Airtable launches Superagent, its first standalone AI product

What do you do when your company sheds two-thirds of its paper valuation? If you are Airtable founder Howie Liu, you double down and launch a standalone product designed to potentially replace your original business.

Airtable just officially launched Superagent, marking the company’s first standalone product in its 13-year history. While secondary markets value Airtable at roughly $4 billion down from its $11.7 billion peak in 2021, the company still holds around $700 million in cash reserves and continues to throw off profits. Liu says that financial cushion lets him adopt a wartime mindset.

Orchestrating a team instead of prompting a bot

Liu is critical of most current AI agents, arguing that they are merely sequential LLM workflows rather than autonomous problem-solvers. In his view, only Anthropic’s Claude and Manus (currently being acquired by Meta) share Superagent’s vision of a long-running, self-correcting agent architecture.

Superagent is built around multi-agent coordination. Instead of answering prompts one by one, a central orchestrator drafts a research plan and deploys specialized agents in parallel.

For example, if you ask how to expand an apparel brand into Europe, one agent pulls competitive data, another examines regulatory filings, and another checks financials. The deliverable is not a wall of text, but an interactive, New York Times-style data dashboard complete with visual maps, demographic breakdowns, and filterable timelines.

The system pulls directly from premium sources, including FactSet, Crunchbase, and SEC filings. To build it, Airtable recently hired former OpenAI engineering lead David Azose as CTO and acquired AI agent startup DeepSky for its team. Pricing will likely range from $20 to $200 per user each month.

Why it matters

We think Airtable’s sharpest decision here is delivering interactive visual reports by default rather than raw text. Knowledge workers and strategists do not need another chatbot spitting out five paragraphs of generic analysis; they need structured deliverables they can immediately present to executives. If you spend hours every week compiling market briefs or competitor teardowns, Superagent looks well worth a spin as soon as access opens.

This release also signals a fascinating shift in enterprise SaaS survival tactics. Airtable built an empire on no-code databases, but autonomous agents threaten to render manual workflow builders obsolete. Instead of waiting for external startups to eat its lunch, Airtable is trying to cannibalize itself first.

The real risk lies in execution costs and customer perception. Running multi-agent parallel workflows against premium financial feeds burns massive amounts of compute. If competitors offer good-enough sequential research at half the cost, Liu’s purist definition of a true agent might not matter as much to budget-conscious enterprises.