Harvard Business School is selling startup practice at a price many founders can actually reach, but the hard part is still proving an AI professor can sharpen judgment, not just mimic it.
Pitch a weak idea to a Harvard professor and you'll usually know quickly. The eyebrow moves. The question cuts through the slide. You either have an answer or you don't.
Harvard Business School is now trying to package part of that pressure into Foundry, its AI-powered startup programme. TechCrunch reported that the eight-week bootcamp costs $699 and uses AI avatars built with HeyGen so founders can rehearse investor pitches, sales calls, and board meetings against digital versions of real Harvard Business School faculty and startup advisers. That price is the story. Harvard isn't just putting a chatbot beside a syllabus. It's taking a kind of access that used to depend on being in the right classroom, or knowing the right investor, and turning it into repeatable practice.
Look, that's useful. It also feels strange.
New York Times reporter Sarah Kessler tested the system by pitching an AI version of Jeff Bussgang, the Flybridge Capital co-founder and Harvard Business School senior lecturer, on an idea she described as "Uber for bananas." The real Bussgang wasn't sold. Neither was his avatar. It had the same general scepticism, according to accounts of Kessler's test, plus the stiff smile that still gives AI video away. Bussgang called the experience creepy, then gave the line Harvard will care about more: "My students love it." Henry AI raises $16.5 million to become the operating system for commercial real estate deals
Henry AI closed a $16.5 million Series A on July 29, 2026, led by FirstMark Capital, with Thomson Reuters Ventures and Y Combinator also backing the New York startup. The raise coincides with the launch of Henry Deal, expanding the platform from document automation into full deal management for commercial real estate firms. - commercial real estate deal software - CRE operating system for brokers
The reps matter more than the clone #
Foundry's own site makes the pitch plainly: founders can test a pitch with an AI avatar as many times as it takes, then walk into real meetings better prepared. It also offers AI conversations with multiple Harvard experts, and practice modes for pitches, sales calls, and board meetings. You can see the shape of the product at once. The avatar isn't there to be magical. It's there to be available.
That matters if you've ever had to prepare for a funding meeting. Most founders don't need one more vague note about storytelling. They need someone to interrupt them at the weak part of the pitch, ask why the market is big enough, press on customer proof, then do it again after the deck changes. A human mentor can do that once or twice. A professor's digital double can do it at 2 a.m. without pretending to be tired.
Reps are the product.
Harvard says Foundry Bootcamp takes founders from idea to first pitch in eight weeks, online and self-paced. Its current public page is aimed at materials science and deep tech founders, with more industries promised later. There's more to come, too: the site says Foundry connects to a later Catalyst programme, and a pitch event at Harvard Business School where founders compete for $100,000 in funding. Those details aren't decoration. They show Harvard is building a path, not only a novelty demo.
Harvard is testing its own ceiling #
The old Harvard Business School model is powerful because it doesn't scale much. HBS says its MBA Class of 2027 has 943 students, and its 2025 annual report lists 1,878 total MBA students. That's a large business school. But it's still a small pipe next to the number of founders who'd take Harvard-flavoured feedback if the price and access problem went away.
Foundry is the answer to that ceiling. Not a perfect answer. A practical one.
You shouldn't confuse an AI avatar with a real investor, and Harvard shouldn't either. Investor judgment is not just a script of hard questions. It's timing, mood, pattern recognition, and the awkward follow-up that comes when someone hears a claim that doesn't quite hold. AI can imitate some of that. It can't fully replace it. Frankly, the creepy part isn't the big risk here. The bigger risk is founders mistaking a polished simulation for the full pressure of a real room.
Travis Kalanick's industrial AI company Atoms raised $1.7 billion in an a16z-led round announced July 22, with Uber itself joining as an investor and Ben Horowitz taking a board seat. The deal merges Kalanick's CloudKitchens, Pronto mining automation, and transport businesses into a single equity structure, built around a contrarian thesis that... - Travis Kalanick raises 1.7 billion Atoms - specialized industrial robots funding round
Still, the direction is clear. Harvard Business School isn't treating AI as a study aid tucked into the corner of an online course. It's using AI to stretch faculty expertise beyond faculty calendars. Make founders practice five more times before they face an investor, and the programme has done a real job. Make them think they've already faced the market, and the lesson gets expensive fast.
For now, Foundry sits in that uncomfortable middle. Bussgang finds the clone creepy. Students like using it. Harvard is betting that practice wins. Also read: Outcome Based Pricing For AI Agents Runs On Escrow, Not Trust • Google's Gemma Models Just Hit a Billion Downloads and Wall Street Shrugged • Nvidia Warns Hyperscalers Its AI Server Prices Are Jumping More Than 15%