The professional automation stack

All four tools here have a genuine free tier, so a working configuration costs nothing while you evaluate it. The 03 COSTS table above breaks down each tool's pricing. Three meters climb with use: Make's credit billing, where native AI modules burn credits far faster than standard steps; Airtable's metered AI credits; and Manus's per-task credits. Make's AI steps tend to bite first, so the sticker price is rarely the bill: watch consumption early, and consider calling AI APIs directly for the heavy steps.

Rarely. Make and Airtable are the working core: the scenarios and the structured records they read from and write to. The numbered steps double as the adoption order. Start with Make and the workflow that hurts most, add Airtable once scenarios need somewhere to keep state, bring in Claude when a step needs judgement not a rule, and trial Manus last, once you can tell which jobs are genuinely irregular.

Then one of the four is already in place, and it is the judgement layer: Claude covers the classification, extraction and drafting that need reasoning. What changes is everything around it. This stack puts Claude inside a running system rather than a chat window: Make triggers those reasoning steps automatically, Airtable holds the records they act on, and Manus takes the irregular jobs no scenario covers. You keep the assistant; you add the spine that runs it without you.

Make and Airtable both automate, the one real overlap here. The dividing rule is reach. Airtable owns what happens on the record: AI fields that classify a row, and triggers that fire inside the base. Make owns everything that crosses systems: multi-step scenarios that branch, loop and move data between apps Airtable never touches. Keep field-level logic in Airtable; reach for Make the moment a workflow leaves the base.

When two things happen together. First, the logic outgrows a visual canvas: you want version control, testing and code-level ownership of how each workflow behaves. Second, data-residency or control requirements mean the automation should run on infrastructure you own, not a vendor's cloud. Owning the runtime, not renting it, becomes the goal. Both point to the world-class tier, the owned automation stack.

The strictest member sets the floor. Claude on its personal plans may train on what you type unless you switch it off, so confidential material belongs on its Team, Enterprise or API use, where training is excluded. Manus runs tasks in a cloud-hosted sandbox by default, so briefed data typically leaves your machine, though its desktop app can run tasks locally instead; check its current data-processing terms before anything sensitive goes in. In Airtable, scope each agent's table permissions before it touches live records.