AI as a research assistant, not an oracle

Every AI-powered project in this portfolio follows the same division of labour:

| Role | Who | Why | | --- | --- | --- | | Reading 10,000 pages | AI | Tireless, fast, consistent | | Deciding what's true | Human | Context, accountability | | Making it legible | Design | That's where value lands |

The pipeline pattern

My client intelligence work runs as a multi-agent pipeline: a research agent gathers sources, extraction agents pull structured fields, classification agents tag and score, and a QA agent fact-checks before anything reaches the dashboard.

The output isn't "an AI answer" — it's a database with receipts, rendered as an interface a human can interrogate.

Why this matters

Trust in AI outputs comes from inspectability. A chat response is a black box; a filterable dashboard with sources is an argument you can check. Design isn't decoration on top of AI — it's the accountability layer.