Airtable is applying its data-first design philosophy to AI agents with the introduction of SuperAgent on Tuesday. It is a standalone research agent that deploys teams of specialized AI agents working in parallel to complete research tasks.
The technical innovation lies in how the superagent’s orchestrator maintains context. Earlier agent systems used simple model routing where an intermediary filtered information between models. Airtable’s Orchestrator maintains full visibility over the entire execution journey: initial planning, execution steps, and sub-agent results. It creates what co-founder Howie Liu calls "a consistent journey" Where the orchestrator made all the decisions along the way.
"It ultimately depends on how you take advantage of the self-reflective capability of the model," Liu told VentureBeat. Liu co-founded Airtable more than a dozen years ago with a cloud-based relational database.
Airtable built its business on a single premise: software should fit the way people work, not the other way around. That philosophy has driven growth to more than 500,000 organizations, including 80% of the Fortune 100, who use their platform to build custom applications suited to their workflows.
SuperAgent technology is an evolution of capabilities originally developed by DeepSky (formerly known as Gradient), which Airtable acquired in October 2025.
From structured data to free-form agents
Liu designed AirTable and SuperAgent as complementary form factors that together address different enterprise needs. Airtable provides the structured foundation, and SuperAgent handles the unstructured research tasks.
"We obviously started with a data layer. It’s in the name AirTable: it’s a table of data," Liu said.
The platform evolved as scaffolding around that core database with workflow capabilities, automation, and an interface that scales to thousands of users. "I think SuperAgent is a very complementary form factor, which is very unstructured," Liu said. "These agents are, by nature, very independent."
The decision to build free-form capabilities reflects the industry’s learning about the use of increasingly capable models. Liu said that as models have become smarter, the best way to use them is to have fewer restrictions on how they move.
How SuperAgent’s Multi-Agent System Works
When a user submits a query, the orchestrator creates a visual plan that breaks complex research into parallel workstreams. So, for example if you are researching a company for investment, it will break down that task into different parts, like researching the team, researching the funding history, researching the competitive landscape. Each workstream is assigned to a particular agent that executes independently. These agents work in parallel, their work coordinated by the system, each making its own contribution to the whole.
While Airtable describes SuperAgent as a multi-agent system, it relies on a central orchestrator that plans, dispatches, and monitors sub-tasks – a more controlled model than fully autonomous agents.
Airtable’s Orchestrator maintains full visibility over the entire execution journey: initial planning, execution steps, and sub-agent results. This creates what Liu calls "a consistent journey" Where the orchestrator made all the decisions along the way. The sub-agent approach aggregates clean results without polluting the context of the main orchestrator. SuperAgent uses multiple Frontier models including OpenAI, Anthropic, and Google for various sub-tasks.
This solves two problems: it manages the context window by aggregating cleaned results without contamination, and it enables optimization during execution.
"It may have tried to do a research task a certain way, which didn’t work, couldn’t find the right information, and then decided to try something else," Liu said. "It knows it tried the first time and it didn’t work. So it won’t make the same mistake again."
Why does data semantics determine agent performance?
From a builder perspective, Liu argues that agent performance depends more on data structure quality than on model selection or rapid engineering. He based it on Airtable’s experience building internal data analysis tools to find out what works.
Internal tool experimentation showed that data preparation required more effort than agent configuration.
"We found that the hardest part to get right was not actually the agent harness, but more of the massage of data semantics where most of the special sauce was concerned," Liu said. "Agents really benefit from good data semantics."
Data preparation work focuses on three areas: reorganizing data so agents can find the right tables and fields, clarifying what those fields represent, and ensuring agents can reliably use them in queries and analysis.
What enterprises need to know
For organizations evaluating multi-agent systems or building custom implementations, Liu’s experience points to several technical priorities.
Data architecture occurs before agent deployment. Internal experimentation showed that enterprises should expect data preparation to consume more resources than agent configuration. Organizations with unstructured data or poor schema documentation will struggle with agent reliability and accuracy regardless of model sophistication.
Context management is important. Simply linking different LLMs together to create agentic workflows is not enough. There is a need for a proper context orchestrator that can maintain state and information with a view of the entire workflow.
Relational databases matter. Relational database architecture provides cleaner semantics for agent navigation than document stores or unstructured repositories. Organizations standardizing on NoSQL for performance reasons should consider maintaining relational views or schema for agent consumption.
Orchestration requires planning abilities. Just as a relational database has a query planner to optimize results, agentic workflows require an orchestration layer that plans and manages the results.
"So the punchline and the short version is that a lot of it comes down to the agent having a really good planning and execution orchestration layer, and being able to fully leverage the model for what they’re good at," Liu said.
<a href