---
title: 'Airtable Omni vs Cobuilder vs Field Agents: Which AI Feature Should You Use When?'
description: 'Confused about Airtable''s AI features? This guide compares Omni, Cobuilder, and field agents — what each does, cost, and when to use each one. Clear decision matrix.'
canonical_url: 'https://www.business-automated.com/tutorials/airtable-omni-vs-cobuilder-vs-field-agents'
md_url: 'https://www.business-automated.com/tutorials/airtable-omni-vs-cobuilder-vs-field-agents.md'
last_updated: 2026-04-21
---

If you've tried to keep up with [Airtable's](/airtable-consultant) AI rollout over the past two years, you've probably heard all of these: **Cobuilder**, **Omni**, **field agents**, **AI field types**, **automation AI actions**. They sound like five different features. In practice they're three — and most of the confusion comes from people mixing up what each one actually does.

This guide draws sharp lines. By the end, you'll know exactly which feature to reach for in any given Airtable AI situation, and you won't have to guess what something costs or whether it's the right fit.

If you need deeper explainers on any of the three, see our [what is Omni guide](/tutorials/what-is-airtable-omni) and our [5 types of AI agents guide](/tutorials/types-of-airtable-ai-agents).

## The One-Sentence Version

- **Cobuilder** is the older starter-schema generator. Its capabilities are mostly absorbed into Omni.
- **Omni** is a conversational AI that builds entire apps (tables, interfaces, automations, field agents) from natural language.
- **Field agents** are AI-powered fields that live inside your tables and automatically analyze, classify, extract, or enrich data at the cell level.

Omni builds the containers. Field agents put intelligence inside the containers. Cobuilder did part of what Omni now does.

## Cobuilder: What It Was and Where It Went

Cobuilder was Airtable's first serious AI builder feature. It focused on one thing: taking a natural-language prompt describing what you wanted to track and generating a starter base template with matching tables and fields.

It was a big deal when it launched because it collapsed the "stare at a blank workspace" problem into a single chat prompt. You typed "I want to track client projects with tasks and milestones," and Cobuilder produced a reasonable schema.

Cobuilder's limits were also clear from day one: it didn't build interfaces, it didn't configure automations, and it couldn't iterate past the initial generation. Once the base was scaffolded, you were on your own.

**Where Cobuilder lives today:** Airtable hasn't formally retired Cobuilder, but its capabilities are now a subset of Omni. New users and builders should default to Omni for everything Cobuilder used to do. If you're still invoking Cobuilder specifically, you're probably leaving productivity on the table.

## Omni: The Conversational Builder

Omni is the successor to Cobuilder and several times more capable. It's a full conversational interface that can:

- Create tables, fields, and relationships
- Build and iterate on interfaces
- Set up automations with triggers and actions
- Configure AI field agents as part of an app
- Analyze data across your records
- Answer questions about your base
- Refine any of the above based on follow-up prompts

In the typical workflow, Omni is where a new project starts. You describe what you want, Omni scaffolds it, you refine, Omni keeps updating. The session feels more like pair-programming with a junior developer than configuring a tool.

**What Omni does well:** first-draft scaffolding, dashboard prototyping, data analysis queries, teaching Airtable concepts to new users.

**What Omni struggles with:** complex business logic, bespoke formulas, legacy data migration, and opinionated design decisions where defaults aren't good enough.

**Cost:** Building with Omni is **free** — no AI credits consumed for creating tables, fields, interfaces, or automations. Running Omni's data analysis against your records **does** cost credits.

Full deep dive in [our Omni guide](/tutorials/what-is-airtable-omni) and [our Omni interfaces tutorial](/tutorials/build-interfaces-with-airtable-omni).

## Field Agents: The Intelligence Inside Records

Field agents are the category people most often confuse with "Airtable AI" in general. They're specifically the AI-powered field types that live inside a table and compute cell values automatically.

A field agent:

- **Reads from source fields** you configure.
- **Runs an AI operation** (classify, extract, enrich, generate, analyze) based on a prompt you provide.
- **Writes the result back into the cell** automatically whenever the source changes.

The five functional types — classification, extraction, enrichment/web-research, generation, and document analysis — are covered in depth in [our types-of-agents guide](/tutorials/types-of-airtable-ai-agents).

**What field agents do well:** per-record intelligence that needs to stay fresh as data changes. Tagging new leads, extracting fields from emails, enriching company records from domains, summarizing meeting transcripts, generating first-draft content for every row.

**What field agents don't do:** build anything. They run inside a table that already exists. If you haven't built the base yet, you're not ready for field agents — use Omni first.

**Cost:** Field agents cost AI credits per run. The exact cost varies by category (document analysis is the most expensive; classification is usually the cheapest) and by the volume of data you're processing.

## Automation AI Actions: The Fourth Thing Nobody Mentions

While Omni, Cobuilder, and field agents are the three names people argue about, there's actually a fourth category of Airtable AI you need to know: **AI actions inside automations**.

An automation AI action is a one-time AI operation that runs as a step in an Airtable automation workflow. For example:

- An automation triggered by "record enters Stage: Proposal" runs an AI step that generates a personalized email body, then a second step that sends the email through Gmail.
- An automation triggered by "new ticket created" runs an AI step that classifies the ticket, then a second step that assigns it to the right team.

