If you've been watching Airtable over the past year, you've seen the platform make a deliberate, very public pivot toward AI. The flagship piece of that pivot is Omni — Airtable's conversational AI builder, and the feature CEO Howie Liu positioned as the cornerstone of "the AI-native Airtable" in his announcement letter.
For new users, Omni is the on-ramp: describe what you want, and a working app appears. For experienced builders and consultants, it's a productivity multiplier that collapses the first day of a client engagement into the first hour.
This guide explains what Omni is, what it actually builds, what's free versus what costs AI credits, and where the rough edges still are. If you're already sold and want the hands-on tutorial, jump to how to build custom interfaces with Airtable Omni.
The Short Version
Omni is a chat interface inside Airtable that can:
- Create new bases, tables, fields, and relationships from natural language
- Build and edit Airtable interfaces (dashboards, record detail layouts, charts, filters)
- Configure automations with triggers and actions
- Set up AI field agents for classification, extraction, and enrichment
- Iteratively refine any of the above based on follow-up prompts
- Analyze data across your base to find patterns, generate reports, and answer questions
You type. Omni builds. You refine. The whole thing runs inside Airtable with no external tools and no code.
See Airtable's own explanation on the Omni / AI app building page or the Using Omni AI in Airtable support doc for the official reference.
How Omni Differs from Earlier Airtable AI Features
Airtable has shipped multiple AI features over the past two years, and it's easy to confuse them. Here's the quick map:
- Cobuilder — the earlier AI feature that generated starter templates and base schemas from a prompt. Most of Cobuilder's capabilities have been absorbed into Omni.
- Field agents — AI-powered fields that automatically read, classify, extract, or generate data at the cell level. These are workflow features, not builder features. See our 5 types of Airtable AI agents guide for the deep dive.
- Automation AI actions — individual AI steps inside an automation workflow (e.g. "generate an email," "classify this record"). These run once when the automation fires.
- Omni — the conversational builder. It's the top of the stack and can create field agents, automations, and interfaces as part of its output.
If you want a structured comparison of Omni, Cobuilder, and field agents — including when to use each — see Omni vs Cobuilder vs Field Agents.
What Omni Actually Builds: A Walkthrough
Let's walk through what happens when you start a typical Omni session.
The prompt: "I need a simple CRM for a consulting agency. Contacts, companies, deals with a pipeline, activity log, and a dashboard showing deal value by stage and the top 10 contacts by most recent activity."
Step 1 — Schema generation. Omni creates four tables: Contacts, Companies, Deals, and Activities. It sets up linked record fields between them so a Deal has a Company and a list of Contacts, and Activities are linked to both Contacts and Deals. Primary fields are named sensibly. Default views appear.
Step 2 — Field configuration. It adds realistic fields to each table: email, phone, title, tags for Contacts; website, industry, size for Companies; stage (single select), value (currency), close date, probability for Deals; type, date, notes for Activities. Single-select fields come prefilled with sensible options (Pipeline stages: Lead → Qualified → Proposal → Negotiation → Closed Won / Closed Lost).
Step 3 — Interface design. Omni builds a dashboard interface with a deal value chart by stage, a counter of open deals, a top-10 list of recent activity, and a filtered list of contacts. You can ask it to rearrange, recolor, or swap components.
Step 4 — Automations. Based on the prompt, Omni may offer to create an automation ("Notify assigned owner when a deal hits Proposal stage") and set it up pending your confirmation.
Step 5 — Iterative refinement. You say "Add a tasks table linked to deals with due dates and assignees." Omni adds it. You say "Show overdue tasks on the dashboard in red." Omni updates the interface. Every prompt builds on the last.
This whole session takes five to fifteen minutes. The resulting base isn't production-ready — you'll want to review naming, add validation, connect it to your real data sources — but it's a dramatically better starting point than a blank workspace.
What's Free and What Costs Credits
Airtable has been explicit about the economics here, and the rule is simple:
- Building apps with Omni is free. Creating tables, fields, interfaces, and automations through chat does not consume AI credits. You can iterate as much as you want during the build process without touching your credit allowance.
- Running AI against data costs credits. When Omni executes AI operations on your records — classifying them, extracting fields, analyzing patterns, summarizing content — those runs draw from your workspace's monthly AI credit allowance.
