Good morning. This is one of the perks of our job. Our creative team work with Rosalía and Sony Music to bring to life her vision. She says, "I belong to the world. The world is so connected." For last album, "Lux," she actually sings in 23 different languages, part of the song. She asked us to transcreate the poetic intent of her 18 songs in her album in 23 languages, including Latin. Rosalía is known to be often provocative, and "Lux" contains religious references. This is really hard to transcreate. You have to have a nuanced approach for markets that are culturally sensitive to these topics. Now you see our works in our concert, in online platforms, including Spotify, that has this new feature of translated lyrics. The reason I show you this is because this is really cultural intelligence at work.
It's the thread that runs through everything I'll show you this morning. Welcome to RWS half-year results for 2026. I'm Ben Faes, the CEO of the company, and with me today, I have Stephen Lamb. Sorry. We have Stephen Lamb, who joined us as CFO in March. Christina Scott, our group CTO, who you've met already in December, and Lewis Whiting, CEO and co-founder of Obviously, which RWS acquired last month. Beyond the usual updates, we'd like to give you today a view on the progress we are making in the strategy that we've announced in December. The company went through significant reorganization, and I want to share how this pivot is going. We are operating in a fast-changing environment where innovations are released daily, where our clients need more help than ever to navigate the new reality of an AI-powered world.
In a rapidly evolving market, RWS grew its revenue by an exceptional 7% organically, showing the importance of our work to our client, and also the value of internally having a lot more clarity and focus. In December, we set out three key priorities, a refreshed go-to-market, an innovation roadmap, and an efficiency plan. Six months on, all three are delivering, and crucially, they are reinforcing each other. We said we'd return to growth, and we have, ahead of expectations. We've been quite busy. The pace at which we operate has significantly accelerated. You'll hear a lot about the financial performance in detail in a second. I just want to insist on a few points here. Our technology roadmap is very strong and has achieved significant milestones.
We've signed a unique partnership with Cohere and built together a product that delivers today the highest quality translation across 30 different languages on any benchmark. Our leadership team keeps getting stronger. Stephen is bringing years of PSC experience. Brajesh, our new CEO of the U.S.A., has been leading very large-scale operation at Genpact and Oracle. Ravi in India was previously leading all offshore operations for News Corp . This isn't a turnaround on paper. It's leadership refresh, logos worn in every segment, new AI patterns, costs down, and an organization that genuinely is operating AI first. As the adage says, what gets measured gets managed. In December, we've put KPI against each of these priorities. Our top 100 clients are spending more with us, not less. Net repeat revenue of 109%. In an AI-disrupted market, that's the number really that matters. That's our largest enterprise relationships are deepening.
On the net promoter score, as expected during a strategic reorganization of this scale, engagement temporarily dipped, but our rapid execution meant that our latest pulse in May bounced back to 53, the highest in two years. A third of the business is now AI-related and growing. We have actually started two new lines of services on the voice AI front. Voice, I believe, will be the interface of the future, and the cultural intelligence that we are building is even more important in the voice environment. We're increasing productivity. Each employee is contributing to higher value creation, which in turn increases our EBITDA margin. On that note, I will pass to Stephen.
Thank you, Ben. Got my glasses. Good morning, everyone. I'm delighted to be presenting my first trading update as Group CFO of RWS. This morning, I'll take you through our strong financial performance for H1, covering a brief summary of group trading, an overview of performance in each of our business segments, summary of cash flows, our outlook, and key considerations for the remainder of the year. This slide focuses on our key financial metrics for the first half. We delivered strong organic revenue growth and double-digit profit growth in the first reporting period under our new operating model.
In December, we provided high-level FY 2026 guidance to the group, and I'm pleased to be able to say that we exceeded our expectations for each of revenue, operating profit, and cash conversion. On revenue, we delivered 7% constant currency growth, a strong outperformance, and our highest growth rate for a number of years. This outperformance was driven by an exceptional growth in the Generate segment, which was linked to a large-scale AI project to support a strategic client through a key phase in the rollout of one of their new solutions. Whilst both Transform and Protect performed in line with our expectations. On profitability, adjusted operating profit margin improved by 130 basis points, with adjusted operating profit up 28% to GBP 26.4 million in H1, compared to GBP 20.7 million in H1 in the prior year.
It's worth noting that the year-on-year growth is despite absorbing a couple of large headwinds, being the accrual of market rate bonuses this year, which were negligible in H1 2025, and adverse FX movements, which I'll come to shortly. Our operational free cash conversion, which is defined as adjusted EBITDA less working capital movements, lease payments, and CapEx, was also a little ahead of expectations, reflecting good working capital management together with focused capital investment in our new products. Turning the page. Looking at our financial highlights. Our first half financial performance resulted in total revenue of GBP 360 million, with reported growth of 5%, and as I said, constant currency growth of 7%. It's worth pausing here to touch on how currency impacts the group. Almost 70% of the group's revenue is in U.S. dollars, and as such, a weakening of the dollar reduces our revenue in GBP.
The global services nature of our cost base means that less than half our costs are in USD, and the strengthening of other currencies increases our cost base in GBP. In FY 2026, both of these factors have gone against us. We actively hedge around half our net USD exposure in the year, but we cannot fully offset these headwinds, either in the current year or the medium term. As we pivot to a technology-first model, we do expect this currency imbalance to gradually correct. Gross margin of 41.6% for the half was below the prior year, with over half of this change being driven by the outperformance of lower margin TrainAI activities, and the balance is related to product mix in the other business units.
