Articles
Operator notes.
Problem-first writing on AI, automation, and the process underneath. Each piece names a real bottleneck, exposes the broken layer nobody owns, and shows what the working system actually looks like.
Law firm AI savings: clients want a number the timesheet can't produce
Banks now ask law firms to quantify their AI savings. Most can't: the timesheet records hours, not steps. The fix is measuring the work, not a discount.
7 min read → AI governanceAI incident notification: OpenAI's Medicare breach landed in an inbox checked once a day
AI incident notification has two ends. OpenAI's Medicare breach notice took 84 days to send, then reached a researchers' inbox checked once a day.
7 min read → Enterprise AIComputer-use agents vs API integration: Amazon's block on Meta's Muse is the contract you never signed
Computer-use agents vs API integration: Amazon cut off Meta's Muse agent with one pop-up. A browser agent on someone else's system is an integration nobody signed.
8 min read → Enterprise AIWorkflow redesign before AI agents: Microsoft's own playbook credits the redesign, not the agents
Workflow redesign before AI agents: Microsoft licensed Copilot to 200,000 staff and usage plateaued. Its own playbook says the redesign did the work.
9 min read → Enterprise AIThe AI agent harness sets the bill. Nobody in your org chose it.
An AI agent harness is everything in an agent except the model. It moved cost 2× to 5× with no change in task success, and it's the layer you actually own.
13 min read → ConstructionConstruction RFI automation: AI made RFIs cheap to write, not cheap to answer
Construction RFI automation now drafts an RFI in a click. The cost, the response time and the claims exposure sit on the answer side and in the log — the part nobody automated.
8 min read →