The draft arrived faster. The language reads well. The person using the tool completed a task that previously took much longer.
Those are useful improvements. They also leave an important question unanswered: does the person understand the work well enough to recognize when the AI has taken it in the wrong direction?
For law firms, that question belongs in the adoption plan from the beginning. It affects how people are trained, how their work is reviewed, and what happens when they need help.
Better output and stronger judgment are different results
In an October 7 research account, David Autor and Tanya Rodchenko described a three-month randomized experiment involving 133 patent lawyers across 11 intellectual property law firms.
Access to an AI drafting assistant improved drafting performance. The more revealing result came when researchers asked participants to review and correct a flawed patent without AI.
Senior lawyers who had received AI access performed better than senior lawyers in the control group. Junior lawyers showed no average improvement on that independent task. The study classified senior lawyers as having at least seven years of experience.
This was a limited experiment involving a particular profession and tool, with firms that had business relationships with Google. It does not establish what happens across every legal practice. It does, however, expose a distinction firms need to take seriously: a person can produce better work with assistance without demonstrating stronger independent judgment. [1]
My reading is that adoption needs to account for both outcomes. Firms want useful work today. They also need people whose understanding grows as they do it.
The training task changes when the answer can look convincing
Traditional software training often has a visible result. A document is formatted correctly. A search returns the intended file. A workflow reaches the right destination.
AI introduces another layer. Someone can follow the demonstrated steps and receive an answer that looks complete while still missing information that matters.
Consider a fictional training exercise in which an approved AI tool summarizes a set of documents. The summary is readable and broadly accurate, but it omits a qualification that changes how one passage should be understood.
Completing the prompt is only part of the task. The learner also needs to compare the result with the source, identify the omission, explain its significance, and recognize when the issue requires review by the responsible lawyer.
That gives trainers something more useful to assess than whether the person reproduced the demonstration.
A practical performance check should clarify reasoning: what did you verify, what did you modify, and why?
Equip those who handle questions after the demonstration
Training and support teams bridge the gap between AI rollouts and everyday work.
They address questions about access, confusing results, missing sources, unfamiliar behavior, and user permissions. While these questions may seem similar, they require distinct responses.
For instance, if a user reports that an AI summary missed an important document, the cause could be an access restriction, incorrect source selection, unclear instructions, or tool limitations. It may also indicate a need for a different review process.
Treating every issue as a prompting problem overlooks these distinctions.
I recommend equipping support teams with a shared framework to investigate tasks, identify resolvable issues, and route questions requiring further decisions. Help desk professionals should have a clear path to the appropriate owner for questions about source access, firm policy, or substantive legal review.
This coordination across training, IT, knowledge management, practice leadership, and security and AI governance is essential. Each group possesses information vital to the others.
Training must reflect actual work
Wider legal-sector research underscores the need for practical preparation. In its May discussion of the 2026 Future Ready Lawyer Survey, Wolters Kluwer reported that 39% of respondents cited inadequate training and resources as barriers to AI adoption. Their expert discussion emphasized training grounded in specific roles and real workflows.
A general introduction establishes shared vocabulary but cannot resolve every task-level decision.
Trainers designing instruction, knowledge professionals assessing sources, and support analysts investigating reported failures require different practice opportunities. Their responsibilities intersect, but their decisions differ.
For firms enhancing an existing rollout, I recommend starting with a defined task and agreeing on four key elements:
- What successful performance entails.
- Applicable sources and information boundaries.
- What users must verify before relying on results.
- Where unresolved questions should be directed.
Develop exercises and support guidance around these decisions. Then observe where users struggle and refine both.
This approach also creates a valuable role for experienced staff: to explain exceptions, assumptions, and warning signs that demonstrations may overlook. Make this knowledge accessible in training examples and support resources.
Measure post-training capabilities
Attendance and usage provide insights into participation but leave many capability questions unanswered.
For the selected task, firms should assess whether users can identify material omissions, select appropriate sources, explain corrections, and escalate uncertainties to the right individuals. They should also evaluate how much rework remains after AI assistance and which issues frequently recur.
These proposed implementation measures focus on the quality of work and the capabilities of the individuals involved.
For professionals, this establishes a concrete development goal: be able to articulate how you reached a result and understand its limitations. For firms, it provides a solid foundation for determining whether a workflow is ready for expansion.
The research encourages law firms to look beyond the finished document.
A successful AI rollout delivers better work and equips individuals to make stronger judgments.
Put this into practice
My live, four-hour workshop, AI Adoption for Law Firms: Equip Your Training & Support Teams, focuses on the people responsible for translating AI capabilities into everyday training and support.
It brings together professionals across training, IT support, knowledge management, and implementation to examine workflows, strengthen practical instruction, and clarify support and escalation.