There’s a common misconception that AI distances us from the work by automating human touchpoints. In my experience, the opposite is true: AI enables boots-on-the-ground leadership.
My approach: Player-coach
My background in consulting first showed me the value of finding problems and shaping solutions. A mentor pointed out I was already doing product work, which helped me name what I liked most: helping teams turn technical requirements into real tools.
That thread eventually brought me to Freestar, where I’ve spent the last three years moving from individual contributor to Director of Product, supporting our product portfolio, teams, and leaders. Lately I’ve been more involved with the data team and publisher tooling, while staying connected across the wider portfolio.
I think of my role as player-coach. I’m in the field with the work, and I’m also helping unblock, mentor, and support others as they own their areas. That combination works because the more time I spend inside the work, the better I understand where teams are stuck, what publishers need, and how decisions connect to customer outcomes.
AI as the enabler, not the endpoint
I was an early adopter in asking for access to Claude Enterprise because I wanted to test where it could help in product operations. Take release notes, for example, which used to mean updating a central log, pulling details into decks, and repeating manual steps every time something shipped. We’ve now automated that flow so Claude can run on a schedule, check Jira for completed work, and post an update to Confluence. That gives the team about five hours of manual work back per week.
But here’s the part I want to underline: When we saved those five hours per person, we didn’t use them to sit back and look at high-level dashboards. We reinvested them directly into our people. It gave me the bandwidth to connect one-on-one with the team members I lead, understand where they’re getting stuck, and unblock them fast. That time also lets us get into the trenches together: talking directly to publishers, sitting with our data and trafficking teams, and addressing operational drag in real time.
AI has changed how I prototype, too. Instead of writing a long requirements document and asking everyone to imagine the outcome, I can quickly show how something could look and behave, testing a direction in the field in days rather than waiting months for a full engineering cycle. That speed matters right now, while our data team collects more first-party data and we explore performance across publisher pages and more granular reporting.
Collaborating across functions
That approach to AI showed up during one of our company offsite retreats. Because we work virtually, the retreat was fully remote and held across two four-hour sessions. The first day introduced tools and examples from team members across the business; the second moved into an idea-a-thon where cross-functional groups had about ninety minutes to identify a problem, use AI to solve it, and present what they’d made.
The range of outputs made the exercise worthwhile. Some teams came back with prototypes, some had working AI-based solutions, others used AI to build a deck or visualize an idea more clearly. That variety showed AI doesn’t have to be one kind of project; sometimes it helps you build the thing, and sometimes it helps you explain it well enough for us to make a better decision.
The cross-functional mix made it stronger, too. Product and engineering staff were paired with teammates from legal, finance, and trafficking, so the problems surfaced came from places we don’t always hear from. One teammate described living in spreadsheets and wanting to eliminate that drag: exactly the kind of operational friction product teams need to hear before they can help solve it.
Listening to the publishers and closing the loop
I find the strongest product ideas sharpen after direct customer conversations. Freestar’s publisher-first ethos is practical for me, not merely a mantra. When I look at priorities, I’m asking what each item does for the customer, what outcome it creates, and whether it solves the right issue, not whether we can check a box.
Our Publisher Forum is one of my favorite examples of “boots on the ground” in action. This year, our second annual event brought together a small group of trusted publishers for candid conversations about their current challenges, and that feedback has directly shaped how we run the business and build pubOS.
The forum surfaced a specific, evolving pain point: publishers were vocal about how AI search and answer engines are pulling clicks away from the open web, making traffic feel more questionable and volatile than ever. It’s the same technology, cutting the other way: not a tool they’ve adopted, but a force reshaping the ground they stand on.
This is why the bandwidth mattered. Because our team had the time back (time AI itself had helped free up), we could take that feedback and prototype a response quickly. That direct input was the catalyst for our shift toward audience packaging and direct deals, helping publishers get closer to buyers and secure their revenue on their own terms.
It’s a reminder that good product work isn’t just about technology. High tech only earns its keep when it buys back high touch: more face time and communication with the people who live these industry shifts every day. Right now, those industry shifts are largely being driven by AI itself, which means the same force creating uncertainty for publishers is also the tool giving us the capacity to help them navigate it.
As we look ahead, the central challenge remains unsolved: how to maintain and prove publisher value as open-web traffic becomes more uncertain. Keeping the publisher at the heart of our roadmap means acknowledging that uncertainty and continuing to build tools that prioritize transparency and direct relationships, using the very technology that’s disrupting the old model to help build what comes next.
