fast means that the average time to fill a job is less than three days, with many jobs actually being filled in less than a day. Lastly, through a combination of platform and service offering, we have been providing those companies with a global and compliant solution. Now, throughout those years and conversations with customers, we also learned that we had some limitations. We learned that essentially we had some barriers to unlock additional growth. I want to unpack those a little bit. First of all, within contingent labor in the enterprise space, there are two buyers. First one are functional leaders and hiring managers. Think of those as your VP of IT or marketing manager. What Upwork Enterprise has been optimized for is to provide those functional leaders and hiring managers with direct access to contingent talent.
As we have seen on the prior slides, it's actually done a phenomenal job at doing that. Now, there's a second buyer: centralized workforce programs. Those programs are specific functions that typically sit within procurement or HR. They have one job. Their job is to identify providers of contingent labor that meet cost and compliance requirements and then make sure that everyone within their companies actually is using those providers to ensure consistency. Upwork Enterprise was not equipped to sell to those buyers. Why? Because it did not meet their requirements in terms of workflows and distribution channels. What you need to remember of all of this: two buyers, we could only sell to one. Second limitation: within contingent, we said before there are five contract types. Upwork Enterprise has been providing directly two of those: independent contractors, agent of record.
For the remaining one, it was dependent on partners. This led to a suboptimal user experience. Now, let's fast forward. After the acquisitions of Bubty and Ascen, we have filled both of those gaps. We do not have those gaps anymore. Why is that? With the acquisition of Bubty, we are now able to sell to our functional leaders and hiring managers, same as before, but now we have a better offering, better product, and service for them. Now we can also meet the workflows and the distribution channel requirements of the contingent workforce programs. Bottom line, before we could sell to one, now we can sell to two. With Ascen, we went from being able to provide two contract types directly to being able to provide all contract types directly.
Let's zoom out of all of this for a second, as there are a lot of contract types and legalese in this. I know today is incredibly cold here in New York, but let's assume you will want an ice cream. You walk into an ice cream shop, and they tell you that they produce an amazing vanilla flavor. Incredible. But you crave another flavor. And they say, "I only have prepackaged ice creams." You walk out. You go down the street. Now you enter another ice cream shop. They also make an amazing vanilla flavor, but they also make every other flavor in store. And they all taste as good as the original vanilla flavor. You are way more likely to buy from this shop. In a way, Upwork Enterprise was the first store, and Lifted is the second. Let's summarize all of this.
We retained our fast access to high-quality talent. This is what we have been doing for many years. With the acquisitions of Bubty and Ascen, we now essentially can meet the clients where they are. What does it mean? We went from being able to sell to one buyer. We can now sell to both, and we can now provide all contract types for them. With this, we launched Lifted, which again is one solution specifically built for large enterprise companies so that they can source, contract, manage, and pay any kind of contingent talent. Now, what you might be thinking is, what does this mean for your existing client base? Let's dive into one example. One of our clients is a very large pharmaceutical company. They spend $1 billion a year in contingent. That's massive. Spend a lot of money in contingent.
Out of that, $50 million is for the two contract types that we used to provide directly. Actually, within that, we had a very good share of wallet, about 20%. However, we did not have access to the $950 million. Why? Again, we could not sell to both buyers, and we could not provide all contract types. Now we launched Lifted. What happens? Two things. First thing that happens, the $950 million becomes available. Again, why? We can sell to both. We can provide all contract types. Second thing that happens is we can actually increase our share of wallet also for the contract types that we were providing before. Why is that? We have a better product and services offering, and now clients have an incentive to consolidate spend with us and decrease the amount of fragmentation in their own ecosystem.
Now you might be thinking, "Okay, not every company spends a billion dollars a year in contingent," and you are absolutely right. I wish that was the case. It's obviously not the case. So what's our ICP? Our ICP, we identified about 5,000 companies that meet our ICP ideal client profile. Think of companies that have thousands of employees and billions of dollars in revenue. So translated, very big companies. Those companies spend anywhere between $50 million and a billion dollars a year in contingent. So that's our ICP. The next question is, how are we actually going to win those companies? So what's the go-to-market strategy here? The go-to-market strategy is very much straightforward. Let's start with existing customers. With existing customers, like I just explained in the pharmaceutical example, we have two jobs to be done.
Job number one is expanding share of wallet within the existing contract types. Again, why? We have a better product offering, and they have an incentive to consolidate with us. Second, it is about starting to gain share of wallet in the contract types we did not have access to before. This involves having conversation with both buyers, the ones we already know and like us a lot, as we have seen from our ratings and feedback, and also starting to engage contingent workforce programs more and more. Over the last year or so, we have done a phenomenal job at actually increasing our operational efficiency and decreasing our cost to serve a lot. Because of that, we are fully confident that we can actually deliver on this growth strategy within the existing customers with our existing account management and customer success teams. Now let's dive into prospects.
