Most companies rethink a team model after it fails. HubSpot moved beyond its pod model for go-to-market artificial intelligence work after the pods succeeded.
The cross-functional pods brought subject matter experts, engineers, data specialists, and other contributors together under one leader. They ran for about a year. Jon Dick, HubSpot's chief customer officer, says they made real progress and moved every key performance indicator they targeted.
But the people doing the work still belonged to different teams. They carried competing priorities. Bigger goals required more coordination than HubSpot wanted.
So the customer-platform company made a bigger organizational change. Roughly three months before Dick spoke with us on Episode 234 of The Artificial Intelligence Show as part of our AI Transformations series, HubSpot moved the specialists working on go-to-market AI systems into direct-line ownership under one leader.
The lesson?
A model built to coordinate borrowed specialists can become the constraint once a company needs the team to commit to bigger goals.
Keep reading or get the full story in Episode 234 of The Artificial Intelligence Show, as part of our special AI Transformations series presented by Google Cloud.
HubSpot did not begin with a centralized team. Its first phase was intentionally loose: Employees received tools, joined hackathons, shared experiments in Slack, and learned by building.
That freedom surfaced ideas and helped employees become fluent in AI. It was well suited to individual experimentation. But HubSpot wanted more than one employee completing one task faster. Dick says the larger opportunity was institutional productivity, where AI could move company results.
The pod model was HubSpot's first answer. The company assembled people from different functions around go-to-market problems and put the pods under Kieran Flanagan, an experienced marketing and growth leader at HubSpot.
"We did that for about a year, and we got really great progress," Dick says. "We moved all the KPIs we were working on."
HubSpot did not move away from pods because they produced nothing. The pods showed that engineers, data specialists, and business experts from different teams could move the same outcomes together.
But the pods also exposed the limits of coordinating those people without fully owning their time.
The issue was that a pod could create one working group without creating one chain of command.
The specialists could share the pod's goal while still reporting into teams with priorities of their own. And while Flanagan did lead the work, the people doing it were not on his team.
That left him coordinating capacity he did not control. Bigger goals still had to compete with other commitments, and each conflict added coordination.
"The pod model was great, but in the pod model there were still competing priorities, and it was a lot of coordination cost," Dick says.
For an experiment, that trade can make perfect sense. A company gets the right people around a problem without redesigning the organization before it knows whether the work matters.
But once the work proves valuable, the same arrangement can limit the size of the commitment. HubSpot wanted the person leading its go-to-market AI work to pursue bigger goals without renegotiating priorities for every project.
HubSpot's next move was structural. The engineers, systems specialists, data experts, product people, and subject matter experts working on go-to-market agentic systems began reporting directly to Flanagan.
At the time of our conversation with Dick, that model was only about three months old. The earlier KPI gains belonged to the pods. The direct-line structure was HubSpot's current bet on what could help the company go further.
Dick says the change lets Flanagan commit to bigger goals and execute at a higher pace. He also expects one team to build fluency across skills that are changing quickly, from go-to-market and software engineering to data science and machine learning. In Dick's view, that concentration should also help HubSpot attract and retain people who want to deepen the craft.
In a pod, the leader coordinates people who still belong to other teams. HubSpot moved to direct reporting so Flanagan could set priorities for the people doing the work.
Putting the specialists under Flanagan required functional leaders, including Dick, to give up some day-to-day control. Dick says he is comfortable doing that because he and Flanagan are measured against the same KPIs.
That alignment keeps Flanagan's team tied to the same business outcomes as the functions it serves. It also gives him room to set priorities without asking each function to release people project by project.
Centralizing people does not create that alignment by itself. The leaders involved must agree on what success means, accept less direct control over the specialists, and trust the shared measures.
HubSpot's new model did not erase trade-offs. It changed which side of the trade-off won.
"The hardest thing about it is always trade-offs," Dick says. "Local priorities versus global priorities."
A centralized team makes global priorities harder to dilute. That focus matters when mediocre execution across many ideas produces mediocre results. But it also means a region or function may have less freedom to pursue a smaller experiment that matters locally.
HubSpot had faced a version of this decision before with search engine optimization. So, the company used two questions to decide between regional and central ownership: Does specialist expertise create a meaningful edge, and would a central team help attract people who care deeply about the craft?
Dick applies the same logic to go-to-market AI. He believes a central team can build deeper expertise and become a stronger home for specialists. The cost is that some good local ideas will wait.
Make no mistake: pods remain useful. They can connect business experts with technical talent, test whether a problem deserves investment, and move results before a company is ready to redraw reporting lines.
The leadership mistake is assuming that a successful pod must therefore be the permanent model.
Once the work becomes important enough to demand bigger goals, leaders should ask a harder set of questions:
If the answers point toward direct ownership, the pod may not have failed at all. At HubSpot, the pod year showed that the work could move KPIs and exposed the coordination cost of depending on people who still belonged to other teams. That was enough for the company to try direct ownership.
Did you enjoy this transformation story? Go deeper on how AI is reshaping work and business with The Artificial Intelligence Show. Each week, we break down what matters in AI and what leaders should do about it.