Keystone Law (KEYS.L): A High-Return Platform Compounder Misread as an AI Loser

A capital-light legal platform with 40%+ ROE, trading at a 13x multiple

Outsized Returns explores investment ideas where outcomes can scale non-linearly because the market is misunderstanding the business, mispricing its durability, or underestimating how value compounds over time. The focus is on downside-aware ideas with embedded optionality and situations where you do not need to be right often, just right when it matters.

Keystone Law (KEYS.L)

Keystone is starting to look like the kind of setup the market creates when it applies a broad narrative to a specific business without looking closely enough at the actual economics.

The broad narrative is easy to understand. AI is coming for legal work. Drafting, review, research, due diligence and workflow all become more automated, so legal services deserve a lower multiple. Keystone has been caught in that downdraft. At around 455p, with consensus next-year EPS around 34p, the shares are on roughly 13.4x forward earnings. Based on the FY2025 ordinary dividend of 20.2p, the ordinary yield is about 4.4% before giving any credit for specials.

But the more interesting question is not whether AI matters. It does. The real question is whether the market is applying the right mental model.

Keystone is not a conventional law firm with a heavy junior pyramid and a lot of fixed-cost leverage. It is a platform model built around self-employed senior lawyers, variable costs, central infrastructure and very strong cash generation.

And the historic quality of the business matters too.

This is not just “cheaper than it was.” It is a business that has compounded for years. Revenue rose from £56.4m in FY2021 to £97.7m in FY2025, the group stayed debt-free, ended FY2025 with £9.7m of cash, and kept returning meaningful amounts of cash to shareholders. Since IPO, Keystone says it has returned just over £45m to shareholders, or just over 145p per share, equal to 96% of adjusted earnings generated over that period.

That is exactly what high-return, capital-light businesses tend to look like in practice.

So this is the kind of Outsized Returns setup that gets interesting: a good business, a simpler bear narrative, and a much lower price than a few months ago.

The core insight

The stock is now cheap enough that the historic quality of the business starts to matter more than the neatness of the AI fear story.

Keystone does not publish a formal ROIC figure, but the underlying economics are unusually strong. In FY2025, the business generated roughly 46% ROE on average equity, while remaining debt-free, highly cash generative, and returning a large share of earnings to shareholders. Because lawyers are paid on a paid-when-paid basis and working-capital needs are minimal, a stricter operating-capital return would likely be even higher. In other words, this already looks like the kind of 40%+ return-on-capital business the market usually does not let you buy on a low-teens multiple.

My variant perception is narrower than the market’s. AI is a real risk to parts of the legal value chain, but Keystone may be one of the better-adapted models if legal production becomes more software-assisted.

Keystone is not a traditional pyramid law firm. It recruits experienced lawyers from mid-market firms, its lawyers are self-employed, they own the client relationships, they are paid only once the client has paid, and they can earn up to 75% of billed fees. Keystone provides the infrastructure, technology, community, compliance framework and central-office support around that lawyer base. The company explicitly describes the model as resilient and highly cash generative.

That matters because the standard AI bear case for legal services is built around the old model: lots of junior hours, lots of document-heavy process work, and lots of billable time that software can compress. Keystone is structurally different. It has less fixed-cost junior bench to disrupt and more variable-cost alignment built into the model. If AI reduces the amount of junior support a senior lawyer needs to service clients, the minimum efficient size of a portable practice may fall. In that world, Keystone’s model may become more attractive to senior lawyers, not less.

The important framing is that this is not an “AI does not matter” thesis. It is a “the market may be overestimating how bad AI is for this specific model” thesis.

What Keystone is today

Keystone is best understood as legal infrastructure wearing a law-firm brand.

The visible layer is the individual lawyer relationship. The economic engine sits underneath: the operating model, billing and collections, compliance, central-office support, IT platform, document systems, collaboration, referrals, and community. Keystone recruits experienced senior lawyers who bring client relationships with them, then gives them a platform that lets them operate with more autonomy and often better economics than in a traditional firm.

That matters because Keystone is not a conventional pyramid law firm built around large junior teams and fixed-cost leverage. It is a senior-led platform model. Many Principals scale their own practices through “Pods” of junior lawyers and support, while others use Keystone’s central resources without needing to recreate a traditional team structure around themselves.

More than 30% of work comes from cross-referrals, which is an important clue that this is a real platform rather than just a collection of lawyers sharing overhead. The internal network matters. The operating infrastructure matters. The technology matters.

The work mix is also broad enough to make the story more nuanced than either the bull or bear caricature. Keystone’s FY2025 matter mix was Commercial 49%, Property 21%, Corporate 20%, Litigation 17%, Employment 12%, Family 8%, Private client 5%, and Other 2%. This is not a pure commoditised contract shop. But it is also not a pure specialist disputes boutique sitting entirely outside the reach of workflow automation. It is a mixed legal platform with exposure to both more standardisable work and more judgment-heavy, relationship-led mandates.

