writing · ai-governance
AI governance
Approval gates, audit trails and human review on AI output. What has to sit between a model and anything a client will read.
2026-09-08
ai-assisted, human-edited
Parallel AI review finds too much, and the volume is the failure mode
Fourteen agent lanes over one codebase returned 63 findings in an afternoon. The count is not the result. What made the list usable was the pass that tried to kill every finding, and the two calls where the obvious action turned out to be wrong.
2026-08-26
ai-assisted, human-edited
Two Agencies, Same Day Rate: One Has an Audit Trail and One Does Not
When a client disputes an invoice or quietly declines to renew, it is rarely about the work itself. It is about what they can and cannot reconstruct after the fact.
2026-07-30
ai-assisted, human-edited
The bottleneck is approval, not generation
This site runs an AI agent that drafts blog posts twice a week. Before this week, the blog had not gained a post in 41 days while four agent drafts sat unpromoted. That queue is the design, not the failure.
2026-06-19
ai-assisted, human-edited
Why I Added a Manual Checkpoint to My Automated Delivery Pipeline
Automation handles the repetitive steps well. But the moment a deliverable touches a client without a human reading it first, you've outsourced your quality standard to the tool that built it.
2026-06-15
ai-assisted, human-edited
The Workflow That Reviews and Delivers an AI Image Is Worth More Than the Image
Midjourney and Firefly now hand you commercial-licensed images for under $50 a month. The generation is free. The approval trail, versioning, and delivery governance still cost real money, and that is where the margin lives.
2026-05-30
ai-assisted, human-edited
Fast AI builds break in front of clients. Here is what to build instead.
Every AI-ops shop is selling speed. Live in two weeks. The problem is that the thing that breaks an AI build is never the part you can ship in two weeks. It is the exception nobody designed for.