Zapier's AI-built workflows are free during early access. That ends on December 1, 2026. Here is what Next Gen Zaps are, how they are billed, and who should wait.
A Classic Zap is a trigger followed by a chain of steps you configure. A Next Gen Zap is different: you describe the work to an AI tool, and it becomes a complete workflow that Zapier runs start to finish. Per Zapier's documentation, you can build one from an MCP client such as Claude, ChatGPT or Cursor, or with the AI assistant inside Zapier's Workflow Manager.
Once published, the workflow runs on its own trigger or schedule with no chat window open. It can also be started on demand from the MCP client, the API or the Workflow Manager, and it keeps a run history. Zapier's launch post lists several capabilities Classic Zaps lack: processing whole lists without the 500-item loop cap, branches that split and rejoin, nested loops, error handling that keeps going, and pausing mid-run to wait for a human reply.
The headline feature is a monitoring agent, called Agentic Management in the docs. It reviews runs that report a problem, diagnoses the failure and prepares a fix. It has three settings: Heal (fix a workflow after a failed run), Optimize (suggest improvements after successful runs) and Automatically apply fixes. Per the docs, all three are on by default.
Some changes always need your approval, including fixes that add a tool call, action or branch, low-confidence fixes, and changes to an asset or authentication. A failed run retries up to five times before the agent steps in, and fixes are tested and published as a new version.
One thing we could not verify: Zapier's promotional post says the workflows get cheaper the more you run them. None of the four help pages we read say anything about repeat AI work costing less. The closest item is that Optimize can turn an AI step that keeps making the same decision into a fixed step that runs faster and more predictably. Treat "cheaper over time" as unconfirmed.
According to Zapier's billing page (updated November 3, 2026), Next Gen Zaps are billed in tasks:
Zapier also says it may change the one-task minimum in the future. Per-run task counts are not yet visible for Next Gen Zaps, and Task History is listed as coming later. That makes cost forecasting harder than with Classic Zaps.
| Area | What the docs say |
|---|---|
| Availability | Early access by request. Not on the Free plan. |
| Sub-Zaps | Not supported. One Next Gen Zap cannot call another. |
| Concurrency | No one-run-at-a-time limit, so overlapping runs can repeat actions such as sending the same email twice. |
| Permissions | No editor role yet. Only owner, super admin or account owner can edit. |
| Importing | Importing a Classic Zap makes a copy. If both stay on, you can get duplicate runs. Zap and global variables are not supported. |
| Enterprise | Several features unsupported in early access, including audit log, workspaces and folders. |
Zapier Professional and Team users: if the list-processing and rejoining-branch features solve a real pain, request early access now while it is free. Test with non-sensitive data, since Zapier itself advises against connecting sensitive accounts. Watch usage before December 1 so the new bill is not a surprise.
Everyone else: nothing forces a change. Classic Zaps keep running as they are. If your workflows are stable, there is little reason to migrate before the product leaves early access.
Comparing alternatives: Make and n8n take a different approach, with visual builders and, for n8n, self-hosting. If per-task billing uncertainty is your concern, our Make vs Zapier comparison covers how the pricing models differ. This is researched from Zapier's public documentation, not hands-on testing, and details may change before December 1.
See how Zapier and Make stack up on pricing and power.
Read Make vs Zapier →