At a net margin level, we have seen the cost actions taken over the last 12 months to continue to improve productivity across the group. The combination of strong revenue growth and disciplined cost management has delivered significant adjusted EBITDA progression, with GBP 46 million in H1, up 20% from GBP 38 million last year, and EBITDA margins expanding from 11.1% to 12.7%. Adjusted operating profit, as I've said, was up 28% year-over-year, with margins up from 6% to 7.3%, whilst adjusted EPS was up 34% to GBP 0.049, so good growth at every profit level.
Cash and net debt was also well managed, and we are declaring our first interim dividend since rebasing our dividend payout in December 2025. At GBP 0.0175, our interim dividend is consistent with our declared progressive full-year policy, and a payout at this stage of around 25%. This compares to around 20% for the prior year interims.
Just moving over. Over the next few of the slides, I walk through trading performance across each of our three business segments. Given this is the first time we are reporting using our new segments, I've also included a simple reminder of what each business does, together with an indicative size of the market. These market assessments were done by a third-party consultancy for us last year. For each segment, we've included the serviceable addressable market, SAM, which is effectively the element to which our services and products are relevant, the vended market, i.e., how much is spent with third-party providers, the medium-term growth rates, and finally, our share of that vended market. I've also included adjusted operating profit by segment for the first time. Now, turning to Generate. Generate helps enterprise clients produce intelligent knowledge and deploy trustworthy AI at scale. It comprises two complementary businesses.
Our content technology business includes enterprise-grade content management solutions, including Tridion, Contenta, and Propylon. Whilst TrainAI provides a full range of end-to-end AI data services for all types of AI. In terms of market context, the total SAM for Generate was estimated at over GBP 5 billion, with a vended market of GBP 2 billion. Although given the recent AI investment, this is likely to understate the opportunity. As you can see, the growth opportunity in AI data services is particularly exciting. In the first half, Generate delivered exceptional revenue growth at 52% constant currency, largely due to the key project mentioned previously. Despite tough comparatives from a strong FY 2025 performance in the higher margin content technology business, this resulted in 19% operating profit growth in Generate.
The key takeaway from the excellent H1 performance, though, is that in TrainAI, we've been able to demonstrate we can rapidly scale activity to meet peaks in customer demands whilst maintaining high-quality ratings and also onboarding new strategic clients. We see this as a fast-moving and evolving market segment, and are focusing investment in TrainAI to ensure that we build a sustainable and scalable business for the future. Looking to the second half, we've taken some targeted actions in content tech to improve our sales capabilities and margins. Whilst it's also worth noting that activity in TrainAI returned to normal levels at the start of the second half. Turning to Transform. Transform is our largest segment, representing approximately 60% of the group. Transform combines AI-enabled language technology with human linguistic expertise to help enterprises transform how their content is created, adapted, and delivered worldwide.
This segment also includes two market-leading technology platforms, Trados, which is our translation management solution, and Language Weaver Pro, our amazing new AI-powered translation solution, which you'll hear about shortly. In terms of market context, the SAM for language services and technology is approximately GBP 17 billion to GBP 19 billion, with around 2/3 of this outsourced to specialist language service providers. The broader language services market is declining in low single digits, but we expect this to be offset by strong growth in technology and automation.
Despite being a global market leader, our revenue of GBP 450 million represents only around 4% of a fragmented global vendor market, highlighting the significant headroom remains for us, albeit our focus is on the large enterprise clients where we can really solve their complex needs on a global scale. In H1, Transform delivered revenue of GBP 210 million, which was a mid-single digit decline as previously guided.
Demand for services revenue remained robust, pricing is holding up, and we continue to win onboard new customers, and increased our share of wallet with existing clients. We are continuing to invest to pivot to a technology-first approach. The launch of Language Weaver Pro at the end of the half and the development of our new translation management platform is also on track. These will form the foundation of our shift to a recurring technology revenue model over the medium term. The Transform segment also benefited from increased productivity reflecting our cost-saving actions, and we are really pleased with the 17% increase in adjusted operating profit. Looking to the second half, we expect to pick up in both revenue and net margin in Transform due to a combination of higher margin seasonal projects, initial sales of Language Weaver Pro, and further productivity benefits. Finally, turning to Protect.
Our Protect business helps enterprises across their intellectual property life cycle from initial idea to commercialization. This enables them to safeguard their brands and maximize their returns from their innovations. In terms of market context, the overall IP services addressable market was estimated about GBP 3 billion per annum, with a global vendor market share of about GBP 1.5 billion growing at low to mid-single digit rates.
I should add that we believe the Obviously acquisition adds a further GBP 2 billion of addressable market opportunity. Our revenue of approximately GBP 100 million represents around 5% of the vendor market share, and we are really well-positioned to increase our share of this, particularly as we broaden our offering. In H1, we delivered revenue of GBP 51 million, representing a growth of 3.3% at constant currency. Within that, the mix was impacted by higher contribution from our renewals products, which carries a lower margin profile.
We also secured a number of new client logos with notable successes in both APAC and the Middle East. We saw a planned increase in our cost base as a result of both investment in our teams and sales capabilities and our processes and technologies. Looking to the second half, we expect Protect to show improved margins with a focus on selling higher margin products, onboarding new customers, and simplifying our delivery model. Clearly, we'll also be focusing on welcoming our new Obviously colleagues and building a combined sales pipeline, and you'll hear from Lewis about this shortly.
