Quarterly business reviews for 40 software accounts: about 80 hours a quarter, and renewals rose without them

For the executive who decides

• The customer acts
• The machine acts
• A person decides
• Nothing runs

An illustration built from the team's own numbers: the days and times are examples, and the account's value is an estimate.

1The job

Preparing and running the quarterly business review: building the deck, holding the meeting with the customer's exec sponsor, and the follow-up. And, since this team stopped, what replaced it: a short written update with a video every two weeks, and an exec meeting only when the customer asks for one. A customer success team at a B2B software company with about 40 accounts, mostly mid-market. The company size and the number of CSMs are not stated. 40 reviews a quarter, 160 a year.

2The number today

Took our CSMs roughly 6-8 hours per QBR including prep, the meeting itself, and follow-up. We were doing this for 40 accounts. Do the math: ~80 hours of CS effort per quarter just on QBR mechanics. His two numbers do not agree. 6 to 8 hours for each of 40 accounts is 240 to 320 hours a quarter, not 80. His own total works out to 2 hours an account. The post does not say which is right, so the low one is used.

At 80 hours a quarter, that is 320 hours a year: about $10,800 a year in wages at US median pay for the closest role ($33.65 an hour for sales representatives of services; there is no customer success category), or about $67 a review. At 6 to 8 hours a review, it is $32,000 to $43,000 a year.

The time was not the main cost. A third of the meetings moved or were cut short, execs sent delegates, and the reviews did not predict who renewed. He stopped them for a quarter. His renewal rate went from 87% to 91%.

The money at stake is the renewals: about $800,000 a year across 40 accounts. The 4-point rise is worth about $8,000 a quarter, about $32,000 a year. It was one quarter, and he has not shown that dropping the reviews caused it.

Estimate. The $800,000 assumes $20,000 an account a year, the low end of a 2026 benchmark for mid-market software deals; at that benchmark's median, $42,000, it is $1.7 million. The $8,000 assumes annual contracts, 10 renewals a quarter. His own account values replace both.

3The verdict

Drop the deck. Hand over the draft update. Keep the conversations.

• Hand to the machine: gathering usage, open tickets and replies for every account every two weeks; spotting what changed against the account's own past; drafting the two-paragraph update with the source behind every number; and keeping count of who has stopped replying.
• Keep with a person: reading and correcting every update before it goes; the video and what to say about risk; the onboarding kickoff, problem reviews, the exec call the customer asks for, and the renewal talk.
• Not yet: sending any update without a CSM reading it; marking an account healthy or at risk on its own; anything about price, renewal terms or release dates for new features.

What the CSM is for now: the conversations that decide the renewal, and the read on which account is drifting. The machine does the stitching. The quarterly deck and the meeting nobody asked for are not handed over. They are dropped.

Part of this job should not exist. A machine that builds the deck faster speeds up the part this team dropped.

4The bill

About $3.50 a month in usage for the 40 accounts, plus 20 euros a month for the software that runs the steps, built from parts; then the build and the upkeep, which depend on where the usage data lives. The other routes: the software made for customer success teams, with AI built in, does not publish prices; the CRM's service plan with a customer success workspace is $180 a month for two CSMs plus $1,500 to start, and it does not write the update; deck builders run $36 to $833 a month and speed up the deck that was dropped; a full-time assistant is $1,600 a month. Checking: a CSM reads every draft before it goes. One team that did this reports 10 minutes a draft, about 14.5 hours a month for 87 drafts, about $490 of CSM time. That is the real running cost. The reviews it replaces cost at least 27 hours a month; writing the update by hand, at the 15 minutes the team quotes, would be about 22. The check time here is not known until it is counted.

Estimate. Usage assumes 87 drafts a month (40 accounts every two weeks) at about $0.04 each; seats assume two CSMs, since the post says "our CSMs" and gives no number.

5Owner and off switch

The head of customer success owns it, because the renewals answer to that role. The off switch: pause the scheduled job in the software that runs the steps. Drafts stop, and CSMs write the two-weekly update by hand, as they do now. It takes under a minute. Try it once in week one.

