HubSpot’s CRO treats early AI revenue as a hidden trap

This article summarizes an episode of Kleiner Perkins’s video series featuring Mark Roberge, founding CRO of HubSpot.

Mark Roberge, founding CRO of HubSpot / Photo credit: Mark Roberge

For an AI startup, early revenue can be a dangerous illusion. Mark Roberge, the founding CRO of HubSpot, alongside Joubin Mirzadegan, CEO of the AI sales software company Roadrunner, warns founders that early sales do not mean a company is ready to scale.

In an era where AI makes software easy to build, the true test of survival is whether customers stick around, expand their usage, and provide enough concrete evidence to justify hiring a sales team.

High stakes are the only real defense

The core problem for today’s AI founders is the copycat threat. Because foundation models are constantly improving, competitors can easily clone basic software.

To survive, a startup must handle mission-critical work that could ruin a financial quarter, breach a contract, or break a customer promise.

In these high-stakes environments, buyers pay vendors specifically to take the blame if something goes wrong. Mirzadegan points out that businesses will never trust a fragile AI wrapper with tasks that could cost people their jobs.

“If [Roadrunner] goes down, people are getting fired and we’re getting sued,” he explains. “If we are running HubSpot’s CPQ (Configure, Price, Quote) system and it is the last day of a quarter, Roadrunner goes down, and HubSpot misses its number, we’re getting sued.”

Deep workflows beat internal IT teams

Even startups that solve high-value problems face another challenge: customers may try building their own AI tools internally.

Roberge and Mirzadegan agree that creating a prototype is easy. Maintaining a production-ready system, however, requires managing complex business rules, pricing logic, integrations, and workflows across multiple departments.

That complexity becomes a competitive advantage. Even when companies have the technical capability to build similar software, maintaining it often distracts engineering teams from higher-value work.

“It is easier to build. That is fair,” Mirzadegan notes. “But you still have to expend a bunch of engineers on something that is just not very strategic to your business.”

Proof must come before salespeople

Once a founder builds this sticky, hard-to-copy product, the immediate temptation is to hire a massive sales team.

Roberge argues this is where most startups fail. Before a company scales its sales force, it must definitively prove that the product actually works for the users.

Venture capital does not dictate hiring