I build AI products where the leverage is real: operating systems, workflow agents, decision surfaces, and platforms that help teams move from scattered experiments to measurable business capability.
The thesis
Pat Mullee is a Director-level AI product leader in Chicago who builds AI copilots, agents, and decision-support platforms for enterprise professionals. He leads AI product management at Storable and spent six years at FourKites as Group Product Manager for AI and Machine Learning, where he grew an AI portfolio to $18M+ ARR across 1,200+ enterprise customers and shipped a US-patented real-time ETA platform. He holds an MBA from Indiana University's Kelley School of Business.
The teams that win the next decade will not be the ones with the most pilots. They will be the ones who can frame a portfolio of bets, govern agents that operate on real data, and build the infrastructure that makes the rest of the work economic.
That is the layer I’ve spent the last decade working in: above individual features, close enough to the product and engineering work to make it real, and close enough to the business to know whether it matters.
How I think
The valuable work is no longer just adding AI into a workflow. It is deciding where AI belongs, what humans still own, how the system is governed, and how the economics improve as usage grows.
Most companies have enough AI ideas. They need someone who can shape the portfolio, choose the right build or buy path, define controls, and move the work from executive interest to shipped capability.
A production agent needs durable state, review gates, clear boundaries, and configuration that lets the same engine serve different teams or customers without becoming custom software every time.
Where I fit
A governed hub where many teams build, review, and monitor their own AI agents. Built one for self-storage operators at Storable; the Agent Control Plane demo shows the pattern.
A real-time ETA platform on millions of daily shipments, a US patent, and 400+ industrial sensor streams at GE. ML run as a product, not a science project.
Guardrails, evaluations, and audit trails shipped in every demo on this site, because in a fiduciary setting an answer is only useful if it is traceable.
Tools judged by whether experts actually use them, not by launch day. Adoption is a first-class measure in how I run a product.
The proof
At FourKites, the work was prediction, logistics intelligence, and machine learning products that had to operate inside real customer workflows. At Storable, the work has been GenAI infrastructure, customer-facing AI assistants, and agentic product patterns that need to be useful, governable, and commercially sensible.
The demos here are not meant to be novelty apps. They are examples of the kind of product judgment I want to bring into a larger role: where to apply AI, how to make it usable, what controls matter, and how to make the business case legible.
Collaboration
I regularly work through AI product ideas, operating models, demos, and implementation plans with people building in similar territory. That can be a focused conversation, a strategic consultation, or a longer project where the scope is turning an ambiguous AI opportunity into something concrete enough to fund, build, or test.