Where Legal AI is Becoming Real Work

Last week in London, we brought together a group of senior law firm innovation leaders for a breakfast roundtable in partnership with Briefing. The group reflected a broader effort to convene a community of AI-focused leaders across UK and international firms — those responsible for shaping how technology is deployed inside organizations of meaningful scale.

What stood out was the level of experience in the room. These were leaders actively working through implementation, comparing notes on what’s changing inside their firms, where progress is taking hold, and where challenges persist.

Ahead of the session, Briefing’s latest research pointed to two themes: the growing importance of workflows and the increasing focus on collaboration. Most firms now have some form of genAI “playbooks” in place, but for nearly two-thirds these apply to just 1–10% of potential use cases — a clear sign that many are still early in scaling. Collaboration is also moving up the agenda, with nearly half of firms actively working toward real-time, platform-based collaboration with clients, even if only a small number have fully enabled it today.

Both themes carried through into the discussion, shaped by the firsthand experience of firms already working through implementation.

From Access to Application

Across the firms represented, lawyers already have AI access available to them. The focus has shifted toward how those tools show up in day-to-day work — whether they are used consistently and whether they meaningfully change how tasks get done.

Several participants described an earlier phase of broad experimentation. What’s taking shape now is more deliberate: identifying where AI can be applied within specific parts of the workflow and where it can deliver repeatable value.

The conversation is becoming more operational, with attention moving toward how AI fits into the lifecycle of legal work rather than sitting alongside it.

The Workflow Gap

Even in firms with strong adoption, there is still distance between individual usage and fully integrated ways of working. Lawyers move across drafting, review, advice, and coordination (often within the same matter) and when AI sits outside that flow, it introduces friction.

What we heard from Harvey customers in particular was a focus on closing that gap. The priority is less about adding capability and more about embedding it — ensuring AI is present where work is already happening, with enough context to be useful across tasks, not just within a single prompt.

That early-stage scaling came through in the data as well, with most firms still applying structured playbooks to a relatively small share of their overall use cases.

Collaboration is Taking Shape

In earlier stages, AI usage often happened individually. Lawyers generated outputs that were useful in the moment but difficult to share, adapt, or build on. That model creates limits, especially inside larger teams working across the same matter.

A more connected way of working is emerging, where AI-assisted outputs are treated as part of a shared process — refined across a team, carried forward, and improved over time.

Several participants pointed to the practical implications:

• Reducing duplication across teams
• Improving consistency in how work is delivered
• Enabling knowledge to move more easily within and across matters

There was also a clear recognition that collaboration extends beyond the firm. As in-house legal teams adopt AI in parallel, expectations around how work is shared and developed together are evolving. The boundary between internal and external workflows is becoming more fluid.

The research reflects that momentum. While only a small number of firms have fully enabled real-time, platform-based collaboration with clients, nearly half are actively working toward it, and many are already experimenting with more direct collaboration using existing tools. Where this is happening, the benefits are starting to show — from faster identification of risks in processes like due diligence to more coordinated document review and shared tracking of matter progress.

Alongside this, firms are working through a more fundamental question around client value — how services are packaged, delivered, and experienced. Several participants spoke to the complexity of evolving delivery models in parallel with AI adoption: balancing efficiency gains with client expectations, rethinking how work is scoped, and ensuring that improvements in speed or consistency translate into something clients can clearly see and value.

The Ongoing Work of Adoption

The operational work of adoption continues as firms navigate training, governance, and internal alignment. The conversation is now more grounded in execution, with leaders focused on supporting sustained usage, guiding lawyers toward the right use cases, and measuring outcomes in ways that reflect the realities of the business. Several Harvey customers emphasized starting with areas where the value is clear and repeatable. From there, usage tends to build with more consistency and less friction.

That shift is also beginning to surface in commercial discussions. Around a third of firms report that genAI is already putting pressure on margins, prompting more active consideration of pricing and delivery models. In many cases, how firms collaborate internally on evolving those models has yet to fully catch up with the pace of technological change.

A More Defined Direction

There is still variation in how far firms have progressed, but the direction is increasingly shared. The focus is moving toward:

• Embedding AI within real workflows
• Supporting team-based ways of working
• Connecting outputs across the lifecycle of a matter
• Pointing adoption toward the highest-value and most profitable opportunities

Several participants noted that as adoption matures, there is increasing focus on where AI drives the greatest commercial impact — not just where it saves time, but where it meaningfully improves how services are delivered and valued. Bringing together this group of leaders made that visible. The discussion reflected a community actively shaping how legal AI is applied in practice — as part of how work gets done.

We’re grateful to Briefing for partnering with us, and to everyone who contributed to the conversation.