Harvard Business School’s pitch for its Foundry bootcamp goes something like this: entrepreneurs need reps, and faculty time doesn’t scale, so the school built AI replicas of its instructors to sit in on practice pitches and mock board meetings. Price of admission, $699.
My first reaction was not outrage. It was recognition. I have reviewed maybe forty AI tools this year that make the same promise in cheaper clothing — an avatar, a voice, a face, feedback on demand. What Harvard added is the one thing none of those tools could buy: a name that makes the output feel like it counts.
What’s actually being sold here
Strip the crimson off and the product is a practice partner. You pitch, it responds, you adjust. That is a real category with real value. Founders are chronically under-rehearsed, and the reason isn’t laziness, it’s that live audiences are expensive. Nobody wants to burn a warm intro on pitch attempt number three. An avatar that will sit through attempt number three, and eleven, and nineteen, without checking its phone, solves a genuine bottleneck.
The mock board meeting is the more interesting piece. Pitch practice is well covered by existing tools. Board dynamics are not. Learning how to absorb a pointed question from someone who technically works for you, in front of four other people, is a skill most founders acquire by failing at it publicly. If the avatars simulate that pressure with any fidelity, that is the part worth paying for.
Where I’d push back
My standing question for any avatar product is what generates the feedback. There is a large gap between a model trained on a specific instructor’s written work, case notes, and grading patterns, and a general-purpose model with that instructor’s face rendered on top. Both look identical in a demo video. Only one of them is teaching you what that person actually thinks.
Nothing in what’s been published resolves this. The avatars are described as AI-generated replicas providing feedback. That’s a description of the interface, not the reasoning behind it. If you’re buying access to Harvard faculty judgment, you deserve to know how much of that judgment made it into the system, and that answer should be in the product page, not inferred from marketing.
Second concern, and this one applies to the whole category. Practice partners that never get tired also never get annoyed, and getting annoyed is information. A real investor who has heard your preamble twice will cut you off. A real board member who thinks your numbers are soft will let the silence sit. Models tuned to be helpful tend to sand those edges down. If you rehearse against something agreeable, you will be calibrated for an audience that doesn’t exist.
The $699 math
Judged as a course, $699 is unremarkable for anything with Harvard Business School on it. Judged as software, it’s steep — you can assemble a rough version of this from consumer AI tools for a monthly subscription and an afternoon of prompt writing. So the honest framing is that you are paying for curriculum, sequencing, and institutional credibility, with the avatars as the delivery layer.
That’s defensible. It also means the value depends almost entirely on the quality of the underlying program, which is the least discussed part of the whole story. The avatars got the headlines. The pedagogy is what determines whether the $699 was worth it.
What I’d want to test
If I were reviewing this properly rather than reacting to it, here’s my checklist:
- Does the avatar’s feedback change meaningfully between a strong pitch and a weak one, or does everything get a supportive nudge?
- Will it interrupt, or does it wait politely for you to finish?
- Does it remember session one when you show up for session four?
- Can you tell which instructor you’re talking to from the substance alone, with the video off?
- Do the human faculty review any of this, or is the loop entirely synthetic?
That last one matters most. An AI practice layer that feeds into eventual human review is a smart use of scarce expert attention. An AI layer that replaces human review entirely is a different product wearing the same face.
My read
I’m not mad about this one. Simulation has been quietly useful in professional training for decades, and rehearsal is exactly the kind of repetitive, low-stakes work AI handles well. Harvard is doing openly what plenty of platforms do without disclosure, and the $699 tag at least sets an honest expectation about what tier of experience you’re buying.
What I want is transparency about the machinery. Tell buyers how the replicas were built, how closely they track the real instructors, and where a human enters the process. Do that, and this is a reasonable tool with a famous logo. Skip it, and it’s a very well-branded chatbot.
🕒 Published: