The avatar looks and sounds exactly like the professor. It's slightly too smooth, slightly too consistent, and according to Harvard's own students, genuinely useful anyway.
Harvard Business School built H.B.S. Foundry, an eight-week entrepreneurship bootcamp priced at $699, around AI video avatars of real faculty.
Jeff Bussgang, a Flybridge Capital co-founder and HBS senior lecturer, is one of several professors who volunteered to have themselves cloned.
How Foundry Actually Works
Bussgang's own read on his AI double is blunt. It's "creepy," he says, but adds that his students love it anyway.
Foundry walks founders through refining an idea, pitching investors and customers, and running a startup, combining HBS course content with a peer community and AI agents built to represent specific instructors. Video call simulations let participants practice pitches, sales calls and board meetings against avatars like Bussgang's.
The product has already been tested at more than 100 universities across all 50 states. About 45% of summer session founders paid individually, with the rest sponsored through university partnerships, and roughly 760 founders found the program through their schools over the summer.
Making the AI Tough, Not Flattering
Harvard grades on a forced curve, and the team building Foundry had to deliberately train its AI to avoid the obsequiousness large language models default to.
Faculty co-chair Tom Eisenmann said the agents are tuned tough enough that almost no startup pitch gets approved on the first attempt. Some participants have spent ten hours refining a pitch against the AI just to get past that first bar.
The agents draw primarily from the specific HBS faculty members they're modeled on, pulling in broader knowledge only when needed. Harvard partnered with AI avatar company HeyGen to build the video versions.
Why Students Say They Trust the Clones
The positive reaction to the video avatars specifically surprised the team building them, according to project director Katharina Rings.
Simply attaching a professor's photo to a chatbot measurably increased engagement in testing. Rings said the pattern was consistent enough to be unmistakable: users trusted the AI more, and preferred it more, when it was tied to one specific, named person rather than presented as a generic assistant.
Founders using the program described real practical value. A Boston University PhD candidate building an underwater propulsion startup said the AI avatars pushed him to narrow a scattered pitch into one focused use case, and valued being able to reach a version of an advisor at any hour rather than working around a professor's calendar.
An entrepreneur in residence at UMass Boston who has already raised more than a million dollars for her company called the board-meeting simulations one of the best features, since they let her rehearse a high-stakes conversation before it actually happened.
Even Eisenmann, who helped build the system, points to something outside the AI as the program's most popular feature: the peer forums where founders talk directly to each other.
"They love the A.I. They're using the A.I., they're learning a lot," Eisenmann said. "But it keeps coming back to humanity."
What This Means for Miami
Harvard's model of scaling faculty through AI avatars is exactly the kind of infrastructure Miami's own entrepreneurship programs could adopt. FIU and the University of Miami both run mentorship capacity that's thin relative to demand.
It's worth reading against a pointed comment elsewhere in the same day's coverage. Representative Ro Khanna, discussing where AI investment and talent are actually concentrating, said flatly that the AI revolution "is not happening in Austin. It is not happening in Miami."
That's a real claim from someone with a national platform, not an easy one to wave off.
Building genuine infrastructure like Harvard's is one way to answer it. A headline-grabbing company relocation isn't.