Highlights
- Legal AI use has nearly doubled this year, yet most firms still can't prove the investment is working.
- Two-thirds of clients are optimistic about AI, but only around half of law firms share that confidence.
- The winning firms build cross-functional AI ownership and measure impact in language finance actually understands.
The legal sector has crossed a threshold. The conversation is no longer about whether AI works; it is about who wins with it.
Legal AI use has nearly doubled this year, and the more than 400 senior legal leaders who gathered at the Thomson Reuters Legal Innovation Summit in London had already moved past the debate. Over an afternoon of candid peer conversation, live product demonstrations, and expert keynotes, what emerged was something more valuable than another forecast: a set of practical frameworks for what winning with AI actually looks like, right now, in practice.
The question keeping legal leaders up at night is no longer technical. It is commercial. Specifically: can you prove the AI investment is working to the person controlling the budget?
That single question is where most AI programmes are quietly coming unstuck, and where the most instructive lessons from the Summit begin.
Jump to ↓
The ROI gap that is stalling progress
What the data reveals and what it means for you
Three frameworks that are working at scale
The architecture that makes the difference
The ROI gap that is stalling progress
By mid-2026, a quiet pattern has emerged across the industry. Pilots have been run. Use metrics have been gathered. And yet dozens of AI initiatives have stalled at the point of scale, not because the technology failed, but because the business case was never properly constructed.
The organisations moving forward share one discipline: they established clear, measurable success criteria before a single tool was deployed. At the Summit, Thomson Reuters General Counsel Bertil Majer and DAC Beachcroft Chief Operating Officer Marie Armstrong were candid about what that requires:
- Set baseline performance metrics before go-live, so you have something to measure against
- Define what “good” looks like across specific task types, not just broad use rates
- Build reporting disciplines that speak the language of finance, not just legal operations
How AI changes legal practice starts with confronting this measurement gap. Those who do are finding the ROI case compelling. Those who don’t are finding their programmes quietly deprioritised when budget cycles come around.
The measurement challenge, however, is only part of the picture. The broader strategic context, including what the market research actually shows about where the profession stands, reveals an equally pressing problem.
What the data reveals and what it means for you
The Thomson Reuters Institute’s 2026 AI in Professional Services Report surfaces a gap that should concern every legal leader:
Two-thirds of clients are optimistic about AI. Only roughly half of law firms share that view.
The client side is setting the pace, and that gap is applying real commercial pressure to firms and in-house teams alike. Clients now expect AI-enabled service delivery and greater productivity. Legal expertise alone is no longer sufficient.
Yet firms are retreating into silos at precisely the moment when their interdependence matters most. Firms are building AI strategies without full visibility into how their clients are deploying tools. In-house teams are evaluating external counsel partly on AI capability, often without making those criteria explicit.
The research also identifies five archetypes for how in-house legal departments will look by 2030:
- Scaled enablement: AI handles volume work; lawyers focus on strategy and judgement
- Advisory-plus: expanded scope into business advisory, enabled by AI productivity gains
- Tightly connected working: deeply embedded with outside counsel through shared platforms and data
- Specialist hub: concentrated expertise with AI managing routine legal operations
- Redesigned function: a fully rebuilt operating model with AI at the core
Lizzy Duffy, Senior Director at the Thomson Reuters Institute, was direct: the choices legal leaders make in the next eighteen months will determine which archetype they become and whether they control that transition or have it happen to them.
So what does it actually look like to take control? The Summit brought together the practitioners who are already doing it, and their answers clustered around three consistent disciplines.
Three frameworks that are working at scale
For the organisations at the Summit that have moved beyond pilot mode, three disciplines keep appearing. These are not theoretical; they come from practitioners at SMBC, Addleshaw Goddard, and DAC Beachcroft who are running AI at enterprise scale today.
1. Start with high-volume, high-value tasks: Apply AI where it can deliver defensible, auditable outputs: document review, contract analysis, legal research across large bodies of material. These are the tasks where performance gains are measurable and the risk of error is mitigated by the architecture of the tools themselves.
2. Build cross-functional ownership from day one: GenAI as a catalyst for new ways of working only succeeds when legal, IT, and finance share objectives and governance from the outset, not brought in sequentially after the technology decision has been made. The legal teams scaling AI most effectively have a C-suite sponsor and an operational owner in legal.
3. Ask the boundary question deliberately: Firms succeeding in this era have answered one question that most AI strategies avoid: what would you not hand over to a machine, even if you could? Cambridge Professor and Thomson Reuters Principal Legal AI Adviser Felix Steffek framed this as the question that defines a mature AI strategy. The answer forces clarity on where human judgement remains irreplaceable and builds the trust framework that makes everything else possible.
Of course, operational discipline only goes so far. The frameworks above assume the underlying technology is capable of supporting them, and not all legal AI tools are. That is where the choice of platform becomes decisive.
The architecture that makes the difference
The Summit’s centrepiece was the live reveal of CoCounsel Legal, and what Rawia Ashraf, Head of Product at Thomson Reuters, demonstrated was a platform built on a fundamentally different architecture from the AI tools most legal professionals have encountered to date.
Agentic reasoning in legal LLMs is what makes this generation of tooling categorically different. CoCounsel Legal is built on Anthropic’s Claude Agent SDK, designed not to execute a series of disconnected prompts, but to plan, retrieve, draft, and validate the way a senior associate reasons: form a theory, check authority, refine, and produce work product a lawyer can trust and cite.
In practice, that translates into capabilities built for the scale and rigour legal work demands:
- Tabular Analysis: process thousands of documents with full citation traceability; run multiple projects in parallel; make midflight edits without restarting
- Deep Research: moves fluidly between Westlaw and Practical Law, grounded in authoritative, expertly maintained legal content with step-by-step research notes
- Citation Integrity (patent-pending): every source the agent reads is verified against a citation ledger before output is returned; click any citation and read the original
- Web search: real-time access for legal-adjacent tasks outside the primary research corpus
- Claude MCP connection (on roadmap): genuine workflow orchestration across legal platforms, moving beyond task-handling tools entirely
If your organisation is ready to move from strategy to action, explore CoCounsel Legal, the platform purpose-built for how lawyers actually work.
The technology, then, is no longer the obstacle. What the Summit ultimately surfaced is that the real challenge, and the real opportunity, is one of organisational will.
The only question left
The Summit made one thing clear above all else. The legal sector is not in an AI debate any more. It is in an AI execution challenge.
The firms and departments that win will not be those with the largest technology budgets. They will be the ones who:
- Chose tools built on trusted, authoritative legal content
- Established the operational disciplines to measure and demonstrate impact
- Asked hard questions about where human judgement remains irreplaceable, and answered them deliberately
The complete AI legal solution is no longer a future prospect. It is what the leaders in that room in London are deploying today. The question is whether your organisation is building the conditions to do the same.
Explore CoCounsel Legal UK today and see how your teams can work faster, safer, and with greater confidence. ↓
