Inside the Conversations Shaping Legal AI’s Next Chapter

Earlier this month, Harvey convened a small group of AmLaw 100 leaders for an intimate dinner at NARO in New York. The event was co-hosted by Oz Benamram, a longtime advisor, operator, and one of the most thoughtful voices in the legal AI space.

Our CEO, Winston Weinberg, joined the conversation alongside firm leaders spanning innovation, knowledge, operations, and practice leadership. The format was intentionally simple: no slides or stage, just candid discussion among people actively shaping how AI is showing up inside their firms.

What stood out wasn’t any single viewpoint, but the strength of the community in the room. These were leaders comparing notes in real time: what’s working, what’s stalling, and where the next set of risks and opportunities are emerging. Judging by the reaction on LinkedIn afterward, there’s growing recognition that these are exactly the conversations firms don’t want to miss.

In my role, I regularly speak with both general counsel and law firm leaders, which made it especially valuable to hear the law firm perspective in isolation. From the GC side, there’s a growing expectation that firms will move beyond simply adopting AI and begin intentionally redesigning how work is structured, staffed, and delivered.

Hearing how firms are grappling with these same pressures internally, without the client lens in the room, surfaced the real constraints and tradeoffs that don’t always appear in client conversations. It also underscored just how much execution, not experimentation, will define what comes next.

The themes below reflect what stood out most to me after reading Oz’s post-event reflections of the conversation that unfolded around the table. This is not a transcript or a point-by-point recap, but a distillation of the ideas that best capture how firm leaders are thinking as legal AI moves from experimentation to an operating model.

The Themes That Stood Out

Talent development is the compounding risk firms can no longer ignore

One of the most animated parts of the discussion centered on talent — not hiring volume, but how lawyers actually develop judgment in an AI-enabled environment.

As Oz reflected, firms are confronting uncomfortable but unavoidable questions. If AI absorbs much of the repetitive work that historically trained junior lawyers, what replaces it? How do firms intentionally design learning, apprenticeship, and evaluation models for a world where “time spent” is no longer the proxy for expertise gained?

This theme also aligns with what we’re seeing beyond firms. Through our work with law schools, including the Harvey Law School Program, we’re seeing growing demand to equip future lawyers with practical generative AI skills earlier in their training — underscoring that talent development will need to be rethought across the entire legal pipeline.

Clients are already planning for change, whether firms are ready or not

Another clear takeaway from the room was how pragmatic in-house legal teams have become about AI’s implications.

As Oz observed, many clients are already operating on the assumption that firm economics, staffing models, billing structures, and collaboration norms will change. From their perspective, the debate isn’t if these shifts will happen, but how.

What stood out was that clients often appear more willing than firms to acknowledge this reality. Firms waiting for full internal consensus before acting may find themselves increasingly out of step with the expectations clients are already forming — and, in some cases, acting on.

AI is quietly reshaping who has influence inside firms

Beyond efficiency, Oz noted that AI is acting as an unexpected organizational catalyst.

Across firms, AI is pulling previously disengaged partners into innovation conversations, enabling earlier and more substantive client engagement, and bringing knowledge and innovation teams into client-facing discussions.

Rather than pushing adoption from the inside, these teams are increasingly being pulled in by client demand. This “reverse inquiry” is subtly reshaping where influence sits inside firms and elevating the importance of roles that bridge technology, practice, and client service.

Change management is the real constraint

If there was one theme that cut across every conversation, it was capacity for change.

As Oz synthesized, even firms with strong intent and capable tools are struggling to align stakeholders, redesign workflows, adjust incentives, and move from pilots to sustained adoption. The bottleneck isn’t access to technology — it’s organizational readiness.

This gap is becoming more visible as firms move from experimentation to execution, and it’s increasingly the difference between firms that make progress and those that stall.

ROI still matters, but the definition is expanding

ROI remains central to how firms evaluate AI investments, but Oz noted that the conversation is clearly evolving.

Time savings still matter, but firms are increasingly focused on value defined by quality and consistency, risk reduction, client satisfaction, lawyer engagement and retention, and competitive differentiation. Many acknowledged they lack clear frameworks to surface, measure, and communicate this value internally and externally.

This mirrors what we’re seeing in Harvey’s own research with RSGI: as AI matures, firms are broadening how they define return and moving beyond efficiency alone toward longer-term strategic impact.

Looking Ahead

The tone of the evening was cautiously optimistic.

As Oz reflected afterward, the firms best positioned for what comes next are not those chasing tools, but those treating AI as an operating capability — investing in people, processes, and foundations alongside technology.

The takeaway from the room was clear: the next phase of legal AI will be shaped less by experimentation and more by execution. And firms that engage early with clients, with their own teams, and with a strong peer community, will have a meaningful advantage.