Role, Customer Success

Claude for
customer success

QBR prep, health check analysis, ticket summarisation, feedback themes, and the writing that sits between account data and action.

Customer success is one of the fastest-paced roles I've seen, and one where the admin-to-customer ratio can creep up until you're doing more writing about customers than talking to them. Claude's role here is simple: take the admin off your plate so the conversations get more of your attention. Nothing on this page is about replacing the customer-facing work. It's all about clearing the path to it.

Three places to start

Three workflows where the time saved goes directly back into time with customers.

1. QBR prep from account data

Paste the account's usage and engagement data since the last QBR, paste your previous QBR decks as a format reference, and ask Claude for a slide-by-slide outline. The good version leads with the biggest shift (good or bad) rather than a boilerplate intro, and ties every slide to something you want the customer to agree or commit to. The QBR becomes a meeting, not a readout.

QBR prep prompt, annotated

The instruction that turns a QBR deck from a readout into an agenda.

YouI have a QBR with Acme next Tuesday. Here is their usage and engagement data since the last QBR, pasted below. 1 Here are our last three QBR decks for other accounts, as a format reference. 2 Draft a slide-by-slide outline for this QBR. For each slide: a clear headline, the data or story supporting it, and crucially what I want the customer to agree or commit to on that slide. Lead with the biggest shift (good or bad), not a boilerplate account summary. 3 If you see a red flag in the data I should raise explicitly in the meeting, surface it in its own slide rather than softening it. Do not invent usage figures. 4
1
Context. The account, the meeting, the data source.
2
Example. Your previous QBR decks as a format reference.
3
The ask. Specific output shape plus "what do I want them to commit to", the line that makes the deck drive the meeting.
4
Escape hatch. Surface red flags rather than softening them. Prevents the deck from accidentally smoothing over a problem.

Where this falls over

If you don't ask for "what do I want them to commit to" alongside each slide, Claude will produce something that sounds thorough but doesn't drive the meeting anywhere. That instruction is the whole difference.

2. Health check analysis across a portfolio

You look after twenty accounts. Claude will score each one green, amber, red with a one-line reason, and then give you the five that most need attention this week, with a specific action for each. "Book a check-in call" is not a specific action. Ask Claude to avoid that default unless it's genuinely the right move.

Portfolio health check, weak vs useful

The difference between "book a call with everyone amber" and a real prioritised action list.

Youwhich of my accounts need help this week
What Claude sends back

A list of all 20 accounts with a rough colour, no specific actions, and a suggestion to "book a check-in call with the amber accounts". You close the tab and do what you were going to do anyway.

YouHere is usage and engagement data for my 20 accounts, pasted below. Score each account green, amber, or red with a one-line reason. Then give me the five accounts that most need attention this week, with a specific action for each. "Book a check-in call" is not a specific action and should only appear if that genuinely is the right next step for that account. Prefer actions like "flag the drop in their reporting feature usage", "share the X feature they haven't discovered yet", "escalate the integration blocker to their engineering lead". If an account is red for a reason you cannot see in the data but is implied by a sudden drop, flag that so I can investigate rather than recommend a generic action.
What Claude sends back

All 20 accounts scored. Five flagged as "attention this week" with specific, different actions:

  1. Beacon Labs (red). Daily active user count dropped from 24 to 8 this week. Specific action: ask their admin whether there was a team reorganisation, because the drop started on a single day.
  2. Marlin Industries (amber). Using the tool heavily but only through the API, not the dashboard. Specific action: share the March dashboard release with their integration lead, it adds the one view they were building in Metabase.
  3. Northward Group (amber). Support tickets this month are concentrated on one feature. Specific action: loop in product about the feature's usability, not a CS check-in.
  4. Owl Systems (red). Renewal in 6 weeks, usage flat. Specific action: set up a goal-reset meeting, not a general QBR.
  5. Penzance Data (amber). New admin this month, no onboarding completed. Specific action: send the new-admin onboarding path and offer a 20-minute walkthrough.
One thing I cannot see in the data: Beacon Labs' drop is sharp and sudden. Looks like something structural (reorg, acquisition, budget change). Worth asking directly rather than assuming.

Where this falls over

Without the explicit instruction to avoid "book a call" as a default, half the recommended actions will be that. Include it.

3. Themes from a month of customer feedback

Tickets, NPS comments, Slack messages, casual feedback in meetings. A month's worth is a lot. Claude will pull out themes, count how often each appeared, quote directly, and crucially, flag anything that's new this month. The newness is the signal you care about most, because it's the thing you didn't know about yet.

SW
Sam W. #cs-team Friday, 16:40

March feedback digest done. Pasted every ticket, NPS comment, and relevant Slack thread from the last 30 days into Claude. Asked for themes with counts and quotes, and specifically "flag anything that is new this month and did not appear in February's digest".

Three new themes surfaced. One of them is about a workflow change we shipped on the 12th that I had not connected to the increase in "can you help me find..." tickets. That connection alone saved an hour in Monday's product sync.

6 heart 2 lightbulb

Where this falls over

Claude will merge similar-sounding concerns into one theme if you let it. Always tell it not to, because the merged version loses information.

Worked examples

Before and after, on tasks from customer success work.

Prompts to steal

Copy any of these, fill in the square brackets, run.

Try this week

Pick one. All of these are under an hour.

The best CS teams use Claude as a way to spend less time on the deck and more time on the conversation. If you find yourself spending the saved time on more slides, you're using it wrong.

If you want to go broader, the Beginner guide covers the habits that apply everywhere. The prompt library has the full set of role-tagged prompts in one place.