Here we have a three-step approach: pilot, adoption, expansion. How can those clients start working with us? Two ways. Option number one, they start by buying one contract type or two contract types, and then they expand from there. Essentially, we make it very easy for them to start and then seamless to scale because it's all within the same solution. Second option, we can work together with them to analyze their total contingent labor spend and assess together how we can drive down the cost of ownership for those contingent labor. This is where the value proposition of Lifted actually becomes unique. That's something that doesn't exist in the market. We are truly the only company that provides a contract agnostic, meaning we can do all contract types, country agnostic, which means we can be global, a solution that is run by a platform.
That's why we can do this for those companies. What we are seeing so far is already positive leading indicators, both with existing clients as well as with prospects. Let's dive into two examples again. Example number one with an existing client. This company is a large services business that has been working with us for many years. Right after we announced Lifted, they came to us, and they invited us to an RFP for a GSV value, which is totally three times the amount of what we currently do with them. Why? Because this is for contract types we did not do directly before. That's how powerful this is. Now let's look at an example with a prospect. This one is one that I personally particularly like.
This company, we had a conversation with them a couple of months before we announced the acquisitions and before we announced Lifted. They told us, "It's going to be impossible to work together." Why? Because we did not provide the contract types they care about directly. Now, fast forward, we launched Lifted. They call us back, and they say that they are planning to invite us to the next RFP, which they will run in 2026 for a contract value of about $10 million of GSV. Obviously, those are just two examples of those positive leading indicators that we are seeing both with existing customers and prospects. Why is all of this happening? All of this is happening because of our value proposition. It all goes back to, first, the fact that we provide companies with high-quality talent fast.
We have been doing this for many years, as we've discussed upfront. With the acquisitions of Bubty, Ascen, and the launch of Lifted, we now can essentially unify and elevate an ecosystem which is otherwise incredibly fragmented. Translated, we can now sell to both buyers and provide all contract types. Because of this, we can unlock massive cost savings for an estimated rate anywhere between 10%-30% of total cost of ownership for contingent. All of this is what gives us confidence that we have a clear path to expand in enterprise. We're going to take a three-step approach here. From now until the end of the first half of 2026, we're going to drive expansion conversation and land discovery with existing customers and prospects. This activity is fundamental. Those companies take time to buy. Sales cycles are long. This activity is very important to be done.
Second is between the second half of 2026. That's where we need to start expanding with those existing clients and then doing pilots with prospects. That will lead to 2027, which is where we're going to scale this business. Let's summarize everything we touched on today because it was a lot of content. The first thing to remember is that with the launch of Lifted, we have significantly expanded the market that is available to us. The second thing is that we start seeing positive leading indicators both from our existing customers as well as from our prospects. Lastly, our plan is to start growing this business in the second half of 2026 and then scale it in 2027. With this, I think it's a great segue to invite back on stage our CFO, Erica. Thank you.
I think we're going to take a quick two-minute break just to adjust a couple of things on the AV. I want to keep your sense of anticipation up for the financials. Give us two minutes.
Music, please. Music.
Our program is now resuming. Please welcome to the stage Erica Gessert.
Thank you. Thank you. Thank you very much, Frisco. Once again, I'm Erica Gessert, CFO of Upwork. I'm here to bring it home and to tie everything that you have all heard today into clear financial outcomes. Just a little bit of my own origin story. I came to Upwork about two and a half years ago after a career in large enterprise. I came from PayPal, where I helped to drive transformation across a global fintech business. I came to Upwork really because I'd seen two very large outsized opportunities. The first, of course, in enterprise. One of the reasons I'm so, so excited about the Lifted strategy, PayPal alone was at the higher end of that range that Ernesto just showed you. We spent about $500,000,000 of spend on contingent labor a year.
I knew that the TAM for this opportunity was enormous. The second was in AI. When Hayden and I were talking about me coming to Upwork, it was April of 2023, and ChatGPT had been out for about seven months. Hayden talked about this vision of transforming the platform through AI enablement and about the possibilities around human and AI collaboration. It was incredibly exciting and compelling to me. Today, at Upwork, we are realizing that vision. It is just incredibly exciting. We have spent the past three years, as we have talked about so much today, reimagining this business for profitable growth and to win in the AI era.
I want to share with you now why we are so confident in the growth outlook that we are putting out there today, built on a profitable foundation that is going to produce operating leverage as we grow, and in a business that has enormous free cash flow yields, building a strong balance sheet that will give us flexibility as we grow as well. Today has been all about showing you all, enabling you to see what we are seeing in our business, that we are really on a foundation of accelerated growth for GSV, for revenue and adjusted EBITDA, and that we're incredibly well positioned to grow with this profitable base. Why is Upwork positioned to win? This is the flywheel that Hayden shared. We've rebuilt our platform to be AI native. Uma is the best tool to help talent and clients find each other seamlessly.
