Howie Liu says Airtable was approaching $500 million in revenue and still growing when he decided to sell it to Bending Spoons. Now he’s building Hyperagent, and his bet is that the next generation of business software looks like a team of AI coworkers.
I ask Howie why he chose to sell, why he didn’t keep Hyperagent inside Airtable, and what starting over lets him do differently. We get into the economics of AI versus traditional SaaS, selling below Airtable’s peak valuation, and what founders and employees should understand about the equity they hold.
Howie calls the personal AI assistant market “a knife fight in a phone booth.” Hyperagent’s next version is his bet on multiplayer: agents with their own roles and accounts, working with each other and the people inside a business. We discuss why he sees an opening there, what he learned from making the first version too complicated, and how much autonomy these agents should actually have.
We also get into why he’s targeting small businesses that need more people than they can afford to hire, with a landscaping example that takes agents from researching a property to preparing a proposal and managing the job. Howie shares how he uses agents to build Hyperagent and what he’d tell software founders deciding whether to keep building on the business they have or start again.
Sources is available wherever you get podcasts, including YouTube, Spotify, and Apple Podcasts.
Thanks to the show’s sponsors: Mercury, Granola, and Atlassian.
Episode highlights
Howie is betting on teams of AI coworkers. He calls the personal-assistant market “a knife fight in a phone booth,” with Meta, Google, OpenAI, and others all chasing the same opportunity. Coordinating a team of agents is a different product problem, and one he thinks is “a lot more winnable for us.” The business opportunity, as he sees it, is giving small companies the staff they can’t afford to hire, with agents taking on work that helps them win customers and generate more revenue.
Selling Airtable below its peak valuation was a bet that waiting could be worse. In his first podcast discussing the Bending Spoons deal, Howie explains why he sold a still-growing business approaching $500 million in revenue to a buyer with a clear operating playbook. He worried that AI would put more pressure on SaaS valuations, making a better offer down the road far from guaranteed. Because Airtable had preserved most of the more than $1.4 billion it had raised, he says the sale cleared investors’ preferred claims and returned money to common shareholders, including employees; the Hyperagent spinout also gave Airtable shareholders upside in his next company. His warning to employees is that a headline valuation tells you little about what your shares will ultimately pay out. “Common could get wiped,” he warns, if the company can’t cover the investors’ preferred claims first.
His advice to SaaS founders is to question the assumptions that made them successful. At Airtable, he had a growing enterprise business with seat-based pricing and strong cash flow, but came to see AI as requiring a different model. “AI has much worse margins by default,” he says, and Hyperagent also needed a different culture and pace of development. Spinning it out lets him focus on small-business owners, whose feedback allows the team to ship and learn quickly, instead of assuming Airtable’s cash and enterprise distribution would make the new product win. For founders navigating the same transition, he argues that being good at the old game isn’t enough. They need a credible case for growing the company’s value as the rules change, rather than simply keeping it afloat.
ICYMI
Every week, Sources reaches the inboxes of thousands of decision-makers in tech, finance, policy, and media.













