Bree Thomas Connect
Bree Thomas

17+ years building product · Former CPO · Denver

Product. Teams. AI.

Product leader who builds agent-first: AI in production, whole teams building with agents, and pricing that holds up when every user costs something different.

AI is the product, not a feature.

The model sets what's possible. The product, the team and the pricing get built around it.

Evals are a loop.

Real conversations get reviewed and coded continuously, and new failures become new evals. They show where quality stands. Product decides what to do about it.

Inference pricing is product work.

What it costs depends on product choices: which model, how much context, how often it runs. Price for how that varies. The average user doesn't exist.

Build AI for the team.

Most AI is one person, one chat, one agent. But the work happens across a team. When a team and its agents share the same context, the coordination drops away and people get more time for the decisions they already own.

Ship.

Homebot logo

Homebot · VP Product · 2025 - Present

A home wealth platform for lenders and homeowners, in the middle of a shift: from software with AI features to AI as the product.

Leading that shift. Evals run as a continuous system: live conversations reviewed, coded and turned into evals that show where quality stands. Pricing starts from how cost varies user to user, not from the average. And the product team builds with AI every day, not just ships it.

AI-first product orgContinuous evalsAI pricing framework
Balance logo

Balance · Co-Founder · 2022 - Present

Banks assume variable means unpredictable. It doesn't. These workers know exactly what a Friday night shift earns.

My hands-on lab for building AI on a real product. A conversational assistant with 12 tools, tested with an LLM judge plus deterministic checks across a 100+ case eval set, and traced in production. The one tool that edits someone's financial records runs only after they approve the exact changes, enforced on the server.

12-Tool Agentic Assistant100+ Case Eval SetServer-Enforced Confirmation
Partytrick logo

Partytrick · Head of Product · 2025

Two years of spinning. Nothing shipped. Team had lost confidence and the product had no clear path to market.

Cut the contractor team by half and brought in laser-focused senior engineers. Realigned the org around one pursuit: PMF. Migrated from Jira to Linear for a fresh start and killed six months of stale backlog. Started from a working prototype in Replit. It aligned stakeholders, set the data structures with engineers and became the first design system. Design took on the UX problems the prototype couldn't solve.

<40 Day Ship60% Burn ReductionMarket Launch
Vangst logo

Vangst · Chief Product Officer · 2020 - 2022

No product function. Stalled releases. Burn too high. Product and engineering disconnected from business strategy.

Replaced contractors who'd over-engineered for scale they didn't have. Brought in senior engineers to fix a stalled data migration. Trained a services-culture org on iterative product development. Shipping went from stalled to weekly.

$19M Series B40% Burn Reduction12-Person Team
Muve Health logo

Muve Health · VP of Product · 2017 - 2019

EMR platform was failing with providers, feature velocity was slow, and the product team lacked clear direction.

Embedded in the OR to study clinician workflow. Discovered EMRs were built for billing, not diagnosis. Built clickable iPad prototypes and ran usability tests with each clinician before writing code. Redesigned around real-time clinical signals that drive treatment decisions.

90-Day Turnaround100% Provider Adoption

Open source.

Bette

Bette

An open-source AI sparring partner for product decisions, built on Claude Code. Opinionated: it pushes back on your product calls, picks the right framework for the question, and stays out of your way until you ask.

View on GitHub →

Story.

My career happened in thirds.

First, marketing. DTC brands like Oakley, North Face, and Beats by Dre. Three Super Bowl campaigns with Audi. These weren't just advertising. We had engineering teams building software for these brands.

Then I learned to code. Took a detour to Turing School because I wanted to understand how things actually get built. Founded the Denver chapter of Girl Develop It. On weekends, I taught women to code.

Marketing foundation plus shipping production code led to product. I've been doing product since before it became the discipline it is today.

I'm the one you call when a product is stalled and needs to ship. Startups, Fortune 500s, stalled products, scaling teams. I've done it across verticals.

Engineers want to work with me. I speak their language, understand their constraints, make tough calls, and protect their focus. I build products WITH my team, not just for them.

I don't rely on assumptions. I study what people actually do. User empathy isn't a phase of the project. It's how I think.

My undergrad was art studio. I painted and drew my way through college. Then law school. Passed the Colorado bar. That breadth made me dangerous: creative thinking, critical thinking, logical reasoning.

I still build. Balance and Bette keep me close to the work my team does every day, and they scratch a creative itch that goes back to art school.

Lately I'm hooked on multiplayer AI. What happens when a whole team and its agents work from the same context, at the same time?

"For years, the design-eng-product dynamic got predictable. AI flipped it on its head. I'm here for it."

Talks.

Continued learning

  • Prototype to Production · Aman Khan, 2025

    Agentic prototypes (RAG, tool calling), evaluation frameworks

  • AI Evals for Engineers & PMs · Parlance Labs, 2025

    LLM evaluation pipelines, LLM-as-judge, CI/CD integration

Education

  • JD, University of Denver
  • BFA, University of New Mexico
  • Full-Stack Engineering, Turing School