Squads Ventures: building a measurable product operating system for Referent

Context & problem

At Squads Ventures, I work on Referent, its AI SEO product. It needed faster, more reliable product decisions while customer feedback, delivery work, analytics freshness, and AI operating costs lived across separate workflows.

Discovery & evidence

I synthesized intelligence from 24 client-operations conversations across 9 accounts, alongside signals from 13 commercial opportunities. This created an evidence base for recurring pain points, product gaps, and opportunities - not a backlog driven by the loudest request.

Decisions

Designed a dual-track, continuous Discovery and Delivery model in Kanban so learning and execution could run in parallel.

Introduced ICE prioritization and WIP limits to make sequencing explicit, protect focus, and expose blocked work.

Made analytics freshness and variable AI cost observable product concerns.

Delivery

I owned the operating design, evidence synthesis, prioritization approach, and initiatives that made critical signals measurable.

A scheduled analytics workflow brought freshness to 21 of 22 organizations and reduced a key manual refresh from 15 minutes to 3 minutes - about an 80% reduction.

A warm-audit approach reduced a relevant AI operation's internal cost by 86%.

What I owned and what the team built

I owned: framing the operating problem, the dual-track model, evidence synthesis, prioritization mechanics, and decision-making around measurability.

The team built: the production workflows, automations, and product improvements that made the system real. The outcomes are team outcomes; my contribution was connecting product judgment to technical execution.

Learning

For AI products, a roadmap is not enough. You need a system that continuously connects customer evidence, delivery capacity, product quality, and unit economics. The job is to make the next decision better than the last.