Harvard Business School’s $699 bootcamp is using AI avatars to coach founders

Harvard Business School is using AI avatars of its instructors to give entrepreneurs feedback in a new eight-week bootcamp, underscoring how quickly generative AI is moving from a classroom talking point to a hands-on teaching tool. The $699 Foundry program blends weekly live instruction with AI-driven practice sessions for pitches and board meetings, signaling that elite business education is experimenting with new ways to scale individualized coaching.

The avatars, built by video-AI startup HeyGen, are designed to let participants rehearse in a more structured setting and receive feedback outside the live classroom hours. The approach is one of the clearest examples yet of a major business school embracing synthetic media not as a novelty, but as part of the core learning experience.

That matters because Foundry is aimed at entrepreneurs who want practical, repeatable guidance on how to present their companies, handle objections and sharpen their messaging. Instead of relying only on a weekly live session, students can practice against digital versions of instructors and, in at least one case, a digital version of a well-known venture capitalist.

What Harvard’s Foundry program is trying to do

Harvard Business School is trying to make startup coaching more accessible, more frequent and more personalized. Foundry is an eight-week bootcamp for entrepreneurs priced at $699, a level that places it well below many executive education offerings while still drawing on the HBS brand and its network of instructors.

The program includes live weekly sessions, but its AI component is what makes it unusual. Rather than limiting instructor input to scheduled discussions, the bootcamp gives participants a way to continue practicing with AI-generated teaching avatars that can respond during rehearsals and provide feedback on their performance.

That setup reflects a broader problem in founder education: people can often attend a workshop once, but they rarely get enough one-on-one practice to improve quickly. By using AI avatars, HBS appears to be trying to extend the reach of its instructors without requiring them to be available every time a student wants another round of feedback.

How do the AI avatars work?

The avatars are used during practice pitches and simulated board meetings. Participants can rehearse in front of them and get guidance in a setting meant to feel closer to a live mentoring interaction than a self-serve chatbot exchange.

According to the program’s description, the AI avatars are not replacing the human instructors who lead weekly sessions. Instead, they function as a supplemental layer of interaction, giving founders a chance to test ideas, make mistakes and refine their delivery before they face actual investors, customers or board members.

The avatars were created by HeyGen, a startup known for generative video tools that can produce highly realistic digital presenters. In this case, those tools were adapted for an educational setting, turning faculty likenesses into interactive coaching interfaces.

Why Harvard Business School chose avatars instead of a chatbot

Harvard’s project director said the school originally imagined a simpler chatbot-style experience, but students pushed for something more guided. That shift helps explain why the final product uses avatars rather than a basic text interface.

Project director Katharina Rings said the initial concept leaned more toward a conventional chatbot. Once the school released a trial version, however, student feedback suggested that participants wanted more structure and direction. The result was a more curated experience, with avatars acting as the visible face of the instruction.

That distinction is important. A chatbot can answer questions, but an avatar can feel like a coach, a critic or a role-play partner. For a bootcamp focused on pitching and boardroom scenarios, that visual presence may make the practice feel more realistic and more engaging.

It also reflects a trend in education technology: when users want feedback, they often prefer systems that simulate a human interaction rather than abstract text outputs. In business education, where confidence and delivery matter as much as content, that human-like feel can be especially valuable.

How students reacted to the AI instructors

Early participants appear to like the avatars, even when they find them a little uncanny. That reaction suggests the technology is landing somewhere between useful and unsettling, which is often where early adoption begins.

Some college students and other observers have been outspoken about their discomfort with AI tools in education, but Foundry participants told reporters they appreciated the new format. For them, the value seems to lie less in the novelty of the avatars and more in the ability to practice repeatedly with targeted feedback.

One participant, Flybridge Capital co-founder Jeff Bussgang, who had his likeness used in the program, said the AI version of him was a bit unsettling, but he added that his students enjoyed it.

That combination of mild discomfort and practical approval may be the best early sign for Harvard. Educational tools do not need to feel perfect to be useful, but they do need to feel credible enough that students keep using them.

What did the AI version of Jeff Bussgang do?

It joined a pitch exercise and responded like an instructor would, but with a digital face. New York Times reporter Sarah Kessler tested the system by pitching an AI-generated version of venture investor and Flybridge Capital co-founder Jeff Bussgang.

According to that report, both the real Bussgang and his AI counterpart were unimpressed by her idea to build “Uber for bananas.” The exercise highlighted two things at once: the avatars can be used in realistic startup role-play, and they can deliver feedback in a way that feels pointed enough to challenge participants.

Kessler also noted that the avatar displayed an oddly fixed smile during the pitch, a reminder that current AI video tools can still produce visible artifacts that make the experience feel artificial. Even so, the larger point of the exercise was clear: the system is designed to create repeated, low-stakes practice for founders before they step into real-world fundraising or board discussions.

