It may sound crazy to launch an entirely new product line when your core business is reducing the value of paper by two-thirds. But Airtable founder and CEO Howie Liu suggests it’s the smartest move he can make.
The company, valued by investors at $11.7 billion during the zero interest rate frenzy of 2021, is now trading at about $4 billion on the secondary market. But Airtable has raised a total of $1.4 billion, half of which is still in the bank, Liu said, even though the company is “burning cash.” The collapse in valuations affected investor returns and employee stock options, but not the business itself.
Liu’s response is to launch Superagent. He suggests this is an AI agent that could eventually surpass Airtable itself. This is Airtable’s first standalone product in its 13-year history, and it captures both the direction the company is headed and the reality of our current AI moment. All serious software players are competing to prove they can handle agents.
To understand what’s particularly interesting about this movement, consider what Airtable is, a no-code platform that democratizes app building. This is essentially a powerful database that allows anyone to create custom software tailored to their workflow. The company currently has more than 700 employees and serves more than 500,000 organizations, including 80% of the Fortune 100 companies. This is not a struggling startup, but a mature company that has staked its future on a new architecture.
Superagent represents Liu’s bet on “multi-agent coordination.” It’s a system that deploys a coordinating agent that, when you ask a question, has experts working in parallel, rather than one AI assistant fumbling through tasks in turn. “We’re not giving instructions to the AI,” Liu explains. “You’re keeping the team together.”
Here’s how it works: When you contact Superagent about expanding your athleisure brand to Europe (Liu’s example), the system first creates a research plan, identifies what needs to be investigated, and uncovers aspects you never intended to ask. Next, deploy specialized agents in parallel. One examines financials, one analyzes competitive position, and one examines management and news. Finally, integrate everything to create a finished product.
The output is not a wall of text. It’s an interactive market analysis with demographic breakdowns, visually mapped competitive presence, and an expanded timeline to filter and explore. “What if everyone could build New York Times-quality data visualizations for every task they have?” Liu told me on Zoom last week. “Ten years ago, five years ago, this would have been unimaginable. You wouldn’t get that quality of output, you’d just get text. But I think it’s a game-changer that now you can get really, really high-quality, rich, interactive output as the default format.”
tech crunch event
san francisco
|
October 13-15, 2026
What Liu distinguishes between Superagent and its competitors is technology. He cites Anthropic’s AI agents Claude and Manus (a new entrant to the AI research space, acquired by Meta) as the only two products with “true, generally capable, long-running, truly smart agent architectures.” Most other so-called agents, he argues, are simply “LLM-powered workflows,” or predetermined steps mixed with AI calls, rather than truly autonomous agents that can course-correct or backtrack.
It’s a fine line in a market where everyone suddenly launches an AI agent. OpenAI started in 2025 by releasing a new agent-building tool, and since then Notion, Harvey, and hundreds of other companies have added agent capabilities. In a market full of agent claims, Liu’s claim that super agents are different will need to be proven in practice.
In a blog post announcing the product, Liu provides an example of what Superagent can do. Ask them to evaluate Google as a three-year investment opportunity, he writes, and you’ll get a structured valuation that includes citations to its earnings reports, analysis of its defenses against OpenAI and Anthropic, and risk factors it hadn’t considered. Before pitching Wells Fargo, get an overview of Wells Fargo’s AI strategy to learn about their regulatory stance, recent AI investments, and the specific pain points their products address. The system pulls from premium data sources such as FactSet, Crunchbase, SEC filings, and earnings records.
The move caps Airtable’s transformation, which Liu has repositioned as an “AI native platform.” Last fall, the company brought on David Azose, previously the business product engineering lead for ChatGPT at OpenAI, as CTO. At the same time, it acquired DeepSky (formerly Gradient), an AI agent startup that raised $40 million. SuperAgent will operate semi-independently from Airtable and will be directed by Deep Sky’s founding trio.
Pricing was still under consideration as of last week, but it appeared ready to follow the new AI product strategy. That means $20 per user per month for the entry tier and up to $200 for power users, with generous inference credits. “Right now, we’re not looking to optimize profit margins,” Liu says.
It remains to be seen whether superagents will become the multitrillion-dollar market that Liu envisions, or whether the big bet will not pay off. Competition isn’t easy, and the distinction Liu draws between “real agents” and others may not matter to customers if other agents can deliver the right results faster and cheaper.
But for the company’s CEO, who lost $7.7 billion in paper valuation while retaining most of his actual capital, the move signals a willingness to bet on the future rather than protect the present. In fact, Mr. Liu reframed the initial valuation compression as a benefit of the hire, telling his employees that they were getting “a stock that was actually much more attractively priced than the $11 billion valuation,” and a big profit if his bet paid off. He has the capital for strategic acquisitions and does not need to raise new financing.
Asked if he thought super-agents ultimately represented a better chance, Liu shrugged. I’m not denying that possibility. Airtable “will probably be bigger than any of our new products, including Superagent, at least in the near term,” he says. “But I also like being able to bet on super agents. It’s good to have an option.”
This is a version of what Liu calls “wartime” leadership. He admits that while he once avoided the term as being unnecessarily violent, he now accepts it as appropriate. “Being able to adapt quickly and reactively is the most value-creating way to do things right now,” he says. He quickly added: “It’s also the most exciting way to do things.”
