Build only after the problem gets sharper.

AI work moves too quickly for heavy theory and too dangerously for pure enthusiasm. I use design work to make the next decision visible.

PROCESS

01
Understand

Explore the existing material, workflow, people, constraints, data reality and strategic goal before treating AI as the answer. I look for role-specific needs, fears, non-goals and the language people already use to describe the work.
02
Frame

Turn the fieldwork into sharper use cases, risk questions and testable hypotheses. The frame should make dependencies visible: definitions, data access, validation, ownership and the point where a tool becomes useful enough to change a decision.
03
Prototype

Build the smallest working version of the workflow: a module, prompt flow, structured knowledge file, blueprint or interface people can test against real work. The goal is to expose trust, missing data, unclear definitions and whether the output actually helps someone move.
04
Adopt

Shape the examples, guardrails, metrics and working routines around the solution as it moves from pilot to real use. Adoption means pilots, feedback loops, validation routines and clear limits on what the system should and should not do.

The model is rarely the first question. The work, decision, person and consequence usually are.

A working draft reveals what people trust, misunderstand, avoid and need to decide before scaling.

People do not adopt AI because a tool exists. They adopt a clearer way to get a real job done.