In terms of financial contribution, we expect Obviously to have a GBP 1 million negative impact on PBT this year, after cost of funding, before becoming profitable at an operating profit level next year, probably break even after funding costs, and making a strong net profit contribution thereafter. Moving on to cash flow.
We started the year with GBP 25 million of net debt, excluding leases. The adjusted EBITDA contribution for the first half was GBP 46 million, reflecting the strong trading performance discussed. Working capital saw an outflow of GBP 5 million, which was broadly in line with our revenue growth trajectory and reflected a solid working capital performance for the period. Total CapEx payments of GBP 7 million in H1 reflected both investments in new products and process automation, and the exceptional cash items totaled GBP 11 million in the period, mainly relate to further payments for productivity investments, both people and IT related.
We also paid the final FY 2025 dividend, which together with the above, brings us to a period-e nd net debt position of GBP 33 million. Net debt to adjusted EBITDA level is below 0.5x after the Obviously acquisition actually, which we paid GBP 16.5 million just after the period end.
Finally, just looking to our outlook. As noted in our interim announcement, whilst we expect profit in FY 2026 to be H2 weighted, it is much more balanced than we saw in the prior year, which was 30/70 split, and that reflects a stronger H1 performance. The H2 profit weighting reflects a combination of the seasonality of the Transform business, good sales pipeline momentum, and further benefit of productivity initiatives coming through. Given we expect the TrainAI business to moderate in H2, we expect the group to return to a normalized level of revenue growth. Across the full-year, this is expected to result in mid-single digit organic constant currency growth. I've also got a few messages to call out on full-year profit expectations. On trading profit, we are not expecting any significant change to current expectations for our full-year outcome.
Clearly, the exceptional TrainAI revenue will dilute group wide margins for the full-year and provide an interesting headwind for H1 next year. We expect to be on track overall. On non-trading activity, currency will be a full-year headwind at current rates. When we set out our expectations in December, the GBP/USD rate was $1.32. Since then, the dollar has weakened to around $1.34 to $1.35. We managed to hedge about half of this in year, but this is still expected to represent a GBP 2 million-GBP 3 million headwind on net profitability for the full-year based on the currency explanations I gave earlier. Clearly, it will also depending on whether rates land impact FY 2027 and beyond. As noted earlier, we expect a small negative impact this year from the Obviously acquisition, but this will be building great momentum getting into FY 2027.
Finally, I've included some further balance sheet details and modeling guidance in the appendix. With that, I think I'm handing back to Ben.
Thank you. Great. Thank you, Stephen. I suggest we dive a bit into each of the pillar of our strategy and illustrate really a bit more clearly what we've been doing. At the beginning of this year, we set ourselves three strategic objectives. I'm a strong believer of the power of focus and executing at pace. We revamped our go-to-market to focus on key growth segments, and we are at an extraordinary moment in the industry where enterprises are rethinking everything they do. To do that, they need to rely on strong partners they trust can bring the work they need, but with the fuel of AI. I will walk you through some of the examples of this work we're doing for our clients across our segments.
We rebuilt the technology team and the product team, basically from the ground up, to do two things fundamentally, bring innovation to our clients and build the internal AI platform to consolidate our assets and help us scale faster. Christina will explain more details the incredible achievement, for example, of launching Language Weaver Pro in nine months, which is now the most advanced enterprise translation engine by any measure. You'll hear as well from Lewis on the story of Obviously and hopefully understand how powerful an addition it is to our Protect segment. Finally, we started a very strict efficiency program, which is delivering visible results. We have a clear plan to continue this throughout 2026 and 2027. This is really a combination of modernization of our workflows, significant automation, and reliance on more offshore skills. Stephen will share more details about this.
The work we do is really hard and really important for our customers. Really hard because it requires complex orchestration of technology solutions and very skilled human intervention at scale. Really important because it's work that helps our clients grow their revenue, reach new customers in a safe and compliant manner. I'll share with you three examples to illustrate this point. In December 2025, one of the most prominent AI companies revolutionizing customer service by building bespoke voice AI agents came to us with a very hard brief. They wanted their agent to sound like home in the U.K. This meant being able to spin up agents that could adopt local accent from 12 different regions in the U.K. They tried that with others. They failed because it's really hard to do.
Anyone will say they can, but in reality, it means casting hundreds of people, building scripts, building an operating platform that guarantees the security of their data. That's what our TrainAI team delivered. Today, we operate this with these amazing clients in many more countries, helping their agents sound more human everywhere, more like home everywhere. The second example is from probably one of the most successful fintech company in Europe today. They are expanding incredibly fast, and they needed to adapt their video ads to 11 new markets. In this occasion, we work initially with their creative agency. We delivered 200 different assets localized using our production platform with carefully chosen artists for the dubbing of their videos. Our solution was comprehensive end-to-end managed service. We didn't just execute, we took full ownership of that work.
This included managing voice talent, localizing scripts, leading recording sessions and acting as a guardian of their brand throughout the process. The client was so pleased that we expanded even further. We now work directly with them. The following year, we delivered 350 assets for them. Finally, this last example is from our Protect segment, and more specifically, the work through our new platform, Obviously. In the previous example, I show how RWS is assisting client expand their brands abroad. When those brands are successful at doing this, they often expose themselves to counterfeit. Counterfeit is a problem and potentially can destroy their brand. Beyond the missed economic opportunity, it is potentially destroying the image of the brand, exposing customers to bad quality and even dangerous product. Obviously helps fight this.