Dave talked about the fact that Uma is touching now about 70% of new clients' work. That's just the beginning. That's new clients. That's up to the pre-match. In 2026, we will be extending Uma into the post-match experience, and it will have an even greater profound impact on our platform. That all leads to natural demand, growth in demand on the platform, nearly 800,000 active clients, drawing that 18 million active talent, active professionals on our platform, over two million of which have advanced degrees. Our category reach is incredible. We have over 50 categories on our platform with more than $10 million of GSV in each year. Clients transact across categories. More than 40% of our clients have projects in multiple categories. This all contributes to our incredibly unique proprietary closed-loop data asset.
We're the only ones in the industry who have the richness of data that helps to train Uma. It helps to drive personalization, and it helps to enable the human plus AI collaboration that we are building. Upwork is the only company in the industry that has strategic advantages across all of these fronts. The flywheel is the result of the platform we've built. Our platform attracts larger customers with broader spend. This has been the very clear pattern over the past four quarters. 85% of our GSV comes from recurring clients. These are clients coming back again and again to our platform to do work. Our GSV per active client has grown for five consecutive quarters. It's over $5,000 per active client and growing. This leads to 90% of our GSV coming from high-value work.
This is what makes our business model so strong and our economics superior. It's because our relationships with clients and talent are richer, are more meaningful, and it means that we can keep investment down as we continue to grow and scale. Now, over the past year, we've been focusing. We've talked so much about our focus on quality over quantity, and this strategy is working for our business. The number of new clients coming on and spending over $1,000 within the first three months of spend, it's up 11% year over year. Now, that spend in the first three months, it is the highest ever at our company right now. In 2025, we are set to grow that first three months of spend every single quarter throughout the year.
That's the first time that's happened in this business, the first time we've driven that since 2021, right in the middle of the pandemic. Our customer retention is improving and continuing this trend of coming down 170 basis points year over year. We're continuing to see that trend continue. This means that as our customers stay with us, they're stickier, more meaningful relationships. Our retained customers spend six times what new customers do on the platform, which means we'll be able to keep marketing spend low as we retain these customers and they build their relationships with us. Now, over the past three years, as we've been rebuilding this business, we have had an incredible track record of inorganic and organic investments on the platform that has truly transformed Upwork. That's what we've been talking about so much today, the new Upwork.
First with Bubty, or sorry, first with Headroom and Objective. We built out our AI talent bench. We brought Andrew and the incredible Objective team on board over the past couple of years. Now with Bubty and Ascen to unlock the enterprise opportunity. On the organic side, we've injected AI into the customer experience. We've introduced tiering on the platform, a profoundly important strategy for our business. We've also built the foundation for human plus AI collaboration that's coming. We've also expanded our monetization strategies, which I'll get to in a minute, with very high confidence levers that will grow both GSV and revenue for our business. Together, all of these moves have created an engine to accelerate Upwork's next phase of growth. We've also consistently outperformed the market over the past few years.
We're taking share in periods of macro growth and also in periods of slower growth within the broader market environment. Over the past three years, we've significantly outpaced staffing, but also the broader online marketplaces. This just shows the durability of our model. With the strategies we've put in place, we are set to accelerate this growth and this lead regardless of the macro backdrop. Our performance is actually the easiest to see when you look against the broader market indicators. This is job growth. In very strong cyclical periods, Upwork is lifted by the market. In periods of slower growth over the past three years, like we've all been experiencing, Upwork has significantly outpaced job-related indicators. With the growth drivers we've now put in place, Upwork is positioned to grow untethered from the macro environment. We're incredibly confident about this.
We will only benefit if and when macro tailwinds return. We are now poised to drive growth again and even more meaningfully in this business because of the foundation that we've built. We're entering a period of durable compounding expansion. This is on top of a stronger base, a broader opportunity, and growing demand across every segment we serve. As we've talked about so much today, we have traction in three enormous growth areas: SMB, enterprise, AI. Our current penetration in each of these areas is actually quite small, which makes the opportunity enormous. This is also why we can confidently say that we're going to grow regardless of the macro environment. It's because we have very, very, very clear traction with initiatives across each one of these opportunities.
Our growth plan is based on the building blocks right here: transforming human and AI work, accelerating SMB growth, and unlocking the enterprise expansion. I'll walk you through the traction in each of these areas. First, the AI opportunity. As you've heard today, the AI opportunity is not one opportunity. It's two high-traction, very near-term opportunities and one longer horizon one. First, on the AI native marketplace, we've embedded Uma across, as I said, the pre-match experience on our platform. Job post creation, search, match, job proposal writing. That's just the beginning. This $100 million that we're experiencing on the platform this year, this is really from experiences that we've launched in the first half of 2025. Things like AI interviewer, things like Uma recruiter, they're not even showing up in our results yet.
