Harvard Business School's $699 HBS Foundry startup bootcamp now deploys AI avatars built on the likenesses of its actual instructors — not generic chatbots — to critique participant pitches and run simulated board meetings.
Key takeaways
HBS Foundry is a $699 online startup bootcamp from Harvard Business School that uses AI avatars modeled on its real instructors.
The avatars provide real-time feedback during practice pitch sessions and simulated board meetings, replacing live instructor face time.
This is one of the first high-profile deployments of instructor-likeness AI avatars in a paid professional-education product, not a free demo.
The program signals a broader shift: AI companions trained on specific human experts, not generic models, are entering professional coaching workflows.
For creators building AI characters or companion personas, HBS Foundry is a concrete, public proof-of-concept for domain-specific avatar products.
What HBS Foundry's avatars actually do
In the program, participants don't just read case studies — they pitch to an AI version of a named HBS instructor and get structured feedback, according to TechCrunch. The avatar can push back on weak assumptions, ask follow-up questions, and simulate the pressure of a board room. That's a meaningfully different use case from a generic AI chat assistant: the avatar carries the instructor's framing, vocabulary, and evaluation criteria, not just a system prompt telling it to "act like a business professor."
The $699 price point matters for context. This isn't a free tier or a research prototype — it's a commercial product where the AI avatar is the primary delivery mechanism for instructor contact. Harvard is, in effect, testing whether a domain-specific AI persona can substitute for synchronous human coaching at scale.
HBS Foundry deploys AI avatars of named Harvard instructors to critique startup pitches in real time.
Building a convincing domain expert avatar requires more than a voice clone and a headshot. The underlying model has to be fine-tuned — or at minimum heavily prompted — on that instructor's actual feedback patterns, the frameworks they favor, and the kinds of objections they typically raise. Get it wrong and participants notice immediately: a pitch coach who never challenges the revenue model isn't useful, regardless of how realistic the avatar looks.
This is exactly the challenge that anyone building AI companion personas for professional or educational contexts runs into. Generic instruction following produces generic avatars. Specificity — the particular critique style of a particular expert — is what makes the interaction worth paying for. If you're experimenting with character design on Charmloop's AI companion tools, the HBS Foundry model is a useful benchmark: the persona's value comes from what it knows and how it pushes back, not just how it looks or sounds.
The likeness question no one has fully answered
Using a real instructor's likeness — name, appearance, voice, and intellectual style — in a commercial product raises questions the industry hasn't settled. The Robin Williams estate's recent move to reclaim control of his digital likeness illustrates how quickly likeness rights become contested when AI is involved. HBS Foundry presumably has instructor consent baked into employment or program agreements, but the model it's testing — pay-walled access to an AI version of a specific named human — will be replicated by others who may not have that paperwork in order.
For creators building character-based AI products, that's the practical signal here: the technology to do this is clearly available and commercially viable, but the legal and ethical scaffolding around instructor or celebrity likeness is still being written in real time.
What this signals for AI-character workflows
If you're a creator who builds AI personas — whether for entertainment, coaching, or interactive fiction — HBS Foundry is worth watching as a workflow template. The architecture is: real human expert → fine-tuned or heavily prompted AI persona → interactive scenario (pitch, board meeting, critique session) → structured feedback loop. That same pattern applies to a fantasy dungeon master persona, a film-directing coach, or a creative writing mentor.
The $699 price also suggests the market will pay for specificity. A well-constructed domain expert persona commands a premium over a generic AI chat interface. Browsing Charmloop's model catalog for character-capable base models, then layering in domain-specific persona prompting, is the practical starting point for anyone looking to build in this direction.
