When a case is about to miss its SLA or has missed it, the customer should hear about the situation from you before they notice themselves. AI watches the case clock and sends the breach notification automatically: what was missed, why, what the new timeline is, what the customer can do. The honest version of this conversation is what AI does better than the manual version it replaces, because the manual version usually skipped the message.

A customer's case had a 48-hour SLA. It is now 47 hours in and the internal owner has not responded. Before AI, the customer would have noticed first and chased. AI sends a proactive message at hour 46: "we promised a response within 48 hours and we're going to miss that; here's what's happening and when we'll be back." The customer is informed before they were ignored.

What people in the field are saying

Service Matters covers the case-management discipline in "Demystifying orchestration: the key...": keeping cases honestly informed is part of the orchestration job, and most teams under-resourced it before AI.

What should the breach message include?

Three things. What was promised and what was missed (the specific SLA). Why the breach happened in plain language (an internal team's review took longer than expected; a required document is still pending). The new timeline, with a specific time, not "soon." Optionally, the next thing the customer can do (provide additional information, wait, contact a different team).

When should the message go out?

Before the breach. The window where the customer would be most upset is the one between when they expected a response and when they realise none is coming. Closing that window with a proactive message is the entire point. Sending the message after the customer has chased is less than half as useful.

Where does this go wrong?

The message goes out without explaining why. The customer feels managed, not informed. The new timeline is vague ("we'll get back to you soon") because no human committed to a specific time. The breach communication happens every time, becoming noise the customer ignores. Each of these is a way to make the discipline worse than not doing it.

What is the practical pattern?

Define the SLA clearly per case type. Define the breach communication: what gets said, by which channel, at what point relative to the deadline. Make the new-timeline field mandatory: the AI cannot send the message without a specific commitment. Audit a sample of breaches monthly to make sure the messages were informative, not formulaic. The AI does the work; the work has to be worth doing.

Related: use case 14: managing a complaint over weeks, when AI should make proactive contact, and how to design escalation.