Powered by Airtable Omni · Claude · ChatGPT

An agentic workflow,
delivered.

Most AI tools add a chatbot to something that already doesn't work. We wire AI into your actual operations inside Airtable, with real triggers, real outputs, and a team that actually uses it. Here's exactly how.

See the full process ↓
10 days
From kickoff to a live workflow — fastest with an existing Airtable base
100%
Built inside Airtable — where your team already works
What You're Actually Buying

Your operations, running
without you.

An agentic workflow is an AI that does something on a trigger — not something you have to prompt every time. A lead comes in and it's already researched, scored, and drafted. A report generates itself. A contract gets uploaded and the key terms are already extracted. Your team just sees fewer things to do.

Where most teams are
  • ChatGPT tabs open all day, prompts pasted by hand
  • Data re-entered between tools your team already uses
  • Reports and follow-ups that only happen when someone remembers
  • AI tools that still need a human to actually run them
After a Forma engagement
  • A trigger fires and the workflow runs — no prompt needed
  • Output lands in your Airtable base, exactly where your team works
  • Scheduled tasks run themselves — no one has to remember
  • Your team sees fewer things on their list, not more tools to learn

The most common mistake teams make is buying an AI tool and hoping it connects to their work. We start from your work and build the AI into it. The order matters.

The Stack

The AI layer inside
your Airtable base.

Airtable has shipped more AI capability in the last 3 months than most tools ship in a year. Here's what we're actually using.

Airtable Omni

Build apps, analyze data, and answer questions — in plain language.

Omni is Airtable's conversational AI layer — available on all plans, embedded directly into records, interfaces, and comments. Ask it a question, tell it to build a view, or have it generate a summary across your entire base. No formulas, no exports.

AI Field Agents

AI that runs on every record, automatically.

Field Agents are AI columns that run on a trigger — a new record, a field change, a schedule. Hover any capability to see what it does:

Categorization
Categorization

Reads a record's content and assigns it to a predefined category or tag — automatically, on every new entry.

Summarization
Summarization

Condenses long text, notes, or documents into a short, readable summary for each record.

Data Extraction
Data Extraction

Pulls structured data points (names, dates, amounts) out of unstructured text and puts them in the right fields.

Web Research
Web Research

Searches the internet to enrich each record — company info, news, or contact details — without leaving Airtable.

Document Analysis
Document Analysis

Reads uploaded files — PDFs, Word docs, spreadsheets — and extracts or summarizes their content directly in the record.

Translation
Translation

Translates any text field into a target language automatically as records are created or updated.

Image Generation
Image Generation

Creates an image from a text prompt or field values and attaches it to the record automatically.

Drive / OneDrive
Drive / OneDrive

Connects to Google Drive or Microsoft OneDrive to read and process documents stored there as field agent inputs.

Scheduled Runs
Scheduled Runs

Runs a field agent on a set schedule or when a condition is met — not just when a record is manually changed.

Deep Match
Deep Match

Uses semantic vector search to find and link the most relevant records across tables by meaning — not just matching exact values.

External Models

Claude and ChatGPT, connected natively to your base.

Airtable has an official MCP server — meaning both Claude and ChatGPT can connect directly to your base, read records, run lookups, and write back results. No copy-paste, no exports, no tab switching.

ChatGPT — MCP Native App · Dec 2025

Available directly in ChatGPT. Query records, update fields, and trigger lookups from the chat interface — your base and ChatGPT stay in sync in real time.

Claude — MCP + Automations

Connects via Airtable's MCP server or directly through Airtable automations. Best for long-form drafts, structured reasoning, and complex multi-step tasks where output quality is the priority.

Airtable is shipping fast — and we track every release.
Deep Match · Feb 2026 Web Search GA · Mar 2026 Scheduled Field Agents Drive + OneDrive Connectors ChatGPT Native App

Every Forma engagement includes a capabilities sweep at handoff so your workflow uses the latest tools — not what was available 6 months ago.

The Process

Five phases.
Ten days to a live workflow.

The whole engagement, end to end. Every phase has a deliverable, a clear handoff, and a check-in with you. You always know where things stand.

1
Trigger and Process Mapping
Days 1–2 · Free

We start by mapping the process, not jumping to the tool. What triggers the work? What decisions happen along the way? Where does data get lost or re-entered? Most teams discover two or three things they weren't planning to automate once we walk through it.

