Weekly active associates using Harvey at a large US law firm fell 30% between the first and fourth month after rollout. Why?

Introduction

A large US law firm rolled out Harvey to all 1,200 of its lawyers. Usage was strong in the first month. But between the first and fourth months, weekly active associates fell by 30%, from 780 to about 545, while partner usage held steady. The firm's innovation director has asked Harvey's customer success team to explain the drop before the firm's budget review, where renewal will be discussed.

Post-launch decay is common in enterprise software, but a 30% drop among associates (the heaviest potential users) is serious. I'll clarify the situation, rule out external factors, map the journey, break down the metric, form and test hypotheses, and propose a resolution plan.

Step 1

Clarifying Questions (3 minutes)

• How is "weekly active" defined, and did the definition or tracking change?

Why it matters: A measurement change can look like a usage drop. Expected answer: Any query, workflow run or Vault action in a week; no tracking changes. Impact on approach: The drop is likely real.

• Is the decline spread across practice groups and offices?

Why it matters: Narrows the search. Expected answer: The largest drops are in litigation (down 45%) and real estate (down 38%); corporate M&A is down only 10%. Impact on approach: Practice-specific causes are likely.

• What happened in the firm during months two to four?

Why it matters: Firm events affect behaviour. Expected answer: Launch training ended after month one; a firm memo in month two reminded lawyers about AI use policy and citation verification after a widely reported court sanction at another firm. Impact on approach: Training and policy effects are suspects.

• Did Harvey ship product changes in this period?

Why it matters: Internal changes are common causes. Expected answer: A new home screen layout and a change to how Workflows are organised. Impact on approach: Discovery changes may matter.