What Peeve resolves inside your product.
The in-product requests users get stuck on, resolved without a ticket. Peeve shows the step on the live screen or completes the action with confirmation, then hands off to a human only when one is genuinely needed.
Each job below is a request users bring to support again and again. Some resolve at Show: Peeve moves the cursor through your real interface to the control while the user watches, so they learn where it lives. Others resolve at Do: on the Growth plan and above, Peeve completes the backend action, such as a plan change or a refund, after the user confirms a message that states the exact effect.
Every action is written to a signed audit log, and the model never holds your credentials: it requests a capability by name. When a request needs a person, Peeve hands off to your team with the full session. These jobs are the in-product half of AI customer support; see Show and Do for how each rung works.
Which use cases need the Growth plan?
Showing a user the step for any of these use cases is on every Peeve plan, including Starter. Completing the request as a backend action, such as changing the plan, issuing a refund or sending an invite, is a Do feature on the Growth plan ($199 a month) and above.
What happens when Peeve can't resolve a request?
Peeve hands off to a person on your team in Peeve Hand-Offs with the full session attached: the intent in the user's words, every step attempted and what was ruled out. The person who picks it up starts from the investigation instead of asking the user to explain again.
Can Peeve take an action the user didn't approve?
No. Peeve runs only the actions you allow, within the logged-in user's permissions, and it confirms the exact effect before anything changes. Every action is written to a signed audit log, and the model never sees a password or token: it requests a capability by name.