You walk away with

A written trigger map showing the exact inputs, decisions, and outputs for your target workflow. Scope confirmed before anything is built.

2
Architecture Proposal
Days 3–4

We translate the trigger map into an Airtable architecture: which Field Agents handle which steps, where Claude or ChatGPT connects via API, what automations fire and when, and what the output looks like in your base. This is a visual spec, not a vague description. You approve it before we touch your data.

You walk away with

A written architecture document with the model choices, automation triggers, field schema, and fixed price. No surprises during the build.

3
Build and Connect
Days 5–7

We build in your existing base, not a sandbox we hand over later. Field Agents get configured, prompts get written and tuned, API connections go live, automations get wired. You have a live link to the base the whole time. We run the workflow against real data and show you actual outputs at the Day 6 check-in.

You walk away with

A working workflow in your live base. Day 6 check-in includes side-by-side before/after comparison of manual vs. AI output so you can see the quality directly.

4
Test and Refine
Days 8–9

Edge cases surface at the end, not the beginning. We run the workflow against your actual data variety — unusual records, missing fields, exceptions your process already handles manually. Prompts get tightened, fallbacks get added, and we verify the output quality holds across your real data range, not just the clean examples.

You walk away with

A refined, tested workflow with documented edge case handling. You know exactly what the system does and doesn't do before you depend on it.

5
Handoff and Go-Live
Day 10 · Final

We walk your team through the live workflow — how it runs, how to spot a bad output, and what to do if something changes in your process. Every engagement includes a written SOP covering what the workflow does and how each piece is configured.

You walk away with

A live workflow, a written SOP, and a handoff session with the people who will actually use it. Your team is unblocked on day one.

6
Ongoing Optimization
Optional · Monthly

Airtable ships updates constantly. A monthly check-in covers whether new capabilities (like Deep Match or Scheduled Agents) should be retrofitted into your workflow, and whether your process has changed enough to warrant a prompt update. Most clients pick this up after 60 days of running live.

Common Questions

Not sure if this is
the right fit for your team?

"We use Zapier for automations already. Why would we switch?"

Zapier moves data between apps. An agentic workflow understands what the data means. If your automation is "when a new form is submitted, add a row and send a Slack message" — Zapier is probably fine. If you want "when a new lead comes in, research the company, score the fit, and draft a personalized outreach email" — that's not a Zap, that's an agent. Different tool for a different job.

"We don't have our data organized enough for this to work."

That's what Phase 1 is for. We've never had a client come in with perfectly organized data. The trigger mapping session reveals what data is actually needed to run the workflow and what's just noise. Sometimes we clean up the base structure first. Sometimes we scope the workflow to what already exists. We tell you which is the right call before anything is built.

"What if the AI gets things wrong?"

It will, sometimes. Phase 4 (Test and Refine) exists to find those edge cases and add handling before you run it on live data. We also document what the system does reliably vs. what it approximates, so your team knows when to spot-check. No workflow we build runs fully unsupervised on day one. That comes after your team has enough data to trust the pattern.

"We're not technical. Can we actually manage this after handoff?"

We build for the person who's going to use it, not for us. The handoff covers the things that break and the things you'd want to adjust yourself. Most of our post-handoff check-in calls are "can we add one more thing?" not "nothing is working."

Forma vs. The Alternatives

The real tradeoffs
before you pick an approach.

Honest about where Forma is the right call and where it isn't.

Forma
Zapier / Make
Custom Dev
DIY with AI tools
Understands your data semantically
Yes, via Deep Match
No, value matching only
If you build it
No
Lives inside Airtable
Natively
Connected via API
Separate system
Copy-paste workflow
Handles unstructured input
Yes (docs, emails, web)
Structured data only
If you build it
Manually
Time to live workflow
10 days
Days, for simple flows
Months
Varies, often stalls
SOP doc + handoff session
Always included
Not included
Sometimes
None
Best when you want
AI that fits your ops
Simple app-to-app pipes
Full custom platform
One-off personal use
Ready to Ship Your First Workflow?

Let's map what you'd
automate first.

We'll walk through your process, show you what Airtable AI can do with it, and tell you whether it's worth building. No pitch, no commitment.

Get in Touch →