Automation AI actions look similar to field agents — they do similar work — but they're structurally different. A field agent lives in a cell and updates continuously. An automation AI action runs once when the automation fires and puts its result into whatever downstream step needs it.

**When to use automation AI actions:** when the AI output is part of a one-time workflow response, not a persistent record attribute.

**When to use a field agent:** when the AI output is a record attribute you want to be able to filter, sort, roll up, or reference from other records.

## The Decision Matrix

Here's how we decide on client engagements:

| What you're trying to do                                 | The right tool                           |
| -------------------------------------------------------- | ---------------------------------------- |
| Build a brand-new Airtable app from an idea              | **Omni**                                 |
| Generate a starter schema from a text description        | **Omni** (formerly Cobuilder)            |
| Create or iterate on an Airtable interface               | **Omni**                                 |
| Set up automations from a natural-language description   | **Omni**                                 |
| Analyze data across an existing base from a question     | **Omni**                                 |
| Automatically classify new records as they come in       | **Field agent** (classification type)    |
| Extract structured data from free text in a field        | **Field agent** (extraction type)        |
| Enrich records with data from the web                    | **Field agent** (enrichment type)        |
| Generate first-draft content for every record in a table | **Field agent** (generation type)        |
| Summarize PDFs, transcripts, or long documents           | **Field agent** (document analysis type) |
| Trigger a one-time AI action when something happens      | **Automation AI action**                 |
| Run AI logic as part of a multi-step workflow            | **Automation AI action**                 |

If the question is "how do I build X?" the answer is usually Omni. If the question is "how do I make this field always reflect AI analysis of other fields?" the answer is a field agent. If the question is "how do I do an AI step inside this specific workflow?" the answer is an automation AI action.

## The Layered Architecture Pattern

Production Airtable systems increasingly combine all three features into a single stack:

1. **Omni builds the base scaffold.** Tables, fields, initial interfaces, foundational automations. This is the fastest path from idea to working app.
2. **Field agents make records intelligent.** New leads get classified automatically. Incoming emails get their key data extracted. Company records get enriched from domain.
3. **Automation AI actions handle one-time workflow steps.** When a deal closes, generate a thank-you email. When a ticket is flagged urgent, draft a Slack message.
4. **Omni handles data analysis on demand.** Once the base is running, come back to Omni and ask "what are the top themes in this month's support tickets?"

None of these layers replace the others. Each does something the others can't. The systems that feel the most "AI-native" are the ones that use all four layers at once.

## Cost Planning Across the Three

AI credits are the currency that ties this all together. Here's the rough shape:

- **Omni building:** free.
- **Omni data analysis (queries across records):** costs credits, usually moderate per query.
- **Field agents:** cost credits per record processed. Volume matters — a field agent on a 100-row table is cheap; the same agent on a 100,000-row table adds up.
- **Automation AI actions:** cost credits per run. Frequency matters — an AI action on a high-volume automation burns credits fast.

Our approach with clients:

1. **Use Omni heavily in the build phase** because it's free.
2. **Be deliberate with field agents** — they're the highest-leverage category, but also the one that can silently blow a credit budget if you apply them indiscriminately.
3. **Monitor credit consumption in the first two weeks** after launching any AI-heavy feature, and adjust scope if the burn rate is higher than expected.

Airtable's billing dashboard shows credit usage broken down by feature, which is the right place to look if you're trying to track down where your credits went.

## Common Confusions, Cleared Up

A few myths we hear constantly:

**"Omni and field agents are the same thing."** No. Omni builds. Field agents run inside what's been built.

**"Cobuilder is deprecated."** Not formally, but Omni has effectively replaced it. Don't start new projects with Cobuilder in mind.

**"AI is too expensive for small businesses."** Building with Omni is free. Field agent costs are manageable if you apply them thoughtfully. The biggest cost risk is running expensive categories (document analysis, generation with images) indiscriminately.

**"Field agents replace automations."** No — they do different things. Field agents keep record attributes fresh. Automations react to events.

**"Omni can replace a consultant."** Omni can replace the grunt work of setting up a new base. The judgment calls — schema design decisions, edge cases, business logic, real-world data migration — are still very much human work. That's where consultants continue to add value in the Omni era.

## The Right Mental Model

Treat the three features as complementary tools in a layered stack, not as competitors:

- **Omni = what you're building.**
- **Field agents = intelligence inside what you've built.**
- **Automation AI actions = intelligence in your workflows.**

Pick the tool that matches the job. Use all three when the job calls for it.

## Need Help Putting It All Together?

At [Business Automated](/airtable-consultant), we ship Airtable systems for clients every week, and AI is now part of every build we do — Omni for scaffolding, field agents for ongoing intelligence, automation AI actions where workflows demand them. We know which feature to reach for when, where the limits are, and how to keep credit costs under control.

If your team is looking at Airtable's AI features and trying to figure out which ones actually matter for your business, [get in touch](/airtable-consultant). We'll help you pick the right stack and ship something useful.


## Sitemap

See the full [sitemap](/sitemap.md) for all pages.