Airtable Just Launched AI Features. We've Been Using AI Agents for Months.
When Airtable announced their AI features, Elena sent a screenshot to our ops channel with the caption "do we still need agents?" It was a fair question. Airtable was adding AI-powered field types, a natural language interface called Omni, and an agent-like feature called Superagent. If the tool we already used for data management was getting built-in AI, maybe the external agents we'd been running were redundant.
We spent two weeks testing Airtable's native AI alongside the agents we already had in production. The answer to Elena's question turned out to be nuanced. Airtable's AI does some things well. It also has clear boundaries. And those boundaries happen to be exactly where external agents start.
What Airtable's AI Actually Does
Airtable's AI features fall into a few buckets, and being specific about each one matters because "AI" is a vague word that means different things depending on who's selling it.
AI fields. You can add a field to any table that uses AI to generate text based on other fields in the same record. Summarize a long description into a one-liner. Categorize a support ticket based on its text. Extract a company name from a messy input. These work at the field level, meaning the AI processes one record at a time using data from that record only.
Elena tested the AI field for categorizing marketing briefs. She had a table of 200 content briefs with descriptions ranging from two sentences to three paragraphs. She created an AI field that read the description and assigned a category: blog, case study, social, email, or video. It got about 85% right. The misses were mostly on briefs that could legitimately go either way, like a case study that was also being adapted as a blog post. Reasonable performance for a built-in feature.
Omni. Airtable's natural language interface that lets you ask questions about your data. "How many deals closed last month?" or "Show me all tasks assigned to Kenji that are overdue." It translates your question into a filter or view. Think of it as a search bar that understands plain English instead of requiring you to build a filter manually.
Superagent. The newest addition. Airtable describes it as an AI agent that can perform actions in your base. Create records, update fields, send notifications based on conditions you describe in natural language. This is the feature that made Elena wonder if external agents were obsolete.
Where Native AI Works Well
I want to give Airtable's AI features a fair assessment because some of them genuinely save time.
The AI field type is useful for any workflow where you need to transform or classify data within a single record. Kenji used it to generate short descriptions from long product specs. His product catalog had 400 items with technical specifications ranging from 100 to 500 words each. He needed a short description (under 50 words) for each one. The AI field handled it. He spot-checked about 30 records and only had to edit five. For a task that would have taken someone a full day manually, that's a clear win.
Omni is helpful for people who don't want to learn Airtable's filter and view system. Our marketing coordinator, Diana, had been asking Kenji to build custom views for her because the filter builder confused her. With Omni, she can type "show me all blog posts published in Q1 that got more than 5,000 views" and get a filtered view. She stopped asking Kenji for help, which freed up about an hour of his week.
For these use cases, native AI is the right tool. It's inside the product, it requires no setup, and it works with the data that's already there.
Where the Boundaries Are
The limitations showed up the moment we tried to do anything that involved more than one record, more than one table, or more than one tool.
Elena's content calendar lives in Airtable. She manages about 60 pieces of content per month across blog, social, email, and paid channels. Her workflow involves checking which pieces are on schedule, which writers have submitted drafts, which items need review, and whether the distribution schedule aligns with the editorial calendar. On Monday mornings, she builds a weekly status report that answers all of those questions.
She tried using Airtable's AI to generate this report. The problem: the information lives across four linked tables (Content Items, Writers, Channels, and Calendar). An AI field can only see data from its own record. It can't cross-reference records across tables, check a writer's submission history, or compare scheduled dates against actual publish dates across the full calendar. The AI field gave her per-record summaries. She needed a cross-table analysis.
Superagent came closer. She described what she wanted in natural language and the agent could create and update records in the base. But it couldn't reach outside the base. When Elena needed the agent to also check whether the WordPress draft matched the Airtable status, or to verify that the social media scheduler had the right publish dates, Superagent couldn't help. It only operates inside Airtable.
That's the fundamental boundary. Airtable's AI is scoped to Airtable. It can read your base, write to your base, and help you work with data inside your base. The moment your workflow touches an external system, you've exceeded what it can do.
What External Agents Do Differently
The agents we'd been running for months operate at a different level. They don't live inside Airtable. They connect to Airtable through the API, which means they can also connect to everything else through APIs. The same agent that reads your Airtable base can check your CRM, query your analytics platform, read a Google Doc, and post to Slack. One workflow, multiple tools.
We replaced Elena's Monday morning report with a content calendar management agent. The agent reads the four linked tables in her Airtable base, cross-references them with the WordPress editorial calendar, checks the social media scheduler, and generates a weekly status report. Pieces that are behind schedule get flagged with specific reasons: "Draft not submitted, writer last active 6 days ago" or "Scheduled for Thursday but no social promotion set up." Elena reviews the report in about ten minutes instead of building it over an hour.
The agent also does something Airtable's AI can't: it acts on the findings. If a content piece is overdue and the writer hasn't been contacted, the agent sends a Slack message to the writer with a reminder. If a blog post is marked as published in WordPress but still shows "In Review" in Airtable, the agent updates the status field. It doesn't just analyze the data. It maintains it.
Honest Comparison
Here's where I've landed after running both for two months.
Use Airtable's native AI when your task is contained within a single base and involves transforming or classifying data at the record level. AI fields for categorization, summarization, and extraction are genuinely good. Omni for querying your data in plain English is convenient. These features make Airtable better at being Airtable.
Use an external agent when your workflow crosses table boundaries in complex ways, touches external systems, or needs to take actions based on analysis. Content calendar management that spans Airtable, WordPress, and social schedulers. Lead enrichment that pulls from external data sources and writes back to multiple tables. Project reporting that reads Airtable and posts to Slack with conditional logic.
The two approaches aren't competing. They cover different ground. Airtable's AI makes individual records smarter. External agents make workflows that span your entire stack automated.
What We Run Today
Our current setup uses both. Inside our Airtable bases, we have AI fields for record-level tasks. Diana's content categorization. Kenji's product description generation. A sentiment classifier on our customer feedback table. These run inside Airtable and they work well.
Outside Airtable, we have agents handling the workflows that touch multiple systems. The content calendar manager for Elena's weekly reporting and cross-platform sync. A lead enrichment agent that watches for new contacts and fills in company data from external sources before the records reach our CRM. A project status agent that reads our project tracking base and generates reports that include data from Google Sheets and Slack.
Tomás, who manages our marketing analytics, made the clearest observation about the split: "Airtable's AI helps me work faster inside Airtable. Agents help me work faster across everything." That distinction has held up.
The Question You Should Actually Ask
When someone asks "should I use Airtable's AI or an external agent?" the real question is about scope. How much of your workflow lives inside a single Airtable base?
If the answer is "most of it," Airtable's native AI features might be all you need. Turn on AI fields, try Omni for querying, and see if Superagent can handle your automation needs. You might be pleasantly surprised.
If the answer is "it starts in Airtable but touches five other tools before it's done," then you need something that can reach beyond the base. That's where agents earn their place. Not by replacing Airtable's AI, but by operating in the space between Airtable and everything else.
Elena no longer wonders if external agents are redundant. She uses Airtable's AI fields every day for quick record-level tasks. She relies on the content calendar agent every week for her cross-platform reporting. Different tools for different layers of the same stack.
Try These Agents
- Airtable Content Calendar Manager -- Manage content schedules across Airtable, WordPress, and social platforms
- Airtable Lead Enrichment -- Enrich new records with company and contact data from external sources
- Airtable Project Status Reporter -- Generate project reports that combine Airtable data with Slack and Sheets
- Airtable to CRM Sync -- Sync records between Airtable and your CRM with intelligent field mapping