Andrew Lewis was Here
Subscribe
Sign in
Home
Resources
Archive
About
Latest
Top
We Were Already Offloading Our Thinking
A response to Yennie Jun: the autonomy we’re afraid of losing to AI was mostly delegated already — what matters is whether anyone reviews what comes…
Jul 20
•
Andrew Lewis
2
Agent Governance Is Access Control You Never Wrote Down
Every requirement a UN body just named for a trustworthy AI agent is the access discipline you already run for people — the only new part is the…
Jul 16
•
Andrew Lewis
What a Government Proved About AI That Vendors Wouldn’t
The Government of Alberta published twenty-one papers on rebuilding its technology with AI, and the real lesson is about the part you can’t buy.
Jul 14
•
Andrew Lewis
2
The Clock You Can’t See From Across the Table
Zack Shapiro’s “The Two Clocks” is right about the biggest problem in AI and law. Here is what it looks like from inside a firm.
Jul 8
•
Andrew Lewis
1
Governance Isn’t a Step in Your AI Pipeline
The Governed Context Pipeline — three phases, one wrapper, and the verification economics that decide what’s worth automating at all.
Jul 6
•
Andrew Lewis
1
The Teams Asking for AI Aren’t Generating the Revenue
When AI effort flows to whoever is most excited, you can ship for a year and never touch the part of the business that pays for it.
Jul 2
•
Andrew Lewis
1
June 2026
Least Privilege Was Built for People Who Pause
Every access model you run assumes a human is standing at the moment of action. Agents void that assumption.
Jun 29
•
Andrew Lewis
Defining “Done” Is the Work the Loop Can’t Do
An agent that runs until it’s finished forces you to say what finished means before the work exists — and code gets away with it in a way a legal…
Jun 25
•
Andrew Lewis
Your Agent Has the Same Permissions You Do
The conversation about AI agents is all about what they can do. The question nobody can answer yet is what they did.
Jun 23
•
Andrew Lewis
The Entry Fee for McKinsey's 60 Percent Smaller Teams
The agentic delivery numbers are real. The preconditions underneath them are the part nobody will quote.
Jun 18
•
Andrew Lewis
You Cannot Review Your Way Out of This
Verification at AI scale is a selection problem, and every guardrail, eval, and playbook only works after a human decides what deserves attention.
Jun 16
•
Andrew Lewis
May 2026
The Last Mile in Legal Has Its Own Geography
McKinsey published the data. HBR named the gap. Both are right. Both are general. Here is what the terrain looks like inside a law firm.
May 14
•
Andrew Lewis
This site requires JavaScript to run correctly. Please
turn on JavaScript
or unblock scripts
