← All projects

Case study

The PM Intelligence Tool

An agent system every PM at the company now uses daily — and the discipline of deciding what not to automate.

Role
0→1 owner — pitched it, scoped it, built it
Timeline
Self-initiated, from the data side
Team
Solo build
Outcomes

Nobody asked for this

This tool wasn't on a roadmap. From my seat on the data side, I watched customer intelligence scatter across four systems — Gong calls, Slack threads, Notion docs, Intercom tickets — with no owner and no way to see it whole. Before writing a PRD, a PM reconstructed reality by hand, one search box at a time.

I pitched the build myself. Seeing the gap from the data side and turning it into a product is the founding-hire version of customer discovery.

The real cost

The hours were the visible cost. The invisible one was worse: PRDs were only as good as whichever fragments a PM happened to find. Decisions inherited the blind spots of a manual search across four tools.

The design decision that mattered

Automate end to end — the agent writes the PRD

Passed

The tempting demo, and the wrong product. A PRD is a decision document; automating the decision removes the accountability that makes PMs trust it — and tools that replace judgment get quietly resisted.

Agents draft, PMs decide

Chosen

Agents own the grunt work — signal detection across the four sources, analysis, a first PRD draft. The PM owns the judgment: what matters, what ships. The boundary is the product.

Why the boundary is the product

The hardest call wasn't technical — it was deciding what not to automate. Tools that eliminate grunt work get adopted; tools that replace judgment get resisted. Drawing the line at “draft, don't decide” is why adoption hit 100% instead of stalling at the demo.

What I built

Three agent roles working over Gong, Slack, Notion, and Intercom: signal detection surfaces what's changed and what customers keep hitting; analysis turns raw signal into patterns worth a PM's attention; PRD generation produces a first draft grounded in that evidence. The PM enters at the judgment step with the reconstruction already done.

Results

100%
of PMs use it — daily, not at launch
~60%
less customer-research time, as reported by the PMs using it
min
to a first PRD draft, down from hours

Adoption is observed; the time savings are self-reported by the PMs — I'd rather label the number honestly than dress it up as telemetry.

What I'd do differently

Instrument the tool itself from day one. A product whose pitch is “better signal” should measure its own: research time, draft-acceptance rate, which sources actually drive decisions. The self-reported ~60% is believable — the PMs vote with daily usage — but the builder of the semantic layer should hold his own tool to semantic-layer standards.