Our Take
Command Center turns adoption anxiety into comparative data, but the real play is DeepJudge: firms have spent decades building institutional knowledge that sat locked away, and now that knowledge compounds directly into every Harvey workflow.
Why it matters
Law firms are spending heavily on AI but have little visibility into whether they're trailing the market or leading it, and even less certainty that the AI they deploy actually reflects how they work. These tools address both gaps at once.
Do this week
Innovation heads: audit your current feature adoption across practice areas this week so you can flag underutilized workflows before Command Center peer data exposes them to leadership.
Two Moves Close Harvey's Knowledge and Visibility Gaps
Harvey launched Command Center, a peer benchmarking and analytics platform that gives law firms two views into their AI usage. The external view compares your firm's adoption metrics against anonymized data from more than 1,500 Harvey deployments globally, broken down by practice area and specific feature. The internal view drills into firm-wide detail: which practice groups lag, which partners or associates engage most, which workflows drive adoption.
The platform includes an agentic layer that accepts natural language queries over that usage data. Ask which practice groups differ in adoption rates, or what behaviors distinguish highly engaged users, and the system surfaces answers directly. Command Center also surfaces "Intelligent Recommendations" that flag which features peer organizations have already enabled, meant to guide rollout strategy.
Harvey also announced a partnership with DeepJudge, a Swiss legal AI startup (founded by ex-Google engineers) that specializes in document classification and knowledge extraction. The integration pipes a firm's past work, decisions, and institutional expertise into Harvey's workflows while respecting access permissions and ethical walls. The result: AI agents that don't just generate outputs, but align them with "what good looks like" within that specific firm's practice, grounded in decades of precedent and firm judgment.
Harvey worked with Foley, Haynes Boone, Clayton Utz, Rajah & Tann, and Dentsu to develop Command Center. Sean Monahan, Foley Director of Practice Innovation, said the tool moves the firm "from anecdotal assessments of AI usage to data-driven management of deployment, training, governance, and value creation."
Adoption Theater Meets Institutional Memory
Law firms adopted generative AI tools rapidly but often blindly. Most firms lack clear signals on whether their deployment is working, whether adoption is shallow, and whether the AI they're using actually reflects their practice standards. Command Center supplies the first signal. Peer benchmarking matters in legal services because no firm wants to be materially behind the market, and most lack the internal data to know where they stand.
The DeepJudge integration solves a different but equally acute problem: firms have built vast repositories of past work, memos, negotiation positions, and client-specific standards over decades. That knowledge is a competitive asset, but it's fragmented across document management systems and effectively invisible to AI agents. An AI model trained on public legal text has no idea how your firm actually structures arguments, what your M&A team considers a deal-breaker, or why your employment group handles severance the way it does. DeepJudge makes that institutional knowledge accessible to Harvey workflows, so outputs are grounded in firm practice rather than generic legal convention.
Both tools address a real cost: the friction of jumping between disconnected systems (Command Center aggregates analytics; DeepJudge pulls knowledge from your DMS into Harvey) and the risk that AI outputs diverge from firm standards (DeepJudge anchors the model in your institutional judgment).
Where to Start
For innovation and operations leaders: treat Command Center as a governance tool, not a vanity metric. Identify the practice areas or cohorts with lowest adoption and diagnose why. Is it resistance? Unclear use case? Training gap? Low adoption is a deployment failure, not a technology failure, and the data will tell you which.
For knowledge and library teams: begin the work of surfacing your firm's institutional knowledge in a form DeepJudge can ingest. This means auditing which prior work, templates, negotiation playbooks, and decision frameworks are worth grounding AI outputs in. Not everything is valuable; discretion here compounds over time.