QBR Prep on Autopilot: A Claude Agent Fills the Template from Your CRM, Calls, and Tickets

Fix the template, and put the customer's goals first

The template is the product. Five sections, same order every quarter, and the first one is about the customer: the goals they stated at kickoff (or last QBR) and where each one stands, with the evidence. Then adoption, meaning who is using what, against the seats they bought. Then risks: friction in tickets, sentiment shifts in calls, a champion who changed roles. Then expansion, the use cases they mentioned and have not bought. Then asks, both directions: what you need from them, what they need from you. The QBR template below is the shape; most teams keep all five sections and rewrite the prompts under each.

Get the goals into the CRM if they are not there. A QBR without stated goals is a usage report, and a usage report is what the customer's VP skips. Add a 'Success goals' property to the company or the deal, ask the CSM to fill it at kickoff in the customer's words, and the agent has the spine of every future QBR.

Decide the length now. Two pages of brief, one page of appendix. If the agent writes eight pages, the CSM will not edit it and the customer will not read it. The template should say the word count per section and the agent should hit it.

Connect the sources over MCP, read-only

Connect HubSpot or Salesforce for the account record, contract, goals, renewal date, and open opportunities. Connect Gong or Fireflies so the agent can pull every call with the account's domain this quarter, with transcripts and attendees. Both are read-only for this play; the agent writes a document, never a record.

Support tickets and product usage are the two sources that rarely have a connector you control. Export them: a ticket CSV filtered to the account and the quarter, a usage CSV with active users, feature usage, and logins by week. Drop both into the run's folder. The agent reads files as easily as connectors, and a CSV export is a five-minute job that most support and product tools automate.

Give it the seat count and the contract value from the CRM explicitly. Adoption is a ratio, active users over seats purchased, and the agent should compute it rather than describe it.

• Reads: account, contract (ARR, seats, term, renewal date), success goals, open opportunities, account team
• Reads: all calls with the account this quarter: transcript, attendees, date
• Reads: ticket export (subject, priority, opened, resolved, CSAT if any)
• Reads: usage export (weekly active users, feature usage, logins by user)
• Writes: one brief (Markdown or Doc), one deck outline, one Slack post. Never a CRM record

Write the brief so every claim carries a source

The brief is the template turned into instructions plus the evidence rule: every claim in the brief cites where it came from, a call and date, a ticket number, a usage figure and week, a CRM field. A QBR built on cited facts survives the customer's skeptical CFO; one built on 'you seem to be getting value' does not. Where the data is missing, the agent writes 'unknown' and says what would fill it. An honest 'unknown' in the adoption section is more useful than a confident guess.

Tell it to quote the customer. The most persuasive line in a QBR is the customer's own VP saying, on the January call, 'if we could get onboarding under a week this would pay for itself.' The agent has the transcript; the brief should carry the quote, the speaker, and the date, so the CSM can open the QBR with the customer's words and the progress against them.

Separate what the agent asserts from what the CSM decides. Outcomes, adoption, and the raw risk signals are assembly; the agent does them. The expansion angle and the asks are judgment; the agent drafts them as options and marks them for the CSM to choose. The brief should make that boundary visible with a 'CSM to confirm' tag.

Prepare the QBR brief for {{ACCOUNT}} covering {{QUARTER}}. QBR date: {{QBR_DATE}}.

Read:
- {{CRM}}: the company record, contract (ARR, seats, term, renewal date), the 'Success goals' property, open opportunities, account team.
- {{CALL_RECORDER}}: every call with domain {{ACCOUNT_DOMAIN}} between {{QUARTER_START}} and today. Transcript, attendees, date.
- files/tickets.csv and files/usage.csv in this folder.

