Separate adoption from useful work
When law firm AI adoption stalls, first check whether the system makes a specific legal workflow easier to complete at the required standard. Logins and training attendance won’t tell you that on their own.
Possible causes include weak task fit, unreliable output, too much review, unclear access rules, and a tool that sits outside the way the practice works. Treat these as hypotheses to investigate, rather than assuming the team resists change.
Start with one practice and one recurring task. Ask the people doing it to show a recent example, including where they stopped using the system or had to redo its work.
Measure use alongside review effort
Define who is eligible to use the tool, what counts as use, and the period you’re measuring. Keep those definitions consistent when comparing results. Record whether use is required or voluntary.
Then measure what happens after the output appears. How much work does the reviewer redo? Does the system make the answer easier to check? Does it help the matter move forward, or create another step?
An adoption number becomes more useful when you can connect it to a specific workflow and quality measure. A firm-wide percentage can hide a strong result in one practice and poor fit in another.
What changed in an AM100 corporate practice
In one AM100 law firm engagement, use rose from 32% under a mandate to 72% after that mandate was removed. The result applies to the corporate practice, not the entire firm.
The earlier baseline came from vendor analytics. The later measurement came from our telemetry during the quarter after launch. That change in measurement source belongs in the interpretation.
The work focused on making the system useful inside the practice’s actual workflow. The point wasn’t to force more activity. It was to give lawyers a reason to return.
Run a pilot that answers the right question
- Pick a workflow. Choose recurring work with a clear owner and enough examples to evaluate.
- Agree on review criteria. Define what an acceptable result contains, how someone checks it, and which failures matter.
- Confirm permitted data and access. Work within the firm’s actual policies and the engagement’s scope.
- Observe completion. Track use, correction effort, and the step that controls throughput.
- Decide what to change. Improve the workflow, change the tool, expand the pilot, or stop based on the result.
Make feedback easy to provide while the task is still fresh. Ask for the example that failed, not just a satisfaction score.
Build around the limiting step
If the work still waits for a small group of senior reviewers, more generation may not create more capacity. The next useful change could involve review support, routing, clearer standards, or a different division of work.
If the tool fits and the process improves, give someone responsibility for keeping it useful. That includes feedback, access changes, maintenance, and measurement after the project team leaves.
Use a constraint-first project brief to connect the proposed change to the result the practice wants. That creates a better basis for investment than asking everyone to use AI more often.
Find your next useful step.
Share your goal. We’ll investigate the likely bottleneck and recommend what to do next.
Start your free assessment