The paradox
A new Chambers report, summarized by Law.com, found that associates at firms deeply embedded in AI advisory work were more likely to report lower confidence using AI tools themselves. The firms most fluent in client-facing AI strategy may be among the firms least successful at turning that fluency into daily practice.
That is not hypocrisy. It is a management failure. Advising a client on AI governance is a partner-led, memo-shaped exercise. Teaching hundreds of associates how to use AI safely inside matters is an operating-system change.
Training cannot be a webinar
Most law firm AI programs still look like technology rollouts: tool access, vendor demo, policy memo, optional lunch session. That model creates awareness but not confidence. Associates learn by doing, and legal AI has to be taught inside real workflows: first draft review, privilege screening, deposition prep, due diligence, chronology building and client-update drafting.
Confidence also depends on permission. Junior lawyers need to know when they are expected to use AI, when they are forbidden to use it, and what level of verification is enough. Ambiguous policy produces the worst behavior: either reckless use or no use at all.
The client angle
The disconnect becomes dangerous because clients are increasingly asking firms to explain how AI affects price, staffing and quality control. A firm cannot credibly sell AI-enabled efficiency if its own associates experience the tools as risky, slow or unofficial.
The next competitive advantage in BigLaw will not be who announces the largest AI partnership. It will be who turns AI into a normal supervised practice skill, taught with the same seriousness as privilege, Bluebook citation and client communication.
The bottom line
Firms should measure AI adoption like training quality, not software activation. The question is not how many licenses were provisioned. It is whether a first-year can use the tool inside a matter, explain the output, spot the failure mode and know when to escalate.