Use a supported AI by Zapier tool configuration with an actual pause before a sensitive action.
Workflow & automation · Changing · Documentation checked 2026-10-04
Zaps, AI by Zapier, Tables, and connected app actions; standalone Agents is migrating.
Paid plan and model tier affect AI features. Tools and knowledge require supported Advanced / Premium models.
An illustrative recipe. This website does not run the automation.
The goal and the right kind of system
Use a supported AI by Zapier tool configuration with an actual pause before a sensitive action.
An agent fits variable lookup and tool choice. Pin destinations and use the documented approval toggle for writes rather than relying on a request in the prompt.
Inputs and connected services
Inputs: supported model tier, narrow read tools, approved Knowledge, pinned test destination, and reviewer.
Services: AI by Zapier and selected app accounts; optional browsing is a separate enabled capability.
The steps
An illustrative sequence to adapt and test before using real data or enabling ongoing work.
- Inspect Tools and Knowledge and configure only required operations.
- Test retrieval and absent-answer behavior.
- Enable Require approval before running on the sensitive tool.
- Test denied and approved sandbox calls and inspect actual destination state.
Outputs and human checkpoints
Outputs: sourced draft, reviewed tool proposal, and reconciled action result.
Human checkpoints: inspect exact tool parameters before approval. A changed proposal needs fresh review.
Exceptions, permissions, and stopping conditions
Permission boundary: account-scoped tools and pinned destinations. Keep equivalent ungated writes unavailable; approval is off by default unless configured.
Stop on denial, missing reviewer, missing evidence, or uncertain completion. Check destination records before retrying a potential write.
Adaptable prompt
Use approved knowledge and lookup tools to prepare the requested draft. Sensitive actions must pass their configured approval gate. Stop on denial, missing sources, ambiguous identity, or uncertain action outcome. Never use an alternative ungated action.
Related reading
Tools, knowledge, browsing, and Agents migration
Understand when AI by Zapier chooses tools rather than simply transforming an input.
Tool approvals and action parameters
Enable a pause before an agent calls a sensitive app action and verify the inputs it proposes.
Tables, product boundaries, and operating cost
Separate stored business records from conversation memory and account for repeated AI tool calls.
Designing an AI workflow
Make a useful task repeatable with clear inputs, steps, outputs, exceptions, and human checkpoints.
Permissions and meaningful approvals
Decide what a system may read, what it may change, and when someone needs to authorize a concrete action.
Sources and verification
Documentation checked 2026-10-04. Examples are illustrative unless described otherwise.
AI by Zapier tools and approvals · Zapier
Agent knowledge: legacy migration context · Zapier
Key concepts in Zapier Tables · Zapier