With this incredibly successful plush toy company, which is growing at exponential rate, they get copied a lot. Fighting counterfeit is a bit of a whack-a-mole game. That's where AI comes quite handy. Obviously's vision AI technology is able not only to recognize the brand attribute, but also the detail of the product and the key features of that product. They are able to make a judgment call to decide whether it's a true or a fake product to initiate a take down notice of the counterfeit. The system is actually so good that it makes counterfeit economically unviable. Counterfeiter move to the next brand, potentially. Finally, we don't just sell AI solution, we run on them.
This is a brilliant example of how we've modernized our marketing function at RWS. In less than six months, with actually half of the team that we used to have, the team has achieved phenomenal results. Today, we're the number one seated localization brand across AI search engine, well ahead of any competitor on generative engine optimization. We've replaced also five tools creative tech stack with one, which represent really an 80% saving for us. We've also built a set of very smart AI agents used by our sales team, working 24/7 to help them perform better in their job, anticipate demand, understand customer needs, detect RFP signals before they are even published. This is the AI-first culture that's spreading in each function. I hand over to Christina to show you more about the technology.
Good morning, everyone. I spoke to you in December and set out our new product vision and how I'd restructured the product and technology team, and also disclosed the deal with Cohere. I'm delighted to say that we're on track with delivering our new products. Today, I want to really focus on the Cohere deal that we did and that great partnership that Ben has mentioned a couple of times already. In March, we launched Language Weaver Pro, which is built on Cohere's Command A+ model. This model was purpose-built for enterprises, and it has 100 billion parameters. To put that into context, a small model of, say, less than 10 billion parameters would be like a junior linguist who is proficient in one domain, but they would guess outside of their domain.
100 billion parameter is like a senior linguist, decades of experience, who is proficient across finance, law, marketing, technology, understands regulatory texts, and knows when not to improvise. As the world of GenAI has evolved, enterprises are beginning to understand the opportunities, but also the pitfalls. I was at London Tech Week yesterday speaking to a lot of senior leaders, and there are a number of challenges in really truly adopting AI at enterprise level. The truth is that generic LLMs are simply not good enough for enterprise use for translation. While powerful, they just lack the nuance that enterprises really need. We see four main barriers to entry to adopting LLMs for localization. The first barrier is quality. How do you get consistent quality across languages, across inputs, and that you ensure that the fluently written outputs are actually accurate?
Customers try to solve this by using lots and lots of input in the prompt engineering, then suddenly you're processing 500 tokens to translate 10 words, and the cost is going up exponentially and the processing is slowing down. The next challenge we see is scale. How do we provide the throughput that customers were able to achieve through neural machine translation? LLMs were typically much slower to do the processing. We have one customer who translates 1 million words per minute. How do we meet those sort of expectations, and not just by throwing more compute power at it, but how do we actually do it by achieving more with less? The next area is security.
How do you offer the quality and scale I've mentioned without passing all your data over to the APIs of the foundational providers, and therefore breaking some of the guarantees you may have given your customers around privacy? A number of our customers today work on on-premise solutions, and particularly are popular obviously with governments and legal companies, financial institutes. The last area people are really worried about is safety, and it remains a big challenge. It's not just the biases that are built into a model, but it's also ensuring that the output is accurate and true to source. Hallucinations can be a simple odd extra word, so not so harmful, or it can be something that sounds really fluent and confident, but it is totally wrong.
When we did the deal with Cohere, we didn't just license a model, we forged a strategic alliance with them, and we co-developed and fine-tuned their model. On top of this, we've built the product and we've added the capabilities to offer something that is unique in the market, and it addresses the shortcomings of the off-the-shelf models. Let's just think about how this has broken down the four barriers I've described. To address the quality issues, we worked with Cohere on training and the evaluation, both machine and human evaluation, and making sure that we'd created something that was the best quality available globally. We chose different languages, different domains, working at paragraph and sentence level, and removing English as the target or source.
This quality is built into the underlying model, which means that you can build a lot less into your prompts, ensuring that you can increase your processing speed. The next challenge is scalability. Most models are just too slow to be able to handle the 1 million words per minute request that our customer needed. Cohere architected into their model the ability that the input gets routed to just one part of the model, an expert in the model, and therefore reducing the processing time it takes. Most frontier models would need a whole data center to host, meaning that it's cost prohibitive for enterprises to go to that solution. Our model runs on just 2 GPUs, allowing us to offer a hosted cloud solution, a private cloud solution, or a fully air-gapped solution.
This is the only product on the market allowing enterprises this flexibility deploy exactly as their security policies dictate. We have our cloud solution with zero data retention, or you can have your own edge solution complete within your own environment. Again, particularly important for many of our clients who need to ensure that they're keeping control of their data. To ensure the output is not just of the highest quality, but in terms of fluency, it is true to source, we've developed a series of hallucination and over-generation detection capabilities. These work to ensure the output is safe and can be trusted by our customers. Language Weaver Pro is a product, but it also produces the machine translation capabilities across all our language technology portfolio. The result of this partnership has created a really unique and marketing-led product.
