A year ago, I was passed over for the VP of Product role at LiveData. This month, I started the job.

Nothing about that first sentence was fun to type. I've written here about hundred-mile races, van builds, and fly fishing, and every one of those was easier to share than this. But the year between those two sentences changed how I work, what I believe product managers can be, and what my company believes AI can do. That seems worth writing down honestly.

"You Don't Know AI Well Enough"

There was no dramatic meeting, no sit-down where someone walked me through exactly where I fell short. The role went to an outside hire that September. I'm not going to pretend any of it felt good. It didn't. (The dogs heard about it. They were supportive.)

The feedback, when I chased it down, fit in a single sentence: I didn't know AI well enough.

I could have argued with it. I'd been building with AI already: prototypes, internal experiments, the habit was there. And I'll be honest, I never got to find out how much of that sentence was diagnosis and how much was justification. But it was the only actionable thing I had, so I decided to take it literally.

There are a few things you can do with a sentence like that. You can decide it's wrong and let it calcify into a grudge. You can take it to another company and hope it doesn't follow you. Or you can treat it like a spec.

I write specs for a living. More to the point, I have a habit of prototyping my own specs.

One sentence isn't much of a spec. I've built from thinner.

So I Built

That December I enrolled in Agentic AI for Organizational Transformation through MIT Professional Education. Partly for the substance, partly for the signal: if that sentence was going to follow me around, I wanted an answer to it that didn't require anyone to take my word for anything.

The course asked for applied work. I already knew mine. For a while we'd been talking at LiveData about a chat tool for Insights, our analytics product: something people could ask questions of directly instead of digging through dashboards. Everyone liked the idea. It just hadn't cleared the bar for engineering resources yet, and there were reasonable things in line ahead of it. So I built it myself. Nights, mostly, while the course had me thinking about agents anyway.

That prototype became the first of five AI tools I've shipped at LiveData since: the chat tool that grew into a customer-facing product, and four internal ones that changed how our own team does competitive research, strategy, customer success, and sales. The one I lean on most is a Context Engine the team connects to over MCP: ask it anything about our product, sales, engineering, or personas, and it hands you the context you'd otherwise spend an afternoon hunting down.

On nights and weekends I went further afield: a dashboard that reads 85 hospitals' worth of field reports so that no human has to, an ROI calculator where the AI narrates but never touches a number, a pipeline that unifies 1.3 million federal surgical records into something queryable. And somewhere in there, my son and I shipped a soccer card site: he's the PM, I'm the build team. The case studies are on my projects page for anyone who wants the details.

The theme is the same in every one of them. I stopped describing what I thought we should build and started handing people working software.

By spring, building was simply how I answered questions. Someone would raise a problem in a meeting, and instead of scheduling discovery, I'd show up a few days later with a prototype, and we'd argue about something real.

Becoming a Builder

There's a word going around the product world for what I was turning into: a Builder. A product manager who doesn't stop at the document. One who ships.

It's not a term we use at LiveData. Yet. But I believe in it, and part of what I want to do with this role is make it normal here.

A Builder is not a replacement for engineers, and the point was never production code. The point is closing the distance between the person deciding what to build and the person who knows what's possible. For most of this profession's history, those have been different people, translating for each other through documents and tickets and meetings. When they're the same person for even the length of a prototype, the roadmap stops being a list of guesses.

AI is what made this possible. Not because it does the job, but because it collapsed the cost of the first working version from months to days. The judgment about what to build, for whom, and what good looks like: that's still the craft. It's just backed by evidence now.

I have a lot more to say about this. It's next week's piece.

The Company Was Building Too

Here's the part I didn't expect. I thought I was closing a personal gap. What actually happened is that the company's sense of what was possible moved with it.

A working demo does something a slide deck has never done in the history of slide decks: it moves the conversation from whether to when. Show a team the analytics chat they'd been debating, already answering questions, and nobody asks about the technology. They ask what else could work that way. Show a sales team an ROI model that explains itself, and suddenly every department has a list.

A year ago, what we believed AI could do for perioperative care undersold reality. Watching that belief get recalibrated, demo by demo, taught me something I now consider a core truth of this job: capability is contagious. My ceiling and LiveData's went up together, and I no longer think that's a coincidence. Building doesn't just answer the question in front of you. It changes what everyone around you thinks is askable.

The Silver Buckle, Again

I wrote earlier this year about the Vermont 100 and the silver buckle: finish under 24 hours and you get one. The line is arbitrary. Committing to it is not, because the commitment reaches backward and changes every decision from mile one.

Somewhere in those first quiet weeks after the news, I picked a standard: do the job at the level of the role I didn't get. Without the title, and without any promise that a title would ever come.

That second part matters, and it's the part I'd want anyone in the same spot to hear. There was no deal. Nobody said "do this for a year and it's yours." I wasn't building toward a promotion I could see, because there wasn't one to see. I had a sentence, a company full of real problems, and evenings. The standard had to be mine, or it wouldn't have changed how I ran.

What Comes Next

This month I stepped into the VP of Product role at LiveData, reporting to our CEO. It's the job I wanted a year ago. I won't claim the delay was a gift. But I know exactly what got built during it, and I know the version of me who wanted the title then couldn't have done what the job actually needs now.

What I want for product here is a longer conversation than this piece can carry: AI in the operating room built carefully enough to deserve the trust it's asking for, product managers who build, and a team where capability keeps being contagious. I'll be writing about all of it: product leadership, building with AI, teaching PMs to become Builders, and what this looks like inside healthcare, where the stakes don't forgive sloppy work. If any of that is your world, stick around.

And to the team at LiveData: thank you. The title is new. The people I get to build with are the best part of it, and they aren't.

A no is a spec. Build to it.