I turn what a business knows into context that helps people and AI do better work.

Every business has customer knowledge, standards, rules and hard-won exceptions that rarely exist in one usable place. I uncover and structure that knowledge so the people and AI producing the work can make better decisions.

See the work Start a conversation
See it, don't just read about it

Raw: how most people explain this

So it's kind of like UX, but for AI, I guess. A business has decisions, exceptions and know-how that never get written down; it just lives in people's heads. If you just point an AI at “be helpful,” it doesn't know any of that, so it either invents something or gives a generic answer that could apply to any company. This matters more every year as more of the actual work gets handed to AI directly. Somebody has to pull that knowledge out and shape it into something a model can actually use. Technically it's nobody's job yet.

The same principle applies whether the work is being produced by a person or a model: better decisions depend on having the right context available at the right moment.

Most UX deliverables were built to be read by a person: wireframes, documentation, style guides. AI doesn't read the way a person does; it needs curated context, not a document written for human pace. That's the practice this site demonstrates directly: the same task, shown at three different levels of curation, so the gap is visible instead of asserted.

Proof, not claims

See the pattern in isolation first, then across three unrelated domains: same underlying discipline, same pattern of failure and fix, every time.

Same brief, three levels of context

One brief, one set of raw materials, three levels of context. The thesis demo.

Customer support

Same raw materials, three levels of curation, three different replies.

Financial anomaly investigation

Same investigation notes, three different audit briefs.

Legal contract clause review

Same clause comparison, three different reviews.
Also worth a look
Signal & Flow A live multi-agent platform, built on the same discipline Services Five applications of one discipline, each linked to relevant work, a demonstration or a practical tool Recent writing Thinking on context design, AI and restructuring
Recent thinking
Context Debt The parallel to technical debt AI is making impossible to ignore Why prompt engineering was the wrong abstraction Wording was never the lever. What the model has to reason from was. Why AI-native companies need more people, not fewer A structural argument against the headcount-cutting story

If your business is starting to lean on AI for real work (support, analysis, contracts, anything) and the results are inconsistent, that's usually a context problem, not a model problem. I'd be glad to talk through what curating it well would look like for you.

Start a conversation