Why this experiment matters for business education

Harvard’s use of AI avatars points to a bigger shift in how elite schools may deliver instruction at scale. Business schools have long sold access to expertise, feedback and networks. Generative AI could change how often that feedback is available, and at what price.

For founders, the appeal is straightforward. Pitching is iterative, and improvements often come from repetition. The problem is that human coaches are limited by time. An AI avatar, by contrast, can be available on demand, letting users run through rehearsals multiple times without waiting for the next office hours slot.

For the school, the technology may offer a way to expand access while keeping the human instructors central to the experience. Rather than turning Foundry into a fully automated product, Harvard is positioning AI as a supplement that broadens the number of practice opportunities available between live classes.

That is especially relevant in entrepreneurial training, where students often need help with the basics: structuring an investor pitch, answering tough questions and sounding credible in front of a room. These are the kinds of skills that improve through deliberate practice, which makes them a natural fit for AI-assisted simulation.

What does the $699 price signal?

The price suggests Harvard is testing a more affordable, high-volume model of executive-style education. At $699, Foundry is cheaper than many programs associated with major universities, but still expensive enough to indicate a premium offering rather than a mass-market online course.

That middle ground may be deliberate. Harvard can preserve exclusivity and brand value while lowering the barrier for early-stage entrepreneurs who might not be able to afford a longer, more intensive program. Adding AI avatars may help the school justify the price by increasing the amount of hands-on guidance each participant receives.

It also raises a strategic question for other institutions: if one faculty member can be represented by a digital version that supports practice sessions, can elite training become more scalable without losing its prestige? Harvard’s experiment does not answer that fully, but it does point to a possible path.

How does HeyGen fit into the picture?

HeyGen is providing the synthetic video technology that makes the instructor avatars possible. The startup has built a business around generating realistic talking-head videos, and Harvard’s use case shows how those tools can move beyond marketing and into education.

In this application, HeyGen’s software is not simply producing polished promotional clips. It is powering interactive instructional avatars that appear to react during role-play exercises. That is a meaningful step, because it suggests that synthetic video can be embedded into live learning workflows rather than used only for one-way content delivery.

The partnership also offers HeyGen a high-profile proof point. If one of the world’s best-known business schools is willing to use AI avatars for entrepreneur training, that may help normalize the technology for other universities, accelerators and corporate learning teams.

What are the limits of the technology?

The limits are still visible in the small details. The stiff smile noticed during the pitch session is a reminder that even advanced avatars can look off in ways that disrupt the illusion.

That may not matter much if the goal is utility rather than cinematic realism. Students are unlikely to judge the system on facial microexpressions alone; they care more about whether the avatar helps them rehearse clearly, challenge assumptions and respond better under pressure.

Still, any educational product built on synthetic faces has to balance usefulness with trust. If the avatar feels too uncanny, students may disengage. If it feels too generic, they may not take the feedback seriously. Harvard’s early response suggests the school is still calibrating that balance.

What this says about AI in elite institutions

Harvard’s experiment shows that institutions once seen as cautious are now willing to use AI in very practical ways. The debate around AI in schools has often focused on cheating, plagiarism and the replacement of human work. Foundry points to another direction: using AI to intensify practice and extend coaching.

That does not eliminate the concerns. Questions remain about how closely an AI avatar can mirror a faculty member’s judgment, whether students may over-rely on generated responses and how much authenticity is lost when a digital likeness stands in for a real teacher. But the program shows that some schools are willing to accept those trade-offs if the educational payoff is strong enough.

It also demonstrates a subtle but important framing shift. Instead of asking whether AI belongs in education at all, Harvard appears to be asking where AI can make a human-led course more effective. That is a more pragmatic question, and perhaps a more durable one.

Timeline of the Harvard Foundry AI rollout

The rollout appears to have unfolded in a few clear stages, from concept to trial to student feedback.

What happens next?

The key question is whether Harvard’s model becomes a template or stays a one-off experiment. If Foundry participants continue to value the avatar-based feedback, the school may expand the approach or refine it for other courses and programs.

Other universities, accelerators and corporate training groups will be watching closely. The combination of a recognizable institution, relatively low cost and practical use case makes this a potentially influential pilot for AI in professional education.

For now, the most important takeaway is that the technology is no longer being tested only in labs or consumer apps. It is being placed directly into a real learning environment, where students are using it to rehearse some of the most important moments in startup life: the pitch, the challenge session and the board conversation.

If Harvard can make AI avatars feel useful rather than gimmicky, the idea could spread quickly. If not, the school will at least have shown that the future of business education may include a digital double sitting across from the founder, smiling fixedly while delivering hard-nosed feedback.

Either way, the experiment suggests that the next wave of AI adoption in education may be less about replacing teachers and more about cloning their most useful moments.