Case study

Law firm ai training

Most teams do not fail at AI because the tools are weak. They fail because nobody changed what gets asked for on a Monday morning,…

On this page The three-week cliff What people are actually taught Measuring one workflow first What good looks like

Most teams do not fail at AI because the tools are weak. They fail because nobody changed what gets asked for on a Monday morning, and the new way quietly became optional.

The three-week cliff

Adoption drops off at about three weeks, once the novelty is gone and the deadline is real. Work with us on AI literacy training and on AI strategy, and you will see the same pattern in the data before you see it in the room. There is a worked version of this in our case study with a fee-earning team.

Nobody went back to the old way, because the new way was already how the work got measured.

The teams that got past three weeks had one thing in common. Somebody senior changed the question they asked in the weekly review.

A team reviewing printed workflow maps on a table
Caption, with a credit if the image is not ours.

What people are actually taught

Most curricula teach the tool. The teams that stick with it taught the decision first, then showed which tool shortens it.

  • Start from one workflow somebody already complains about.
  • Measure it for two weeks before anyone is trained.
  • Train the people who own that workflow, not the whole company.

Measuring one workflow first

A baseline takes two weeks and settles most of the arguments that come later. Without one, every claim about time saved is a matter of opinion.

Callout box. A baseline is not a business case. It is the thing that makes a business case arguable.

A worked example

One legal team measured drafting time across eleven matters before any training happened, which is exactly why the change was defensible afterwards.

Week Hours on drafting Change
Baseline 19.5 —
Week 4 14.0 -28%
Week 12 8.5 -56%

What good looks like

Adoption is a management problem, not a curriculum problem. The teams that kept going had a manager who changed what they asked for, and kept asking for it after the interesting part was over.

Web PM

Role, two lines of bio.