A retailer is building a canvas app for store inspections. An agent should comment on the visit only after the required photos are attached. When should the app call the agent?

In a canvas process, the agent call belongs at the step where its required inputs are present. Because the inspection comment depends on the required photos, call the agent only after those photos are attached—not when the inspector merely opens the screen. Calling on every screen open can run the agent on an empty visit and produce a comment before the evidence exists. Replacing the inspection form with a separate chat bypasses the canvas process and skips the retailer’s required photo step. A background timer can also run for a visit that still has no photos, leaving the agent without the input the comment needs. Treat the required photos as the gate: neither opening the screen nor a half-finished form satisfies it. Once generated, the comment is written back to the inspection the employee is already completing. Exam caveat: waiting for the photos does not mean hiding the agent; its comment still returns to the inspection in progress. Practical check: open the screen with no photos and confirm the agent is not called, then attach the required photos and confirm the call happens once.