That is exactly why the stock is interesting here. The market can clearly see the first half of that sentence. It may be underweighting the second.

Why this is a quality business

Keystone is a quality business because the model itself is structurally attractive, not just because the recent numbers have been good.

The business is capital-light. Keystone does not need to fund a large branch network, carry a heavy salaried junior bench, or tie up large amounts of working capital to grow. Its lawyers are self-employed, they own the client relationships, and they are paid only after the client has paid Keystone. The firm provides the infrastructure around them rather than employing a traditional hierarchy of fee earners on the balance sheet.

That creates unusually attractive economics for a professional-services business. Growth comes mainly from recruiting successful senior lawyers onto the platform and then helping them scale further through brand, infrastructure, cross-referrals, and systems. In other words, Keystone does not have to build demand from scratch in the way many services firms do. It often recruits the demand along with the lawyer.

That is a big part of why the company has been able to grow while remaining debt-free and highly cash generative. The model has some resemblance to other partner-led professional-services platforms, such as Kelly Partners in accounting, where the central entity provides systems, alignment, and growth infrastructure while senior professionals remain economically motivated operators in their own right. The structures are not identical, but the family resemblance is there: aligned senior professionals, central infrastructure, and a platform that scales by making productive practitioners more effective.

The moat is limited, but real. This is not a hard-moat business in the way an exchange, ratings agency, or credit bureau might be. But it does appear to have a modest and valuable moat built around reputation, recruitment, culture, internal referral density, and operating infrastructure. Once a professional-services platform gets enough scale and credibility, that can become self-reinforcing. Better lawyers attract better lawyers. A stronger internal network creates more referrals. More scale supports better systems and support. Better support makes the platform more attractive again.

So the quality here is not about claiming Keystone is unassailable. It is about recognising that this is a better business than the market may instinctively assume when it hears “listed law firm.” It is a capital-light, partner-led platform with a sensible degree of alignment, real cash generation, and a modest but meaningful operating moat.

Why now

Because the stock is being valued more like a threatened professional-services business than a still-growing, cash-generative platform.

The recent numbers do not look like a model that is already breaking. FY2025 delivered £97.7m of revenue, £11.7m of profit before tax, and £9.7m of closing cash with no debt. Management proposed a 14.0p final ordinary dividend and a 15.0p special dividend, bringing total declared distributions for the year to 35.2p, including 20.2p of ordinary dividends.

The momentum then continued into FY2026. H1 FY2026 revenue rose to £54.2m, adjusted PBT increased to £7.3m, operating cash conversion was 104.2%, and net cash remained £6.5m even after paying both the final ordinary and special dividend during the period. Recruitment conditions also remained healthy, with 164 qualified new applicants in the half and 30 new Principals added.

Then the February 2026 trading update reinforced the point. Keystone recruited 61 new Principals in FY2026, increased total fee earners by 13.5% to 654, and said revenue per Principal was up by just under 10%.

That does not read like a model already being hollowed out by AI. It reads like a business that is still recruiting, still scaling, and still monetising the platform.

And yet the shares, at around 455p, are now trading on roughly 13x to 14x forward earnings, with an ordinary dividend yield of about 4.4% before giving any credit for specials.

That is why the setup is interesting now. The quality of the business has not obviously disappeared, but the valuation is already pricing in a much darker future. That is often where the best Outsized Returns setups begin.

The AI threat, and why it may be over-priced

There is a real bear case here.

Keystone itself warns that technology is increasing the commoditisation of services, and that legal clients increasingly expect work to be delivered more quickly and cheaply. That matters because the biggest disclosed work bucket is Commercial, where some meaningful share of activity is likely to include contracts, drafting, review and process-heavy work that AI can attack first.

So the real AI risk is not that lawyers disappear. It is that the value of a chunk of legal work migrates away from billable human time, clients push harder on price, and firms like Keystone lose some combination of hours, pricing power, and growth duration. In that outcome, Keystone may still remain capital-light and cash generative, but it becomes more of a yield stock than a compounder.

That is the serious bear case.

But there is another side to it.

Keystone is already set up as a senior-led, variable-cost model, not a classic leveraged associate pyramid. If AI lowers the amount of junior support needed per senior lawyer, that could actually increase the appeal of Keystone’s platform to experienced, portable partners who no longer need as much institutional staffing around them. In other words, AI may not just compress legal work. It may also lower the organisational advantages of the traditional firm. That could be a tailwind to Keystone’s recruitment proposition.

It also helps that Keystone is not sitting still. In H1 FY2026 the company said it had rolled out generative AI tools that allow lawyers to produce file notes from Teams meetings in seconds, use generative AI across documents held within NetDocuments, and interrogate the firm’s operating manual using an internal tool built with generative and agentic AI elements. Management framed the AI strategy around delivering real value, real-life solutions, efficiency and enhanced user experience.

