MIT Harnesses AI to Accelerate Startup Ambitions

In an era where artificial intelligence has thoroughly transformed the startup landscape, MIT is using the technology as the foundation for new training programs for entrepreneurs. The school, of course, is known for nurturing startups. A 2015 report found that alumni across the university had started 30,000 then-active companies, generating about $1.9 trillion in yearly sales. The JetPacks (the software to foster the process of initiating a business from an idea to final product) are part of an effort to support and connect the many entrepreneurs across the MIT campus in Cambridge—and around the world—through the Trust Center’s Orbit website.

To that end, the software compliments a pedagogy known in Kendall Square as “disciplined entrepreneurship,” developed by Sloan professor Bill Aulet, the center’s managing director. Disciplined entrepreneurship takes a founder through 24 steps to scale up a business. Nineteen of the 24 steps focus on the customers—who they are, how the entrepreneur can help them, how they would buy the startup’s product, and how to sell to them.

When a founder straps in to the Disciplined Entrepreneurship JetPack by typing a business idea into the prompt, the AI scrapes the web to compose five potential markets for the product—and further subdivide each one if you choose—then analyzes them to propose a priority “beachhead” market to concentrate on. It combs through additional data to suggest market sizes, conversion rates, pricing and more, until the founder comes out with a “minimum viable business product” and a plan to develop it. A second JetPack automates a sequence of 15 “startup tactics” formulated by Cheek that guide founders through executing the plan.

The JetPacks funnel these queries through a platform developed by Stack AI, an MIT startup. “Stack AI keeps each user, and each idea from each user, separate from every other user and data,” says Doug Williams, who oversaw the software’s development for the Trust Center. And it prevents the AI from training on those queries. Currently the JetPacks use OpenAI, but, says Williams, “we’re continually evaluating the other models.”

The Orbit website and the JetPacks are currently available to the MIT community, including students, faculty, graduates, and one-off visitors such as Hotchkis. For everyone else, there’s a waitlist. Cheek says the school is giving priority to educators teaching entrepreneurship around the world.

For now the software is free, but Cheek says MIT plans to commercialize it to some degree, both to maintain the software and to improve it. (Williams says the school has invested $1 million in Orbit and the software.) He and his colleagues have in fact fed the JetPacks into the JetPacks themselves. Besides developing specialized versions, including for businesses focused on climate change and energy, Williams is working on tools to help founders prioritize their time and, as Cheek puts it, “present them with the thought-provoking questions that they need to consider but haven’t yet.” Still, says Cheek: “We’re a nonprofit. We’re not attempting to build a software company to scale this.”