As we continue, we see this accelerating because we have new experiences launching on the platform every day that are going to drive this growth. Next year, as we expand Uma into the post-match experience, we expect that to accelerate. The second, of course, within AI is the AI category itself. The annualized AI-related GSV on the platform is now set to surpass $300 million by the end of the year. The tailwinds behind this are huge in the marketplace. Further down the road, and a catalyst in 2027 and beyond, is the AI Agentic work on the platform. We're already seeing demand for this work. The requests for talent using Agentic AI on the platform are up 12x what we saw a year ago. It's also up 20% quarter over quarter in Q3.
As Andrew talked about today, we are already testing and experimenting with AI agents on the platform to benchmark them and to collaborate with humans to get real work done, a true differentiator from any other benchmark that's out there. While it's early, we expect to monetize this work at take rates above the rest of the platform because of the productivity and speed to market that this kind of work produces. We're getting feedback from clients on this. We're already getting significant ecosystem interest from larger partners out there, other AI companies to partner with us on human-in-the-loop solutions. Upwork's the only company in the industry that is poised to take benefit in all three of these areas from this new technology. Next is our traction with SMBs. Business Plus is leading the way to greater penetration with mid-market customers.
This is a very important segment for us, and we are moving up market. The average GSV you heard about is 2.5, what our core platform is. Our growth in Business Plus grew 33% quarter over quarter in Q3 and continuing to increase. We had not even started marketing this product yet when we produced this result. Now, Business Plus, importantly, also drove 20 basis points of take rate accretion in the quarter in Q3. We only had low single digits penetration on our platform that drove this take rate accretion. This is a strategy that will drive both GSV and take rate. The tiering of the marketplace, as I said, it is a profoundly important game-changing strategy for Upwork. It promises to be a structural change for our business. Just looking at current momentum in this, we expect that Business Plus volumes will double next year.
That will only bring us to about 5% of penetration on the platform. This is going to be a multi-year growth lever for this business. Finally, of course, close to my heart, enterprise is our largest growth opportunity. The work we have done over the past two years has completely reshaped our ability to access this $650 billion market. We have an incredible base, as we have talked about, of marquee customers already with Upwork that are eager to expand their relationships with us: Microsoft, Lime, Scale AI, Cloudflare, to name a few. We are competing for much larger contracts. What Ernesto talked about with the client that is on the platform who will have 3x to spend with us, that is a small comparison. We are easily going after contracts that are 10x the size of our legacy base.
This momentum positions us now to drive a minimum of 25% growth in GSV next year. We think this is a small growth rate. Remember, this is going to scale at the end of next year in 2026. This is actually small compared to what we expect in 2027, in 2028, and beyond. The GSV growth levers are very clear. I just walked through them. AI from the fastest growing work category to the enablement of the platform through AI-enabled experiences. We are just at the beginning of the tailwinds from this technology, more deeply serving SMBs with the momentum on Business Plus. Enterprise, of course, the huge opportunity now fully open to us with Lifted, also promising to be a multi-year growth opportunity for us. Finally, ads and pricing.
The new strategies, well-tested strategies on our platform that we are launching and in the middle of expanding. These are strategies that will grow both GSV and revenue on the platform. All right, on top of those GSV growth strategies, we do have multiple high-confidence take rate strategies. Now, the use of Connects continues to grow, continues to be a driver of take rate on our platform. Connects revenue grew 18% in Q3. That is really driven by increased use of ads and bids as we dynamically price to the size of contracts on our platform. That is continuing to drive that use of Connects. Freelancer Plus revenue was up 24%. There is still opportunity to add new tiering to that offer and those subscriptions. Of course, the experimentation that we have been doing with the variable freelancer fee. This is nascent on our platform.
Again, an early strategy that we've only expanded to a few categories. Now, in 2026, we are set to expand further and drive take rate and GSV as we actually dial it down in places in order to drive demand. Looking beyond 2026, we also expect as we grow into broader relationships with SMBs through Business Plus, we have the opportunity to launch additional value-added services. This will be a catalyst in 2027, 2028, and beyond. With these drivers, we expect to expand take rate to the low to mid 20% range by 2028. As in pricing, like I said, that is just one driver of take rate. Agentic work on the platform, Business Plus with SMB growth, enterprise expansion, these are all drivers of take rate on the platform as well as GSV. We're expanding take rate by adding value.
