AI copilot for customer success calls

A check-in where the customer mentions, almost in passing, that something broke last week and nobody told them why. A QBR where they ask what happened to the integration promised two quarters ago. A renewal where the sponsor wants to know what their tier actually includes before committing budget. An escalation where the question is simply when this gets fixed.

None of these is hard in isolation. What makes them hard is being asked without warning, in a call about something else, by someone who will remember whatever you say.

In sales that phrase loses momentum. In customer success it spends trust, and trust is most of the job. The customer has heard it before, often about this exact issue, and each repetition confirms a suspicion that nobody on your side is holding their account in their head.

The information usually exists. The known-issues list has the bug. The escalation matrix has the path. The release notes explain the change that broke their workflow. The problem is retrieval under pressure, mid-sentence, while staying present in the conversation.

Blackbox listens to the call on your machine, recognises what is being asked, and puts the relevant passage on your screen — from runbooks, release notes, the escalation matrix, entitlement terms, whatever you uploaded.

Because answers are constrained to those documents, what appears is what your company has actually written down. That matters on a support call more than most places: an invented SLA commitment or a guessed fix date becomes a promise the customer will hold you to, and that someone else will have to break.

Customer calls are a poor place for a visible notetaker. Enterprise customers often have policies about external participants, and on an escalation call the last thing you want is to spend the first two minutes explaining a guest in the attendee list.

Blackbox does not join. It reads the audio already playing on your machine, and its overlay is excluded from screen capture — which matters on the QBR deck and the shared dashboard that make up most of these meetings.

Support runbooks and troubleshooting guides. The known-issues list, kept current. Release notes, especially the breaking changes. The escalation matrix with severity definitions and response times. SLA and entitlement terms per plan. The onboarding checklist.

The known-issues list earns its place fastest. “Yes, that is a known issue, here is the workaround and the fix is in the next release” is a completely different call from “that sounds odd, let me look into it” — and the first one is available whenever someone has written it down.

It does not know your account history unless you put it in writing. It will not tell you whether a customer is a churn risk, and it has no view of your health scores or ticket queue. It covers the question you could not have predicted, not preparation you skipped.