It might sound crazy to some, launching an entirely new product line while your flagship business has shed two-thirds of its paper value. But Howie Liu, the founder and CEO of Airtable, suggests it’s the sanest thing he could do.
The company that investors valued at $11.7 billion during the zero-interest-rate fervor of 2021 now trades on secondary markets at roughly $4 billion. But Airtable has raised $1.4 billion total, and Liu says the company still has half of that in the bank while “throwing off cash.” The valuation collapse affected investor returns and employee stock options, but didn’t undermine the business itself.
Liu’s response is to launch Superagent, an AI agent he suggests could eventually eclipse Airtable itself. It’s Airtable’s first standalone product in its 13-year history, and captures both where the company is headed and the reality of the current AI moment: every serious software player is racing to prove they can deliver on agents.
To understand what makes this move especially interesting, consider what Airtable is: a no-code platform that democratizes app-building. It’s essentially a supercharged database that lets anyone create custom software tailored to their workflows. The company now employs more than 700 people and serves over 500,000 organizations, including 80% of the Fortune 100. This isn’t a struggling startup but instead a mature business betting its future on a new architecture.
Superagent represents Liu’s bet on “multi-agent coordination”—a system where you ask a question and get not one AI assistant fumbling through sequential tasks, but a coordinating agent that deploys specialists working in parallel. “You’re not prompting an AI,” Liu explains. “You’re orchestrating a team.”
Here’s how it works: When you ask Superagent about expanding your athleisure brand into Europe (an example from Liu), the system first builds a research plan, identifying what needs investigation and surfacing dimensions you didn’t think to ask about. Then it deploys specialized agents in parallel – one investigating financials, another analyzing competitive positioning, another reviewing management and news. Finally, it synthesizes everything into a finished deliverable.
The output isn’t a wall of text. It’s an interactive market analysis with demographic breakdowns, competitive presence mapped visually, and expansion timelines you can filter and explore. “What if every person could have New York Times-quality data visualization built for every task they have,” Liu told me over Zoom last week. “This would have been unfathomable ten years ago, or five years ago, where you don’t get that quality of output – you just get text. But to be able to now get truly extremely high quality, rich interactive outputs as a default format, I think that’s a game changer.”
Techcrunch event Disrupt 2026 Tickets: One-time offer Tickets are live! Save up to $680 while these rates last, and be among the first 500 registrants to get 50% off your +1 pass. TechCrunch Disrupt brings top leaders from Google Cloud, Netflix, Microsoft, Box, a16z, Hugging Face, and more to 250+ sessions designed to fuel growth and sharpen your edge. Connect with hundreds of innovative startups and join curated networking that drives deals, insights, and inspiration. Disrupt 2026 Tickets: One-time offer Tickets are live! Save up to $680 while these rates last, and be among the first 500 registrants to get 50% off your +1 pass. TechCrunch Disrupt brings top leaders from Google Cloud, Netflix, Microsoft, Box, a16z, Hugging Face, and more to 250+ sessions designed to fuel growth and sharpen your edge. Connect with hundreds of innovative startups and join curated networking that drives deals, insights, and inspiration. San Francisco | REGISTER NOW
The distinction Liu draws between Superagent and competitors is technical. He name-checks Anthropic’s AI agent Claude and Manus (a newer entrant in the AI research space that’s being acquired by Meta) as the only two products with “a true, generally capable, long-running and really smart agent architecture.” Most other so-called agents, he argues, are merely “LLM powered workflows” – predetermined steps with AI calls mixed in, not true autonomous agents that can course-correct and backtrack.
It’s a fine line in a market where everyone is suddenly launching AI agents. OpenAI kicked off 2025 by launching new agent-building tools, while Notion, Harvey, and hundreds of other companies have since added agent functionality. In a market steeped with agent claims, Liu’s insistence that Superagent is different will need to prove itself in practice.
... continue reading