Write qbr/{{ACCOUNT}}-{{QUARTER}}.md using the template in templates/qbr.md, five sections in this order and these lengths:
1. OUTCOMES VS GOALS (max 250 words): each stated goal, status (achieved / on track / at risk / no data), the evidence, and a direct customer quote with speaker and date where one exists.
2. ADOPTION (max 150 words + one table): weekly active users vs seats purchased, trend over the quarter, top 3 features used, features purchased and unused, named power users and named silent seats.
3. RISKS (max 200 words): ticket volume and severity trend, any ticket open > 14 days, sentiment shifts in calls, champion or sponsor changes, renewal date and days remaining. Cite ticket numbers and call dates.
4. EXPANSION (max 150 words, tag 'CSM to confirm'): use cases the customer mentioned on calls and has not bought, teams adjacent to current users, the quote that supports each.
5. ASKS (max 100 words, tag 'CSM to confirm'): what we need from them (exec sponsor, case study, reference, data), what they have asked us for and its status.

Then write qbr/{{ACCOUNT}}-{{QUARTER}}-deck.md: a 6-slide outline, one line per slide, customer's goals on slide 1.

Rules:
- Every claim cites a source: (call, date), (ticket #), (usage, week), or (CRM field). No uncited claims.
- Where data is missing write 'unknown' and name what would fill it. Never estimate adoption.
- Quote the customer verbatim where it helps; always attribute.
- Read only. Write nothing to the CRM or the call recorder.
- Post a link to both files in the account's Slack thread #{{ACCOUNT_CHANNEL}} with the three most important lines.

Run it ten days out, per account, on a schedule

Set the job to fire ten days before each QBR date in the CRM, as a scheduled Claude Code job that reads the upcoming QBR dates weekly and runs the brief for each account inside the window. Ten days leaves the CSM time to edit, send the brief to the customer a few days ahead, and adjust when the customer replies with the thing they actually want to talk about, which happens more often once they have a document in hand.

Run the first three by hand, with the CSM who owns the accounts. Read the outcomes section against what they remember; where the agent missed a goal or over-read a call, the fix is a line in the template or a missing field in the CRM. The usual discovery is that the goals were never logged, and the fix is a kickoff habit, not a prompt.

For a book of thirty accounts, the run costs a few dollars a quarter and replaces roughly a day of assembly per account. The CSM's day goes to the twenty minutes of judgment and the meeting itself.

The CSM edits the judgment, then sends the brief ahead

The draft lands in the account's Slack thread with the three most important lines up top: the goal most at risk, the adoption ratio, the days to renewal. The CSM reads the outcomes and adoption sections for accuracy, which takes five minutes because every line has a source to click, and then spends real attention on expansion and asks, the two sections tagged 'CSM to confirm'. Choosing which expansion angle to lead with, and what to ask for, is the job. The assembly around it was never the job; it just took the day.

Send the brief to the customer three days before the QBR. A two-page document the customer reads in advance changes the meeting: they arrive with reactions instead of waiting for slides, the VP stays because page one is her goals, and the forty minutes go to the risks and the expansion conversation. The deck outline exists for the teams that still want slides on the screen; it is six slides and the customer's goals are on the first one.

Keep the brief as the record. Next quarter's run reads this quarter's brief and the CRM, and the outcomes section becomes a trend: goal one achieved in Q2, goal two still at risk, a new goal added in Q3. Three briefs in, the account has a written history that a new CSM can read in ten minutes, which is the part of this play that outlasts any single QBR.

Calibrate the template against what the meeting used

After a quarter of QBRs, ask the CSMs which sections the customers engaged with and which they skipped. Outcomes and risks almost always earn their place. Adoption tables sometimes need to shrink to one line. Expansion is the section that varies most, because it depends on how the agent reads the calls, and the fix is usually sharper instructions about what counts as an expansion signal.

Watch for the agent over-reading sentiment. A customer venting on one call about a bug is a ticket, not a churn risk; the brief should cite the call and the ticket and let the CSM decide the weight. If the risks section reads like an alarm every quarter, tighten the definitions, the way the churn risk brief does with a scorecard.

Every quarter, compare the goals section across the book. The accounts where goals were never logged are the accounts running a usage report instead of a QBR, and that list is the kickoff coaching agenda for the CS team.