An evaluation shared by Cohere on LinkedIn a couple of weeks ago shows that the Command A model was second, beating all of the competitors, so Google and everyone else. Actually, RWS is at number one, and that's because of the capabilities we built on top of their model. As we look to the future, our partnership with Cohere will remain a cornerstone of the strategy. It means developing further models, whether it's specific for industry or whether it's a 1 GPU model for efficiency, we will continue to push the boundaries of what's possible with AI to deliver the solutions that will empower our clients to navigate the complexities of a global market. I'm now delighted to hand you over to Lewis to talk about the Obviously platform.
Thank you, Christina. Good morning, everyone. Lewis, Co-founder and Chief Executive of Obviously. IP is hugely valuable. AI is accelerating its pace and growth, and where content has never been easier to create, our clients have to work harder and spend more to keep their own unique propositions clear in a crowded landscape. We started Obviously with a simple belief that the right mix of expertise, understanding of the market, and new technology could create something that is better for the IP industry. A way to manage IP, trademarks, patents, commercial data together in one place, far more efficiently than ever before. We had to invest heavily. We built three connected solutions into one platform. This started with my career as an IP attorney. I trained, I qualified as a trademark attorney, and we built the Obviously business out of an award-winning law firm, Stobbs.
As such, this is an industry that we really do understand. This deep domain expertise and our first-hand grasp of the problems that our clients face, coupled with our talented engineering team, has really helped us to solve our clients' problems at scale. Through doing so, we've built, in a relatively short period of time, our own brand, the Obviously brand. We've worked hard to meet the needs now of around 50 of the world's largest brands across tech, financial services, fashion, entertainment, consumer goods. We've made sure that our tech works across all of these different challenges. Looking at the fit of Obviously and RWS, couple of things stand out for me. RWS' sales team, their customer success team, and the scale at which they're able to solve really complex problems for the world's biggest brands is super exciting for the team at Obviously.
In return, Obviously can deliver an accelerated ability to support the roadmap from a tech innovation point of view at RWS and bring our ability to quickly deliver AI-native products into effectively the same customer base that RWS currently serve. One month in to life at RWS, it's been really great to see the teams working really closely together. We have incredible people at Obviously and really incredible people at RWS, and this has given me increased conviction that we can continue to attract world-class talent, and we can drive forward with our plans to scale the most comprehensive AI-powered IP productivity platform, and continue to serve all of our customers across the full IP life cycle. What is Obviously? Clearly, being an IP professional, we take this for granted, but I'll spend a bit of time to tell you in a bit more detail what it is.
As I said, one platform with three connected parts built by IP experts, powered by AI. If you take one of our customers, think about one of the largest brands in the world, it brings all of their intellectual property needs, their management into one place. It breaks down the silos between departments, finance teams, legal teams, brand creation teams, all of the same teams that RWS are serving. Creates a single space for them to manage their most valuable assets, and it provides something fit for purpose for their continued journey from an AI adoption point of view. Our modular approach across the three pillars gives clients the ability to start with us in one area and grow our relationship across the different services and technologies that we offer. Where the solution is highly automated, it scales really well.
That revenue, by nature, is reoccurring. Quickly to move through the product. Obviously Manage, this is the home of our clients' intellectual property. It's the main system of record that brands rely on as the single source of truth. It's the operating system. It's the key component that allows us to replace old tech, siloed tools, manual processes. Importantly, from an AI point of view, that system of record provides the context that then drives agentic workflows, AI automation, and revolutionizes, honestly, how our clients are able to work within this global ecosystem of protecting intellectual property. Obviously Protect, Ben mentioned, is our AI native, I would say industry leading, brand protection, anti-piracy, content protection system. We pioneered our own computer vision.
This is using the kind of legal concepts of what makes a brand strong, legal concept around how you protect your brand and the identity of your corporation, coupling that with machine learning techniques to create really powerful tech. As Ben mentioned, it's AI that understands the DNA of a client's product. One of our clients, famous manufacturer of shoes, they have a very distinctive yellow stitch, a real-life example is building AI that can detect a yellow stitch running across the welt of a pair of boots, and then think through how that would scale to process millions of images a day to find copies or dupes or replicas, lookalikes of that same product. Finally, Obviously Discover. This is a really interesting area for us. This is about connecting commercial data with IP rights. Naturally, this happens day by day anyway.
This is the real world. People sell products. Those products are protected by IP rights, that generates revenue for our customers. The Discover platform connects those data sets together and allows legal teams to make the same data-driven decisions that the rest of their organizations rely on across all of their different departments. In combination, use of Manage, Protect, and Discover provides a very powerful engine when it comes to our clients' ability to see where their brand lives and act where it matters. Hopefully you're all seeing where this opportunity leads us to. It fits right into the strategy that Ben and Christina shared with you back in December. Obviously fills a trademark and brand size gap in the Protect segment. As you know, the Protect segment has a deep heritage in terms of its patent capability.
It's where I believe the business started, it's never served customers from a brand and trademark point of view. That's now changed. Where's our focus? Accelerated delivery. The work undertaken by RWS to build the AI-powered IP attorney productivity platform, bit of a mouthful, fits extremely well with the Obviously technology. From the work we've already been able to do in May, we can say that in combination, working together, we can speed up the existing roadmap and deliver growth sooner. The experience our engineers have in terms of building AI native solutions for this industry is also invaluable in terms of reducing the risk of that rapid pivot or program of development that we've set out to the market. What this means is growth and market expansion. Naturally expanding into a new segment of intellectual property comes with it new addressable markets.