The dividing line is useful in practice: the building work is free because Airtable wants you to build more apps. The per-record AI work is metered because it has a real compute cost and scales with your data volume.
Field agents (see our types-of-agents guide) are on the "costs credits" side of this line. Building them with Omni is free; running them across thousands of records is what eats the budget.
Five Real Use Cases Where Omni Shines
From client engagements, these are the places Omni delivers disproportionate value:
1. First-draft scaffolding of a new base. Instead of starting with a blank base or hunting through the template gallery, describe the system you want and let Omni build the skeleton in minutes. The result is usually 70% of the way to what you want, which makes the next hour of refinement dramatically more productive.
2. Rapid interface prototyping for client demos. Before you invest a day building a full dashboard, use Omni to generate two or three interface concepts from different prompts. Show the client, get their reaction, then build the winner properly.
3. Automation setup for non-technical owners. Describe the workflow ("when a new support ticket is marked urgent, send the assigned agent a Slack message and create a follow-up task"), and Omni wires up the automation. Great for handing off the "last mile" of a build to a non-technical client admin.
4. Data analysis over large record sets. Ask Omni "What are the top three reasons deals are being lost this quarter?" and it'll analyze the closed-lost records, categorize them, and return a summary. This is the feature that starts to feel more like a business analyst than a builder.
5. Teaching moment for new team members. Have a new team member describe what they want in English, watch Omni build it, then explain what Omni actually did. It's the fastest way to teach Airtable fundamentals because every step maps to a specific Airtable concept.
Five Places Omni Still Struggles
Equally important: where Omni will disappoint if you expect too much from it.
1. Complex relational schemas. Anything with many-to-many junction tables, multi-level rollups, or unusual cardinality tends to confuse Omni. It'll build something, but it's often not what a human schema designer would choose. Plan to review and refactor.
2. Specific formula logic. Omni can write formulas, but it doesn't always get the business logic right. Always review formula field contents it generates — particularly anything involving dates, currencies, or multi-field conditions.
3. Custom integrations. Omni can configure built-in automation triggers and actions, but it doesn't know about your specific Make scenarios, Zapier zaps, or external APIs. You'll still design integrations manually.
4. Legacy data migration. Pointing Omni at a pile of CSVs and asking it to "figure out the right schema" works poorly. Start from the schema you want, then import data into that structure.
5. Opinionated design decisions. Omni optimizes for "build something that works" rather than "build the right thing for your business." If you have strong opinions about naming, view configurations, or interface design, Omni will fight you a little. That's where human consultants still add value.
Omni and Field Agents: The Stack
The most powerful Airtable systems we've built recently combine Omni and field agents into a layered stack:
- Omni builds the base scaffold — tables, fields, interfaces, initial automations.
- Field agents handle per-record intelligence — classification of new leads, extraction from incoming emails, enrichment from public data.
- Omni analyzes the aggregate — running queries across the enriched data to answer business questions.
This layering is new and still evolving. Expect the story to keep getting better over the next year as Airtable continues to expand Omni's capabilities and field agent model selection.
Should You Use Omni Today?
If you're new to Airtable: yes, unambiguously. Omni is the fastest way to get from "I have an idea for a system" to "I have a working app." Start with Omni, then learn the manual controls as you refine.
If you're an experienced Airtable builder or consultant: also yes, but with clear expectations. Omni is a productivity multiplier for the scaffolding phase and a time-saver for dashboard prototyping. It's not a replacement for schema design, validation logic, or the client-specific judgment that separates a working base from a great one.
If you're worried about AI-generated quality in production systems: your instinct is correct. Use Omni to accelerate, not to abdicate. Every Omni-built base we ship to a client goes through a human review pass before it's called done.
Get Omni Into Production, Properly
At Business Automated, we build Airtable systems for clients every week, and Omni is now part of that workflow. We use it where it shines — rapid scaffolding, interface prototyping, first-draft automations — and we layer in the judgment and specific business logic that Omni alone doesn't have.
If your team is curious about Omni but uncertain about how to evaluate its output, or you want help shipping an Omni-assisted base into real production use, get in touch. We'll walk you through what Omni does well for your specific use case and where the human layer still earns its keep.