These are well-tested strategies that we're executing on. Of course, we also have very clear room to grow our take rate. Even at 20% today, we are well below industry averages. Together, these GSV and take rate drivers create a very strong, diversified path to both GSV and revenue growth. For revenue, we expect that we will drive 13%-15% revenue growth CAGR through 2028 because of the combination of these strategies. Now, while we've been rebuilding our growth engine, we, of course, have also done a phenomenal job. It's been a team effort optimizing our cost base. We've created a disciplined growth engine that will create operating leverage as we grow. Our track record on this is incredibly solid. It shows up in our results. We're going to post about 28% adjusted EBITDA margin in 2025.
This is an average of nearly 10 percentage points a year of adjusted EBITDA margin accretion every year for the past three years. Now, we've done this through targeted optimization across nearly every part of our business, first in sales. We have lowered both our cost to acquire and our cost to serve in sales. Since 2022, we reduced our total sales expense by 35%. We reduced our cost to acquire by 40% and our cost to serve by 20%. Now, we will make some incremental investments in 2029 or in 2029, no, in 2026 to invest in the Lifted strategy. Beyond 2026, we expect to grow sales expense at a very small fraction of the revenue growth coming from this business. This business is going to scale extremely nicely as we go after these very large contracts. In marketing, an even more impressive result coming from marketing.
Over the past three years, we have reduced our marketing expense by 55%. We have increased the GSV yield from marketing by nearly 25% at the same time. We're getting more yield out of lower dollars. Marketing-led contracts that we are achieving from marketing, the GSV per contract is up over 100% over the last three years. Our growing focus on these high-quality, lasting customers is working very well for us. It's showing up in our GSV per active. It's showing up in our marketing yield. This means that we will be able to keep marketing costs low as we grow because we're getting such a high yield out of every incremental marketing dollar. Lastly, in R&D, we have produced an incredible pace of productivity through optimizing our R&D spend. We've embarked on a comprehensive program to implement AI internally across different functions.
The place it's showing up best right now is in engineering productivity. In any one month in Upwork today, about 25%-35% of our code is AI-assisted. That has resulted in a 32% reduction in the cost per line of code produced and a 16% increase in code going into production. This means more improved customer experiences coming to our customers faster every day. That is what's behind the pace of our execution today. We will plan to slow our margin expansion in 2026 because of all of the outsized opportunities we see for growth in this business. We still have many more levers to pull from a cost point of view in this business, and we have clear line of sight to the 35% margin target that we've set and incredible confidence that we will get there.
One of the best things about this business, this is what we talked about when we talk about superior economics in this business. Our profitability converts incredibly efficiently into cash. We are a very low-capital business. About 85% of our EBITDA, it converts to free cash flow. This gives us the flexibility to invest, to acquire, and to return capital as we go. We are projected to produce well over $200 million of free cash flow in 2025, which will be up over 50% from 2024. This will give us an incredibly strong balance sheet to continue to drive the growth that we're talking about. All right, our capital allocation strategy has been very consistent. It's very clear. We are focused on a discreet set of organic investments, narrow big bets that we talked about with AI, SMB, enterprise, while always maintaining a relentless focus on margin accretion.
We will also continue our track record of high ROI M&A, consistent with the track record with Objective, Headroom, Bubty, Ascen. We will continue to use our balance sheet to enable our product roadmap and our growth. Of course, we will consistently return capital to shareholders. At a minimum, we will offset stock-based compensation, but we will also strategically and opportunistically reduce our share count as we go. We've had an incredible track record over the past three years of capital return to our shareholders. We plan to continue that well into the future, particularly as we see these low valuations out there in the market. All right, this brings us to our 2026 guidance. Q3 of 2025 was about the return to GSV growth. 2026 is all about accelerating that growth. Our GSV growth rate in 2026 will accelerate to 4%-6%.
We're very confident about this based on all of the traction that I described to you. Our revenue growth rate will go up to 6%-8% next year. Our adjusted EBITDA margin will expand by a point while also absorbing a couple of points from the investments in these high growth strategies that I've talked about. All right, this all rolls into our three-year outlook. Our exit ramp for 2025 is up and to the right. We are very confident about our 2026 outlook. With the Lifted strategy, marketplace tiering, the agentic opportunity, and so much more, we also feel very confident that we see very clear growth levers for 2027, 2028, and beyond.
That's why today we have released our new three-year growth outlook: GSV CAGR of 7%-9%, revenue CAGR of 13%-15%, and adjusted EBITDA CAGR, that operating leverage I talked about, of 20% over the next three years. We are executing from a position of strength, scaling growth, expanding profitability, and building sustained shareholder value for years to come. In closing, this is the new Upwork, and there has never been a more exciting time to be at this company. I know I speak for all of us on this. We've fundamentally rebuilt this business across our platform, our targeted customer focus, our operations to be at the center of growth in the AI era. We've got growing traction in these three massive opportunities, and we're poised for accelerating growth and profitability. Many thanks to all of you for your time and attention.