The trademark space itself, the online brand protection space, and across our wide range of service offering, we unlock an additional market of GBP 2 billion. In addition to that, we can also use our technology to better serve customers in relation to the existing IP services delivered by the Protect segment. That unlocks, again, further addressable market and increases the amount of customers we can serve. Finally, on to a real guiding proposition for us, brand guardianship in the trademark life cycle. Brand guardianship really matters to our clients. Brands take a lot of energy to create, nurture, grow, reach customers, and clients use those brands to ultimately drive revenue. It's an existing proposition of RWS that customers create content in Generate, localize in Transform, and protect in Protect.
The step from patents into trademarks mean we can continue serving customers through now into the brand and trademark space. Why this is really exciting, where that content and localization work is brand heavy, and I would say for a business like RWS serving 85 of the top 100 or biggest brands in the world, much of that work is brand heavy. By continuing to be able to deliver services throughout the life cycle and beyond the current capacity of the business, that will naturally lead to increased growth. I think that's covered all I wanted to share with you today, but generally to wrap up, I think the most important message for me is this acquisition allows us to continue solving some of the world's largest and most complex problems at scale, and yeah, happy to take questions later. Thank you.
Thank you, Lewis. Such a great business. I have one slide on efficiency, so just to share this with you now. As I mentioned earlier, a key part of our H1 trading performance was driven by the delivery of our productivity improvements. This is actually becoming a core part of the DNA at RWS. There is no single lever we can pull to reduce our cost to serve, but I set out a few examples here to show how we are achieving this goal. On people, as technology improves, we've been able to simplify and automate more, and this has allowed us to reduce our overall number of FTEs by about 6% in the last 12 months. You see on the chart here the improvement of revenue per FTE. We've also been making more use of our offshore shared services hubs to standardize processes in lower cost locations.
On products, we are simplifying our portfolio through a program to actively end of life, end of sale legacy platforms with a significant progress being made, transitioning towards our core Trados solution. On IT, we have a global program to rationalize our legacy infrastructure, and that's allowing us to move to a more secure, more scalable and efficient cloud platform across the group. Not only does this save in IT, but it also makes it much easier for us to exit a large number of our legacy locations. Finally, we're also championing the use of AI across the business. We've given AI tools to our teams, challenged them to learn and encouraged them to show off their successes. Ben has already mentioned our achievements in marketing, but AI is now starting to benefit all areas of the business.
One notable success has been how much faster we've been able to develop our new products across the group, at significantly lower cost too. A brief summary on productivity, with that, I shall hand back to Ben.
Thank you. Let me close for you. When I stood before you six months ago, I didn't just outline a vision, I laid out a strict blueprint for this company. A new go-to-market strategy, an aggressive innovation roadmap, and a rigorous efficiency plan, all tied to hard KPI. Today, you're seeing the result of that execution. We delivered 7% organic growth, significantly ahead of expectation. Profits are up 28%. Our top 100 clients are spending 9% more with us. In a market where the consensus was that AI would shrink this industry, we're proving the exact opposite. We aren't just participating in the AI revolution, we are leading it. We built and launched Language Weaver Pro in just nine months, today it outranks the engine built by the biggest tech giants in the world.
With the acquisition of Obviously, we are forcefully expanding our footprint into a GBP 2 billion market that we didn't even play in last year. Internally, we are at an home-based case study. We're running an AI-first playbook across every function, starting with the marketing that I showed you. This disciplined operating model is exactly why we can accelerate growth while driving margin expansion. On the outlook, we've upgraded our revenue guidance to mid-single digit growth, whilst our margin will still expand this year. The mix is shifting, it's shifting because we are aggressively capturing market share with TrainAI at an exceptional rate. More top-line growth, higher profit, and rising margin will make that move every single time. This company is executing at a pace it hasn't seen in years. The strategy is locked in. The culture is revitalized. We said we'd return to growth. We did.
Now we intend to win. Thank you very much. I'd like to invite my colleagues for Q&A. Katie?
Yeah. Katie Cousins from Shore Capital. I've got two, if you don't mind. Thanks. Katie Cousins from Shore Capital. Two questions from me. First on Obviously, talking about the GBP 2 billion TAM, just kind of what's your market share of that and the competitive landscape, and also if there's a crossover of RWS clients that you feel that you could upsell and make better use of the customer wallet. I'll go with that first.
Thank you. I'll start, and Stephen can chip in if you'd like to. Obviously is a small company. We focus hard on innovating, investing, building the technology. Our market share is small in terms of the competition, but our scale and breadth is-- or our breadth rather, is large in terms of where we can service the market. Yes, there is a really strong overlap, both in terms of our existing clients, the clients in the Protect segment, but also importantly the clients in the Transform and Generate segment. We know our product serves those clients well, and we have lots of use cases in terms of serving similar clients already in the business. The competitive landscape is interesting. There's a good market there. It's an interesting market to play in.
As we said, AI is pushing forward in terms of the necessity of our products to the clients.
I would say what triggered us to build the Obviously business in the first part is our clients' frustration over legacy tech. What we benefit from is a fresh business that's built this product from the ground up in a relatively short period of time. By definition, that means our tech is really new, it's innovative, and that's where we compete most effectively with our competition.
