Customer re-engagement automation that knows when not to send
When a user goes quiet for a week, the system researches the account, drafts a relevant message, and decides whether an automated email is appropriate at all.
The situation
When a user has not touched the product in seven days, the usual response is a win-back email that reads as though it was sent to everyone who went quiet that week — because it was.
A useful re-engagement message needs context: which services the account uses, how much it used them before it stopped, whether it is a priority account, and whether someone in Customer Success spoke to the person three days ago. Without that context, an automated email is ignored at best and sent to someone mid-conversation with their account manager at worst.
The constraint
The tempting build is to let a model read the account, write the email, and send it. That works in a demo and creates the wrong architecture. A model that can trigger an action is a model whose bad day reaches a customer.
The agent returns structured fields rather than prose with an action attached: a subject, a body, the right Customer Success owner, and an assessment of whether relevant email has already been exchanged. The workflow reads those fields and applies its own rules. The model contributes the writing and assessment; it never decides what happens next.
Not every inactive user should receive an automated email. A priority account, a user who had been sending significant volume, or someone with an active thread may need a personal response. In those cases, an automated message can signal that nobody is paying attention. The workflow routes the full context to Slack instead, and a person decides whether to reach out.
Person and company research run in parallel: CRM records, meeting history, LinkedIn, prior contact, company data, and external research. Running them in parallel keeps the workflow fast enough for the signal to remain useful.
What we built
Seven days of inactivity creates an event carrying the user’s identity and product usage. The workflow finds the HubSpot contact, retrieves the associated company, lifecycle stage, and priority status, then runs person and company research in parallel.
The agent returns structured output. The workflow then routes it. Standard eligible accounts receive a tailored email from the right Customer Success owner’s inbox rather than a generic address. Higher-touch accounts go to Slack with the context and the assigned owner, so the relationship owner decides what happens next.
Process diagram
Where it stands
Inactive users receive messages that reflect their account rather than their status. Accounts where an automated message would cause harm do not receive one.
Next
Find similar companies from your best customers
Ready to boost business results with AI automation?
Book a 30-minute operations review. We will look at where manual work is limiting your operation and whether there is a practical opportunity to create capacity.
Book a 30-minute operations reviewSwiss precision. Valley speed.