That does not remove the threat. But it does make the simple “AI loser” label feel too lazy.

The part the market may be underweighting

Even if Keystone ultimately does settle into a slower-growth future, the downside endpoint is not obviously terrible.

This is a business that remained resilient even in a highly disrupted backdrop. In FY2021, which captured the COVID shock, Keystone still grew revenue 10.9% to £55.0m, increased adjusted PBT 3.6%, converted operating profit to cash at 100%, and ended the year with £7.4m of cash and no debt.

That does not prove recession immunity, but it does suggest the model has more resilience than the market may assume for a listed legal-services name.

And if the real bear case is “this becomes a decent yield stock rather than a great compounder,” that is not the worst place to start from when the stock already trades on a low-teens multiple with a 4%+ ordinary yield. The downside may be less “thesis implodes” and more “returns are merely okay rather than exceptional.”

That is not a bad backstop.

Returns Framework: thinking in IRR

This is deliberately order-of-magnitude. The point is not precision. The point is to think about what you actually need to go right.

Downside case: ~5% to ~10% IRR

In the downside world, AI-driven pricing pressure is real, growth slows, and Keystone gradually loses its premium-compounder status. Revenue per Principal softens, recruitment becomes less exciting, and the market continues to value the stock as a mature legal-services yield name rather than a platform compounder.

But even in that world, the business likely remains capital-light, cash generative, and debt-free, with a meaningful ordinary dividend doing part of the return heavy lifting. That is why the downside case is not “capital loss disaster.” It is more like a decent yield stock with muted growth and limited rerating.

Base case: ~12% to ~18% IRR

In the base case, Keystone keeps doing what it has been doing: recruiting solidly, growing Principals and fee earners, increasing revenue per Principal modestly, and converting profits into cash. AI creates noise and pricing pressure in parts of the market, but not enough to break the model. The multiple stays roughly in the low-to-mid teens, so returns are driven mainly by earnings growth plus dividends rather than a heroic rerating.

That feels like the most plausible middle ground: no dramatic rerating required, no heroic assumptions, just continued platform execution and ongoing cash returns.

Upside case: ~18% to ~25%+ IRR

In the upside world, the market gradually realises that AI is not simply a threat to Keystone but may actually increase the relative appeal of a senior-led, variable-cost legal platform. Recruitment remains strong, Principals keep joining from more traditional firms, revenue per Principal keeps compounding, and the stock gets re-rated from a depressed low-teens multiple to something more appropriate for a high-quality, cash-generative platform model.

You do not need a return to the old peak multiple for this to work. You just need the market to stop assuming that AI turns Keystone into a structurally worse business.

What I need to see next

Recruitment is the first thing to watch. If Keystone keeps adding strong Principals and the applicant funnel remains healthy, the platform proposition is probably still working.

Revenue per Principal is the second. If that starts flattening materially, the AI/pricing bear case becomes more credible.

AI adoption is the third. I want to see Keystone keep using software to raise lawyer productivity and preserve its relative attractiveness versus traditional firms.

And finally, cash returns matter. Keystone’s combination of ordinary dividends, occasional specials, and a clean balance sheet is a meaningful part of the downside-aware case.

What breaks the thesis

This is not a one-quarter miss story.

It breaks if AI compresses pricing materially faster than Keystone can offset through productivity and mix, recruitment quality deteriorates or partner inflows slow sharply, revenue per Principal stalls or falls for the wrong reasons, the legal market ends up valuing senior portable partners less, not more, in an AI-assisted world, or the market is right that legal platforms deserve structurally lower multiples because software captures too much of the value pool.

Outsized Returns scorecard

Downside protection: ⭐⭐⭐⭐½
Clean balance sheet, real cash generation, and a meaningful ordinary dividend provide a genuine backstop.

Base-case confidence: ⭐⭐⭐½
You do not need AI to be irrelevant. You just need it to be less damaging than the market fears, while the recruitment engine keeps working.

Upside optionality: ⭐⭐⭐⭐
If AI lowers the need for large junior benches, Keystone’s model could become relatively more attractive to senior lawyers rather than less. This is the most interesting part of the thesis.

Mispricing / misunderstanding: ⭐⭐⭐⭐
The market seems to be treating AI as a broad sector de-rating event. It may be missing that Keystone is one of the few models in the space that could plausibly adapt well to that shift.

One-sentence takeaway

Keystone is not just cheaper than it was. It is a historically high-quality, capital-light platform with strong cash returns that may be getting marked down as an AI loser precisely when its relative advantages are becoming more visible.

Disclosure: The author may hold positions in the securities discussed. Views expressed are personal and not investment advice. [Disclaimer]