I think actually the only thing to add is that we did this deal in record time, we did that deliberately to land just ahead of the trade show, the biggest trade show for the industry in the U.K. I think over that weekend, we generated some phenomenal leads, both from Obviously clients to RWS products, RWS clients there, but also a much broader set of global clients that we can now target as well. It was a really exciting opportunity. I am really hoping my guidance on it is very conservative.
Great. Thank you. The second one is just around the dividend. If we think about a standard phasing of a third, two- thirds, doesn't look like it's going to get to that progressive stance without either a lot higher second half or final dividend, also that implies a higher payout ratio, which I think is probably not consistent with your rebasing previous language around being aligned to earnings.
After the rebasing in December, I think we settled to GBP 0.07 for last year in total. Traditionally, we've paid 20% in the first half and 80% in the second half for some historic RWS reason. This time around, we are expecting a progressive dividend for the full year compared to that GBP 0.07, we've gone with GBP 0.0175 as a conservative quarter. We're kind of nudging in the right direction is probably the way to think about it should be progressive for the full year.
We can use that GBP 0.07?
Yeah, GBP 0.07 is the base.
Okay. Clear setup. Thank you.
Thanks. James Beard at Deutsche Bank. I've got two questions, please. Firstly, on Language Weaver Pro, can you talk about the early progress that you've made in terms of marketing and selling that product, and also how we should think about that product in terms of differential, I don't know, pricing or service versus previous iterations of Language Weaver, and therefore how this product might contribute to future growth expectations. Secondly, on TrainAI, I'm going to ask a question that I think I asked about a year ago, and see how different your response is now. A year ago, I wanted to get your perspective on the longer-term viability and sustainability of the revenue model within TrainAI. What has happened in the last year to have changed, if anything, your viewpoint on that?
Obviously, we had in H1 what appears to be a very large contract project, which has since finished. Is that the sort of catapult to further growth? Just interested to get some more thoughts on that as well.
Certainly. On Language Weaver Pro, it is a software solutions that takes a bit of time to install. It is not just a click to play. We have already signed a few of our customers. There are two motions. One is upgrading our existing customer from Language Weaver, and that is in terms of price, roughly double the price, the Pro version. A great upsell opportunity. On the new customers, we have now a very strong pipeline of people that are interested because they want to try it, because they are curious about this benchmarking quality. More importantly, from what Christina was saying, we see a lot of company now being a bit more wary of their token consumption. If you translate hello on Claude, it will burn through a lot of tokens that you do not need to burn through because it is not meant for purpose.
Those are the two reason why we see a lot of attraction and upsell potential. On TrainAI, I remember your question. I remember my answer. I feel actually a lot more confident today about the perspective. That is confirmed by all the market studies on the data services segment. I think the two elements I would like to point out, one, you are probably aware of the CapEx level of the large LLM builders. I think last year when Google said they would spend, I think it was GBP 60 billion last year, everybody thought it was outrageous. They are going to spend GBP 180 billion this year. That is meant to deploy more AI models. There is a commitment from more builders to invest in future models. We have signed actually two more foundational model builder so that we are diversifying also our customer base.
I think the new models are fundamentally harder to train because they are multimodal, they are voice and that requires a whole level of input that is hard to find on an open source model. All the models trains on text on the internet. The voice and the spatial models are much harder to train. We are investing a lot in what is called the world models and tomorrow, the robotic agentic also.
Tom.
Thanks. Tom Callan from Investec. I have got two as well. Just on the headcount reduction, during the period, I think you took about 500 heads out of the business. Can you just give a bit more color in terms of where those reductions actually landed, and also where you think headcount might be by the end of the year? Just on Transform for Christina, actually. I just wondered, to what extent do you think clients at the moment are paying for that sort of outcome risk piece as opposed to just words or licenses, and how do you see that potentially evolving over time? Do you think more clients are going to turn to you guys for absolute guarantee that they are getting a safe translation and an accurate translation, and therefore can push back on you if the translation is inaccurate?
The headcount savings are, they're spread quite broadly across the business actually, as I touched on. We've made some productivity improvements in our translation area, in our client services area, in some of our overhead areas, and also in terms of our business product development areas as well. There's no one go-to spot. I think, as I said, it's in our DNA. We are making the right decisions. We're not doing anything that doesn't feel right for the wider business. I think the headcount will slowly tick down over time, but we've not got any dramatic changes coming.
Just on LXD specifically, to use the old terminology, I think we were at what, 1,500 heads at the end of last year, something like that. Where are we today on that?
It's a little down.
It's down also. I can't give you the exact number, but it's down. Yeah.
Yeah.
If I can pick up the second question. It's obviously very dependent. There's lots of different use cases, for use cases where enterprises really care about quality, that is exactly the conversation that we're beginning to have. With the testimony of new products, obviously there's more products than just Language Weaver Pro, we are speaking to early alpha customers and testing with early alpha customers, some customers are already making that transition and thinking in that way. Others will be further behind. What we think it will be a phased approach. There'll be early adopters who we are now working with to make sure our solutions are fit for purpose, we see the rest of the market will start to move in that direction.
James.
Hi, all. James Bayliss from Berenberg. Three, if I may. On the Cohere partnership, can I ask how the unit mechanics work? The output model's yours, but the development was with their input. How does that all come together as you look to push that out into the wider market? Perhaps specifically on that one, given it's their own model, sorry, they have their own model. Is there competition there, or are they now fully kind of pushing the RWS output, given it is superior to their own offering? Lewis, one for you perhaps. I thought that was a really clear run through of the strategic fit and the opportunity for Obviously as part of RWS, but I'd expect it was quite a competitive space when you were looking at who the kind of the natural owner for the business might be.
