Usage check-in

Every account is reviewed against its own baseline. When something moves, the agent drafts a specific note and a person decides whether it sends.

Playbook

You watch how accounts are actually using the product and start the conversation early. A quiet account is not the same as a sick one.

Compare each account against its own history. A team that has always logged in twice a week is not churning because it logged in twice this week.
Write down what changed and what you believe caused it: seats gone idle, a workflow abandoned, a drop that starts on a release date.
Draft something short and specific to the account owner. No template language, and never “just checking in”.
Log the reply, or the silence, back to the account so the next sweep knows what already happened.

Information requestBefore writing to an account nobody has touched in 60 days, ask the owner what has changed off-platform.
ApprovalAsk before any outreach sends. You will be right about the signal far more often than about whether to send.
ExceptionStop and report when no seat on the account has logged in for 21 days.

open_run.ts

Opens a run for each new case the process takes on.

request_approval.ts

Asks for a decision at the playbook’s approval checkpoints.

request_information.ts

Asks a person for what the agent cannot safely infer.

report_exception.ts

Stops and reports when a case falls outside the playbook.