Case Studies · Product · AI

Decisions I have made.

Product manager at OPENLANE, currently shipping AI features inside a $20M ARR subscription suite. The pattern across all of it: the hardest part of an AI product is not the model, it is designing the moment where a user has to decide whether to trust it. These case studies are about that moment, covering what I recommended, what I pushed back on, and what shipped.

anthony.brady.f@gmail.com · LinkedIn · Toronto

Secret-cleared · Government of Canada · 2022

What I'm looking for next

Cohere is building AI for institutions that cannot afford to get it wrong. Every case study here runs into the same wall. Users do not yet trust the output, and earning that trust without hedging the capability is the entire product problem. That has been my actual work for three years. I want to solve it where the stakes are highest.

Case studies

Agentic AI

Designing Human-in-the-Loop for Agentic LLM Workflows

Two patterns I shipped at Delo and the decision logic behind them. Why I routed the user's trust contract through the document, not the model.

7 min read · May 2026 →

Product discovery

From Mixpanel Anomaly to LLM Extraction with HITL

A Mixpanel anomaly pointed at the wrong problem. The right one was in the data we had not looked at yet. How I pivoted the scope before the sprint locked.

6 min read · May 2026 →

AI UX

From Forecast to Signal

The pricing model was accurate. Dealers were not coming back. How I pushed to convert a depreciation curve into a decision a dealer could actually act on.

8 min read · May 2026 →

Product strategy

Building a Daily Habit

The strategy bet behind ProPower. How I framed an inventory product as the dealer's daily money-at-risk surface.