I am now delighted to invite Hayden up to join me for Q&A. All right, I think we have some mic runners. You may have some questions from the webcast as well. And some seats. Thank you. I'll take a seat. Thank you so much. Thank you. All right, let's do it. Looks like Bernie has a question here.
Great. Thanks for taking the question. Great job on the presentation. It was great to learn about the new Upwork. I think it would be just helpful you provide that really helpful bridge in terms of GSV growth and how it's going to accelerate. Can you just talk about some of the puts and takes from the 4%-6% GSV growth in 2026 and how we get to, I think, double digits in 2028?
Yeah, sure.
As I was kind of going through it here at the end, many of the growth catalysts that we have on the platform, Lifted first and foremost, right, is ramping it towards the end of 2026. We expect that 25% GSV growth from enterprise that we're expecting next year is really a back half event for us because of the length of contract times. Obviously, we're going through platform integration right now. That gives us very strong confidence in even stronger growth just coming from enterprise in 2027 and beyond. Make no mistake, the growth drivers are both on the marketplace and on the enterprise. We are seeing growth catalysts across both of these businesses right now. Similarly with Business Plus, while we expect to ramp through 2026, we expect that to be a multi-year growth catalyst as well.
Expect kind of ongoing acceleration as we continue to build out that product. There are further opportunities for marketplace tiering just beyond Business Plus and other things on the marketplace. Thank you for taking my question. My first one is just on the 25% growth for the enterprise in 2026. I think that would imply for just the core marketplace business, a more modest pace of acceleration in 2026. Can you just talk about, yeah, just what are the key inputs to the marketplace? Can AI drive some sort of growth on top of that low single-digit growth? I think that's implied. Yeah, I think that it should imply sort of an even GSV kind of growth driver from both marketplace and enterprise. It should be relatively even.
Aside from that, I would say the growth drivers on marketplace are kind of, as I articulated, I would say marketplace tiering. The AI category itself, of course, that category is accelerating right now. We are doing kind of focused work in order to drive strategies that will harness that growth. We certainly expect that to also be a driver in 2026. Yeah, and I just add, we certainly have been cautious about modeling in a lot of expansion from things like AI agents, which are such a new opportunity. That is part of our outlook, and that could be even bigger than we expect. Right now, we can see the traction with Business Plus, with the AI categories. I think those are big building blocks for us on the marketplace side.
Other things could happen sooner or further out, and we're not reliant on them to hit our targets.
Thanks so much for all the details. Maybe two, if I could. One, in terms of the three buckets of growth you laid out, can you talk a little bit about your go-to-market strategy in SMB, enterprise, and AI, and how much they're already built out today versus there's elements of still putting in place go-to-market strategies over the long term as opposed to just executing against the opportunities? The follow-up would just be on the marketing. There was some interesting data there in terms of the efficiency on the marketing side. How should we be thinking about continued levels of efficiency at that rate of change going forward? Thanks.
Eric, so we have the building blocks for that go-to-market already in place.
I'd say on the enterprise side, Ernesto has done a fantastic job really retooling how we go to market there with a smaller, more nimble team that is going after these big contracts. That work will continue into 2026. A lot of the kind of foundation for that is already set. Erica shared some of those sales efficiency metrics, which speak to how we're set up for this new and kind of different profile of enterprise go-to-market. On the marketplace side, we have done a lot to tune our engine already for AI and AI workloads. There's scaling of that that we can definitely do across channels and kind of optimizing our channel strategy and overlaying that with our SMB focus. I'd say those are things where on the SMB, Business Plus side, we're definitely earlier in terms of marketing that.
Dave shared the campaign that we just launched in October. There's more coming. We know how to do great digital marketing as a business and use that as a lever. Doing that with SMB and then layering on AI, I think, is going to be a great kind of compounding opportunity that we're just building into. Yeah, I'd just say just on the marketing efficiency, it takes time to really turn the dials from a channel point of view and really turn the dials to achieve what we've achieved in terms of attracting these larger, better customers. Obviously, a lot is changing within marketing channels right now, right, with kind of SEM and SEO changing a lot as the environment changes.
We are going to continue to turn those dials and make sure that we are achieving these targets of attracting these high, high, high ROI customers. It will take time to continue to scale that up. As we see the yields from the business and see the kind of revenue and GSV accretion, it certainly is possible that we would dial it up more as we really see that ROI, always with a focus on our margin targets as well.
Hi, thanks for taking my questions and great job on these presentations. You talked about monetizing the AI agents because they add so much value. I was just curious on how we should think about the interplay there because you would think that AI agents would make projects complete a lot faster because of the obvious speed of them.
With your hourly contract work, how should we think about that being accretive because you're able to monetize that? Have you kind of considered that in your forecast?