Is there one kind of overriding reason why you went with RWS as opposed to anyone else? To round that off, one for Stephen, just a quick one on CapEx. I think your guide's evolved from a modest increase as a percentage of revenue back in December to a modest decrease this year. Is there anything specific to call out on that, or is that just evolution now that you've started executing the strategy and you're landed in the business? Thanks.
Should I start with the Cohere? I think the important thing to remember is that the models, the LLMs, they are not a final product. What Cohere do are they build a model, we've built a product, and it's the product that we're selling to customers and that's a really important distinction. We worked on Cohere on improving the model, but the model is completely hosted for us, and we have built our product on top of it. They can, the Command A model, there's some exclusivity clauses in the contract. They can sell it to other people. Actually it's the product that is most valuable. All the hallucination detection, identity access management, all the things that are built around the outside, those are the things that make it really important for enterprises and fit for purpose for enterprises.
Lewis.
From an Obviously point of view, yes, the process was competitive, and it was a great experience in terms of looking at the competitive landscape through that lens. Why RWS? Why was that such a compelling choice for Obviously and for RWS? Really, I think it comes down to the fit of the two businesses. From a competitive landscape point of view, Obviously has driven hard to change the way that our clients operate. From a legacy existing competition point of view, we have the choice in terms of being acquired by one of those legacy players in our market or moving into a business where actually the vision is so clearly aligned, but we do not have competing legacy technology. That offers a really fresh approach to our clients, to our team, to the market. I think that probably answers the question in a snippet.
Yes, it's a great result for everyone involved.
Brilliant. Thanks.
CapEx, I think probably the biggest change, I think we spent about GBP 7 million in the first half this year, is just focus really. We're spending less than we were on back office systems. We're spending less than we were on IT platforms because we're moving everything to the cloud, and we're being very disciplined in where we put our dollar in terms of investment in our product development. Questions online?
Questions online. The first question is from Chris Morrison of Jupiter. What is the profile of exceptionals and how long will they last?
Yep. We are guiding to around GBP 25 million this year in terms of exceptionals. The two big items in there are the people related exit costs for the headcount reduction we touched on, and also there's a big element of the IT re-platforming to the cloud. I think that will continue into next year, possibly ticking down a little bit, and then we expect to see it slowing a bit after that. I do think as we continue this journey, we will continue to evolve.
Thank you. Next question is from Gareth Evans from Progressive Equity Research. It's interesting that a lot of what you do relates to large customers building trust and belief in their own brands. Could you talk a bit about how you're evolving RWS to make sure that these customers trust RWS to help them manage their AI implementations?
I think I would reframe. RWS, one of the greatest assets that we have is this trust from large customers. We have relations that span 20, 30 years with some of the largest brands around the planet. That trust comes from this very, very high benchmark of quality that RWS has always delivered in core localization or IP management. I think we are building on that trust platform. We are very careful on how we deploy new product because we do not want to risk any of that trust we have built over the years. This is one of the strongest assets that we have in the company, is the long-lasting relationship with some of the most prestigious brands.
Thank you. Another question from Gareth. Post the acquisition of Obviously, please could you talk a bit more about your future M&A pipeline?
I do not want to talk in detail about the M&A pipeline. It is a moment of massive change. I think we see two things. One, as we build this internal AI platform, that is becoming more fit for purpose to be an AI consolidation platform. We are adding more services into that platform to help us scale without increasing cost. There is an appeal there. The second one is, we will make an informed decision this time. Acquiring Obviously, we acquired momentum. They demonstrated that top brands were voting for their solutions. That is a much faster access for us to the trademark industry than building a platform on our own. There are, I think, a lot of appealing technology out there. Getting to our list of customers is something that is really hard for a new tech company. I think that there is probably more opportunities there also.
Thank you. Another question from Hamish Adam from Patchwork Investment. If Cohere was to be acquired by a larger model company, where would that leave the business relationship with RWS?
We've built some very strong wording in the contract that guarantees us this success for many years. At the same time, we are also working with other foundational model. The architecture that Christina has built allows us to have this level of agility and flexibility also in how we use different models.
Thank you. A last question online, from Hamish again. You've answered some of this on the TrainAI side, I think. There was a question around whether you might be sort of sidelined by clients once those models become stronger. He's asked, once there are lots of AI models with, for example, flawless regional accents, how do you continue to generate fees for that work now that future models and clients can copy from what's already out there? Can you structure the deals as royalty deals instead of one-off fees?
I guess there's three element. One is we are building also a unique data sets for ourselves, unique methodology, which is building on our IP. The second thing is the work that those companies are asking is really changing. The voice work that I mentioned didn't exist a year ago. I think that we're building more a capability of finding skill set to render a task rather than building a specific thing that will be obsolete in the future. To that point, I think the third element is, let's imagine a world where, okay, AI is built and there's no more training needed. The trends that are growing more and more are ones of traceability of AI, inspectability of AI. We already work with some company on helping them test and inspect the AI output rather than help in the input.
AI is going to be central to our lives, if it's not already. The services that need to be built around how we manage AI at an enterprise level, I'm not worried about the future.
That completes the questions online.
Thank you very much. I think we should close that meeting. Thank you