How I think about — AI Workflows

Architect the solution. Don't just visualise it.

AI is rewriting the cost-to-serve curve faster than the billing model is adapting. The interesting question isn't 'which tool' — it's where in the workflow the savings actually accrue, who captures them, and what the client is willing to pay for once the work takes half the time.

Do

Practice
  • +01

    Pilot on the seams between tasks, not inside them

    The biggest wins are at the handoffs (associate → partner, firm → client). Single-task tooling is where AI looks impressive and changes nothing.

    Source — Thomson Reuters Institute · GenAI in Legal

  • +02

    Decide who captures the dividend before you deploy

    If the firm pockets all the efficiency, the client renegotiates. If the client pockets all of it, the firm under-invests. Name the split at the pricing table.

    Source — Clio · Legal Trends Report 2025

  • +03

    Tell clients what you use

    71% of clients are unsure whether their firm uses GenAI. Opaque adoption erodes trust faster than disclosure does.

    Source — Thomson Reuters · 2025 Future of Professionals Report

Don't

Patterns
  • 01

    Don't price AI-assisted work at the old hourly rate and hope nobody notices

    Clients are running their own benchmarks. The bill is the slowest-moving variable in the relationship right now.

  • 02

    Don't pilot a tool without a measurement plan

    If you can't say what 'better' looks like in dollars or hours, the pilot will be declared a success and changed nothing.

    Source — Altman Weil

  • 03

    Don't outsource judgement to the model

    AI is excellent at drafting and synthesis. The lawyering — risk calibration, client read, negotiation posture — is still the deliverable.

§ References

Where this thinking comes from

  • Thomson Reuters InstituteFuture of Professionals Report 2025
  • ClioLegal Trends Report 2025
  • Altman WeilLaw Firms in Transition

Synthesised from publicly available reports and commentary. All views my own.