Yeah, so a couple of things there. First and foremost, some of the things that Andrew showed on the HAPI index and all the work we're doing on the benchmark, what that's really showing is that the future of AI agents on our platform, on any platform, is with human in the loop. We're seeing that very clearly in the data. These agents, they may be completing tasks, but it requires human oversight. It is really going to be the combination of the two. Now, we're talking to our customers about this. To the extent that things can get done faster and productivity is better, they're willing to pay a premium for that work.
That is how we see it. Now, we have not modeled in too much kind of take rate accretion from this at this point because it is so new and nascent. We do see that opportunity, and that is the feedback we are getting. Yeah, and we see in the data that, again, our AI clients that are leaning in the most are also spending the most. That is even including the fact that these searches and appetite for AI-capable talent that are using agents themselves is increasing. Even as talent are using more agents, even as clients are leaning and looking more for that talent, they are spending more. Their contract sizes are going up. What they are having the humans do is not replaceable by the humans, even as the technology gets better.
Because for all of the work, there's setting the goals, setting the kind of concept of what needs to get done. And that's a human task. Agents cannot set goals for themselves. Agents can help execute the work, but then a human needs to come back in and qualify that the work was done correctly. And so even as agents get better at that middle part, there is so much runway where humans need to kind of start and finish projects. We see that already happening.
Great. I had a second question. You mentioned more tiering in the marketplace across your different offerings. Just curious, do you see equal opportunities kind of maybe a lower level and an upper level? Or do you think you'll be skewing more just like up more premium products? Or how should we think about that?
I think the future here, look, right now, we're super focused on optimizing Business Plus. It's actually, like I said, it's very low penetration on the marketplace. There's huge runway there. That is our first and biggest focus right now. I think the future here is in more premium products on the marketplace. I think we'll have our basic marketplace offer. That is why this is such a key strategy for our business because there's really opportunity to continue to cater to larger and larger customers on the marketplace.
Thanks for taking the question and for the great presentation. My question was just on Business Plus. Could you talk about the key opportunity in terms of the post-match experience? What gives you confidence that you could double GSV in 2026?
Absolutely.
The post-match experience in Business Plus is really about what Dave shared with UMA, helping project manage kind of everything that happens after the hire has been made. We have this incredible data asset where we alone see what happens to project completion, what makes a good outcome versus a bad outcome. We're using that data to train Uma to be a project manager that's kind of always on and helping the project stay on track. That is a huge value add because today, either the customer or the client has to do that, or the talent has to do that. Maybe they're not even aligned on what that looks like. Now we can come with something in the middle that really helps both sides be successful. Incredibly sticky, incredibly valuable. It literally saves hours of work every week from people who are doing it today.
That's the new experience. That is something that we do view as a premium product as part of this Business Plus expansion. As we alluded to, with Business Plus, we've really just launched that into the ecosystem, our existing customers and existing product. We've been adding new entry points in the product. We haven't yet really done concentrated marketing to educate SMBs about this product and get them in. Organically, we're seeing some of that, but there's a lot still that we can do with our marketing channels to build that. In addition to expanding the value of Business Plus, we're also expanding kind of access and visibility through our marketing channels. That's what really drives that 5% GSV goal. I'd just add the 33% quarter-over-quarter growth that we saw in Q3, that was pre-marketing, right?
And so just the momentum of that product right now is making the projection pretty confident.
Thanks. Hi, Hayden and Erica. Thanks for the presentation. Question on the AI GSV. Definitionally, is it very easy for you to discern what is and what isn't AI on the platform? I guess numerically, what's the growth that's baked into the 2026 outlook for the AI portion of it?
Yeah, sure. It's a great question about how easy it is to identify the AI work on the platform. Still the largest subcategories within this AI category are dedicated AI work. We're talking prompt engineering. We're talking AI infrastructure work, data labeling to a certain extent. This is relatively easily identifiable AI work. That said, there's AI work proliferating cross-category.
That is harder for us to identify because it doesn't fit into the classic lexicon that we have on the platform. We do have kind of methods to actually, we've trained an LLM to scan our data and identify some of this work. Then we've kind of backtested it. We are working to identify across category, things like in video animation and other things. It's an ongoing work in progress, I'd say. In terms of the growth of the AI category, we're not providing a specific projection on it. I would say that growth is accelerating. There are real tailwinds behind it. We expect it to continue to be on the rise quite significantly.
Okay, great. If I can ask a follow-up on the financial targets. Obviously, there is a gap between GSV CAGR and revenue CAGR.
That makes a lot of sense. It seems like that gap increases beyond 2026 into 2027, 2028. Could you just talk about kind of the driver there?
