Capability area
AI problems are rarely just AI problems.
The quality of an intelligent system depends on more than the model. I look at the surrounding ecosystem—users, workflows, information architecture, incentives, data, interfaces, organizational constraints, and human decision points—to understand what actually determines whether an AI experience succeeds.
How I work through an AI question
- 01Understand the systemUsers, workflows, data, ownership, incentives, and the decisions people are accountable for.
- 02Locate the opportunityWhere intelligence changes an outcome — and where it would only add a surface to supervise.
- 03Define what good meansOutcome quality, effort to verify, and recoverability, with test sets behind each.
- 04Design the human loopWhere judgement sits, what evidence it needs, and how the system fails safely.
- 05Translate into strategyPriorities, sequencing, experience principles, and the decisions a roadmap can act on.
Capabilities
AI opportunity identification
Finding where intelligence meaningfully improves an outcome, and where it only adds surface.
Human-AI interaction research
How people supervise, verify, correct, and abandon intelligent systems.
AI experience strategy
Principles for how a product should behave when it is uncertain.
Evaluation and test-set development
Scenario sets and measures that make quality arguable with evidence.
Agent and workflow evaluation
Assessing multi-step systems against the work they are supposed to carry.
Human-in-the-loop design
Placing judgement where it changes the outcome, not where it merely signs off.
Guardrails and responsible AI
Failure modes, recoverability, disclosure, and accountability by design.
Research-driven AI decisions
Evidence that resolves roadmap questions rather than decorating them.
Complex workflow diagnosis
Tracing breakdowns across tools, roles, data, and handoffs nobody owns.
Adoption and organizational readiness
Whether the surrounding organization can absorb what the system assumes.