Harvard sells a $699 bootcamp with AI avatars of its faculty

699 dollars. That is the entry price for HBS Foundry, the entrepreneurship program Harvard Business School has packaged with a detail worth looking at closely: a good part of the feedback does not come from its instructors in person, it comes from AI avatars built from them. The avatars step in exactly where it hurts, during rehearsals: practice pitches and mock board meetings. TechCrunch reported it on 22 August.

The price tag is half the story. A program signed by Harvard Business School at 699 dollars lives in a different order of magnitude from the same school's MBA, and that gap only holds up if something gets automated. What gets automated here is precisely the most expensive thing in a classroom: the time of a senior professor listening to a student repeat their pitch for the sixth time.

What gets automated and what does not

It helps to separate two things that headlines always bundle together. One is content: videos, readings, cases. That has been packaged for decades and needs no avatar. The other is the response: listening to someone, spotting where the argument falls apart and handing it back with judgment. That second part is the one that does not scale, the one that gives an in person program a capacity limit and a waiting list.

HBS Foundry goes after the second one. Putting the avatar in the practice pitch and the mock board, and not in the lecture, is a sensible product decision: those are the two moments where a founder needs repetition, and repetition is exactly what a human with a calendar cannot provide.

Why it matters beyond Harvard

The move reads clearly for anyone who trains people inside a company. Sales onboarding, negotiation drills, code reviews on junior teams: they all share the same structure. Someone practices, someone returns judgment, repeat. If a school with that brand agrees to put its professors' faces into an automated feedback system, the corporate market will look at the precedent before it looks at the technology.

And that is where the nuance gets lost in the avatar conversation: the video is the easy part. Generating a believable face and voice is bought technology, not built technology. The hard part is the judgment behind it. An avatar that returns generic praise is worth nothing; one that catches that your unit economics do not line up with your CAC is worth quite a lot. That is not a video generation problem, it is a rubric, context and evaluation problem.

The cost of lending your face

There is a bill schools have not finished calculating. An instructor's credibility comes from the fact that they answered. When their likeness answers a thousand students at three in the morning all at once, that credibility transfers to the system, and the way back is not obvious. Licensing your image and voice is a contract with long consequences, especially if the program changes vendor or owner.

There is also the quality floor. In an in person cohort, weak feedback gets noticed and someone corrects it in the next session. In an automated system, an error of judgment replicates identically a thousand times and nobody raises a hand, because the student has nothing to compare it against.

Who this is useful for

For an early stage founder with no access to mentors, 699 dollars for an environment where you can fail a hundred times without spending anyone's social capital is a reasonable proposition. For an internal training team, the pattern is reproducible today with pieces that already exist: an agent with a clear rubric, real business context and a channel to practice in. The avatar is the optional part.

What that price does not buy is the network. A Harvard program is worth a lot because of who sits next to you, and an avatar can fix your pitch but will not introduce you to anyone. We think it is an honest experiment as long as the school stays explicit about which part of the feedback is human and which is not.