Again, Business Plus, it's such a great product because it drives both GSV and take rate. Obviously, as enterprise scales, that's also enterprise overall tends to be at an average higher take rate than the marketplace. That's another very nice driver for us. Just to be clear, our marketplace product today has a 5% take rate on the client side. Business Plus has a 10%. It's a big step up. That's what's really a big driver of that. We have these other drivers like the variable freelancer fee is super early with that. We're seeing lots of success there. There's just a lot of levers for us.
Hi, this is one from the webcast. You mentioned the volume of proprietary data on the platform and how it's enhancing AI-driven work delivery. Can you talk about how meaningful a competitive advantage this is for your platform? And could you help frame how this data advantage feeds into your broader monetization strategy?
This advantage really can't be overstated. It's huge. And we're really early in unlocking it. I mean, it's just in the last year and a half that we really have Uma-powered features that are taking advantage of this data asset. As we talked about, we haven't even launched the Uma capabilities that are in the post-hire experience, the project management, the things that happen.
We just are launching things like insights for customers where Uma can scan work diary snapshots coming from their talent and give them insights about how people are working, where there might be issues. We have a corpus of data that we ourselves have not yet fully unlocked and turned into features, but we have a huge roadmap to do that. There is also a lot of interest from partners in the ecosystem. Our data is so unique and uniquely at scale. We get a lot of interesting conversations with other folks, foundation model companies, other agent builders, etc., who really want to be part of seeing and learning from that data.
What you heard from Andrew was very interesting in terms of us creating an environment where we benefit and our talent benefits, but AI agent builders, foundation model companies can come in and launch AI agents that once they pass our quality bar can be operating in the marketplace and using the marketplace experiences they have as a learning kind of modality. That is new and different. The next phase of AI-trained models are really they have to have experiential learning. They can't just be trained on static data sets. We are uniquely creating an environment where that can happen and where all parties can monetize that as a commercial activity, which is huge. We're early in really, I think, tapping the value here.
We're seeing the data convert into value with that $100 million of GSV from the AI and other feature enhancements we made just in the first half of this year. There is a long runway here. Yeah, in terms of monetization, I think underlying that question is probably asking if we're going to sell our data or something like that. The data is too valuable for the work that we can do on the platform and the yield that we're going to get. We don't have any plans, nor is that included in our guidance.
All right. One more from the webcast. AI GSV annual run rate is $300 million by end of 2025. Is there a way to quantify across all of 2025 GSV or in Q3 as a percentage of GSV? On AI agents monetization, how is the opportunity different between SMBs and enterprise?
On the first, I think we've provided the quantum on the go-forward AI category. I think that's what we, and we've provided the growth rate. I think you can probably take it from there. Yeah, on the second part, I think the quantum at this point, it's really hard for us to forecast exactly what that kind of mix looks like. These are super new technologies and new experiences that in the AI agent case, we have them in testing. There's a lot to do and learn as we roll them out to both types of customers, SMB and enterprise. If there's anything, we just see a lot of upside and runway there, but kind of probably too early to give some specifics.
Yeah, we have one more from the webcast.
How can Upwork sustainably grow enterprise GSV and revenue without simply cannibalizing existing SMB marketplace activity, especially as AI tools change how lower value work is done on the platform?
These two products, our Lifted product and our SMB product with Business Plus, are going after completely different segments of the market. I mean, we've shared the $530 billion on SMB, the $650 billion of spend from enterprise. When you heard Ernesto talk about our ideal customer profile in the enterprise, these are very large companies, more than 1,000 employees, huge budgets, and the hundreds of millions to billion dollars of spend on contingent. That is really the sweet spot we're going after. Those companies look totally different from the SMB target with Business Plus. Business Plus customers are a small to mid-market company. They definitely have programmatic spend that they maybe want to move or make contingent.
We are helping them do that. I'd say the businesses are very different, so the risk of cannibalization is zero. We're really kind of just going on two paths towards two different customer types with those products.
We have another one from the webcast. Are the 2028 GSV to revenue targets organic, or do they include potential M&A?
No, there's no additional M&A included in any of our outlook. We certainly will utilize our balance sheet to enable further growth and to accelerate our plans, but there's no incremental M&A in our outlook.
Anybody else? All right [crosstalk]
I want to thank everyone. This has been such a fantastic morning. Hopefully, you have all seen what we see in this business, that it is a completely reinvented business.
The work over the last three years on our product, on our customer focus, on our operations is setting us up for incredible performance and acceleration in this next chapter. We've outlined our three major growth building blocks, each one unique, each one significant, and each one with traction. We have tailwinds in AI we are capitalizing on. We have tremendous progress with our Business Plus offering for SMBs going after a huge market. Of course, we've launched Lifted and are seeing great feedback and traction with a new model of serving the enterprise with something that doesn't exist anywhere else. We are thrilled that you've joined us today. We'll be sharing more updates and milestones as we go. Thanks, everyone.
Thanks, everybody.
Thank you for being part of our program. Please join us for lunch in the café.