It may seem crazy to some, launching a new product line when your flagship business has shed two-thirds of its paper value. But Howie Liu, founder and CEO of Air tablesuggest it’s the best you can do.
The company that investors valued at $11.7 billion in the spirit of zero interest rates in 2021 is now trading on the secondary market for about $4 billion. But Airtable has raised a total of $1.4 billion, and Liu said the company still has half of that in the bank while “throwing cash.” The collapse in value affected investor returns and employee stock options, but did not hurt the business.
Liu’s response was to launch Superagentthe suggested AI agent could eventually destroy Airtable itself. This is Airtable’s first standalone product in its 13-year history, and it captures both where the company is and the reality of AI’s current moment: every serious software player is racing to prove they can deliver agents.
To understand what makes this move so interesting, consider what Airtable is: a no-code platform that democratizes app building. It is a supercharged database that allows anyone to create custom software that fits their workflow. The company currently employs more than 700 people and serves more than 500,000 organizations, including 80% of the Fortune 100. This is not a struggling startup, but a mature business that is following the future in a new architecture.
Superagent represents Liu’s bet on “multi-agent coordination” – a system where you ask and get not one AI assistant fumbling through sequential tasks, but a coordinating agent that deploys specialists working in parallel. “You don’t ask for AI,” Liu explained. “You manage the team.”
Here’s how it works: When you ask Superagent about expanding your sports brand to Europe (an example from Liu), the system first creates a research plan, identifying what needs to be investigated and which dimensions you haven’t considered. Then deploy special agents in parallel – one investigates the financial, the other analyzes the competitive position, management and other news. Finally, synthesize everything into a finished post.
The output is not a wall of text. It’s an interactive market analysis with demographic breakdowns, competitive presence mapped visually, and expansion times you can filter and explore. “What if everyone could create a New York Times-quality data visualization for every task they do,” Liu said on Zoom last week. “This would have been unimaginable ten years ago, or five years ago, where you didn’t get that quality of output – you just got text. But to be able to get interactive output that’s very rich in quality as a standard format, I think it’s a game changer.”
Techcrunch event
San Francisco
|
13-15 October 2026
Liu’s distinction between Superagent and its competitors is technical. He checked the names of the Anthropic AI agents Claude and Manus (a new entrant in the AI research space who bought by Meta) is the only product with “a true, generally capable, long-term and intelligent agent architecture.” Most of the other agents he calls, he says, are just “LLM-powered workflows” — predefined steps with AI calls mixed in, not truly autonomous agents that can correct and backtrack.
It’s a fine line in a market where everyone is suddenly launching AI agents. OpenAI kicked off 2025 by launching a new agent building tool, while Notion, Harvey, and hundreds of other companies began adding agent functionality. In a market with many agent claims, Liu insists that Superagent is different and must prove itself in practice.
In a blog post announcing the product, Liu gave an example of what Superagent can do. Ask to evaluate Google as a three-year investment opportunity, he writes, and you get a structured assessment with quotes for earnings calls, defensive analysis against OpenAI and Anthropic, and risk factors that have not been considered. Ask them to tell you about Wells Fargo’s AI strategy before you make it, and you get regulatory posture, new AI investments, and specific pain points your product addresses. The system pulls from premium data sources like FactSet, Crunchbase, SEC filings, and earnings transcripts.
The move is a transformation for Airtable, which Liu has repositioned as a “native AI platform.” Last fall, the company brought on David Azose, who previously served as engineering lead for ChatGPT’s business product at OpenAI, as CTO. At the same time, it is acquired DeepSky (formerly Gradient), an AI agent startup that raised $40 million. Superagent will operate semi-independently from Airtable, led by DeepSky’s founding trio.
The price is still being worked out as of last week, but it seems poised to follow the emerging AI product playbook: $20 per month per user at the entry level, up to $200 for power users, with generous inference credits. “Right now we’re not trying to maximize profits,” Liu said.
Whether Superagent becomes the trillion-dollar market that Liu envisions or a big gamble that just doesn’t work out remains to be seen. The competition is not trivial, and Liu’s distinction between “real agents” and others may not matter to customers if others can produce satisfactory results faster and cheaper.
But for a CEO whose company has lost $7.7 billion in paper value while retaining most of its real capital, the move shows a willingness to bet on the future rather than protect the present. Indeed, Liu has reframed the previous price compression as a recruiting advantage, telling employees they are getting “equity that is actually more attractively priced than $11 billion worth” with significant value if the bet pays off. They have the capital for strategic acquisitions and don’t need to raise another round.
Asked if he thinks Superagent is ultimately a bigger opportunity, Liu shrugs — he’s undecided. Airtable “is probably going to be bigger for at least the near term than the new products we’re doing, including Superagent,” he said. “But I’m also happy to be in Superagent. The choice is good.”
This is Liu’s version of what he calls “wartime” leadership — a term he admits he once shunned as unnecessarily violent but now finds it appropriate. “Being very quick in the round to be able to adapt,” he said, “is the most creative way to open things now.” He quickly added, “It’s also the most fun way to do things.”