Insights
Notes on making AI work inside real organizations.
Writing published here will be drawn from engagements and from the product work — what actually happened, including the parts that did not go to plan.
What this will be
Working notes, not thought leadership.
There is no shortage of confident writing about AI. Most of it is written by people who have not had to operate the thing afterwards.
What goes here will be narrower and more useful: how a specific assessment was structured, why a particular automation was not worth building, what a data model looked like before and after, where a governance model held up and where it did not.
Nothing is published here yet. Rather than fill the page with placeholder articles, it stays empty until there is something worth your time.
Likely subjects
The questions that come up most often.
These are drawn from the conversations that recur across engagements, not from a content calendar.
Deciding what is worth automating
Most automation candidates fail on exception handling rather than on the happy path. How to tell the difference before building.
Why reporting disagrees with itself
Nearly every reporting dispute is a definition dispute. Fixing the definitions before building the dashboard.
Governing AI-assisted engineering
Traceability, independent review and approval controls — what it takes to make agent output auditable.
Identity and continuity in AI media
Why generated media drifts between shots, and the anchor-and-review approach that holds it together.
Adoption is the hard part
Systems that work and are not used. What changes when adoption is treated as part of the build rather than as an afterthought.
When not to use AI
The cases where a documented process, a database constraint or a conversation solves the problem better.