6The brief to hand your builder

The job: every two weeks, draft a short update for each of our 40 accounts from its usage, open tickets and replies, for the CSM to read, correct and send; and list the accounts where something changed or nobody has replied.

Today: quarterly reviews took 6 to 8 hours each, prep, meeting and follow-up, for 40 accounts. We stopped them. CSMs now write a two-paragraph email and record a short video every two weeks, by hand.

Done when: for two weeks (one round of all 40 accounts), every account due an update has a draft by 9:00 on Monday; every number in 10 drafts chosen at random matches the source system; no update reaches a customer without a CSM approving it; CSM time per update is under 10 minutes, against a timed week before switch-on; and every account with no reply to three updates in a row is on the list.

It may never, on its own: send anything to a customer; say anything about price, renewal terms or release dates for new features; or mark an account healthy or at risk in our records.

You get from us: read access to usage, support tickets and the CRM; the last 10 updates our CSMs wrote by hand, good ones and weak ones; and our short list of changes worth flagging.

We own: the data, the accounts, the email and video accounts, and the billing.

We get back: every Monday, each draft with the source behind every number, the accounts flagged and why, and the accounts gone quiet.

Ask whoever sets it up	A good answer	A bad answer
Where does each number in the draft come from?	Each one shows the report and the date it came from. Missing data shows as a gap, never a guess	"The AI works it out"
What does it do when usage looks fine but the customer's main contact has gone quiet?	It counts replies and meetings as well as logins, and puts an account with three unanswered updates on the list	"Usage is the health score"
Can it send the update without us?	No. Every update waits for a CSM, who sees what the draft says and what they changed	"Yes, fully automatic. That is where the time is saved"

For the operator who builds and runs it Paid layer

7The working set

• Trigger: every Monday at 6:00, for the 20 accounts due that week (the 40 split in two, so each account gets an update every two weeks). A daily check at 7:00 for quiet accounts.
• Steps: pull the last 14 days of usage (active users, seats used, the two or three features that matter) and compare with the previous 14 days and the account's 90-day average; pull open tickets and any open longer than 14 days; pull contact activity (replies to the last three updates, meetings, the main contact's last login); pull the last update sent; draft two paragraphs (what changed, what we see, what happens next) with each number tagged to its source and time; apply the flag rules; put the draft in the CSM's queue; the CSM edits, records the video and sends; log the final text and the send time.
• Data: one record per account per update: account, CSM, ARR, renewal date, period, each metric with its source and pull time, flags raised, draft text, final text, numbers the CSM changed, minutes from draft to send, sent at, reply received (yes or no) and when, exec call requested.
• Review step: the CSM approves every draft before it goes. The head of customer success reads 5 sent updates at random each week against the source data.
• Alerts: no draft for a due account by 9:00 on Monday; a data source that failed to pull (the draft shows the gap); an account with no reply to three updates in a row; an account inside 90 days of renewal with any flag; and the wrong-but-green count, the numbers CSMs had to correct in drafts.
• If it breaks: what you see: drafts missing, numbers that do not match the usage report, or the same text every two weeks. Check first: the data connections are still authorized; the account IDs match between the usage tool and the CRM. Safe fix: pause the schedule; CSMs write the updates by hand, as today.
• Cost per account: about $0.04 a draft in model usage (Claude Sonnet 5.5, about 15,000 tokens in and 1,000 out), about $0.26 an account a quarter, plus n8n Starter (20 euros a month, 2,500 executions; this uses about 120). CSM time: 6.5 drafts a quarter at 10 minutes, about $36 an account a quarter, against $67 today (his quarterly total) or $202 to $269 (his figure for each review).
• Baseline, before switch-on: one round of all 40 accounts by hand: minutes per update (time it, do not ask), replies per update, accounts with a ticket open longer than 14 days, and exec calls requested.

Estimate. Token counts, 10 minutes a draft (one other team's figure) and two CSMs are assumed. The baseline round replaces all three.

Tools by name: the workflow runner n8n (Starter, 20 euros a month billed annually); the model Claude Sonnet 5.5 ($2 per million tokens in, $10 out); the video tool the team already uses, Loom (Business + AI, $24 a user a month, adds auto summaries). Inside the CRM: HubSpot Service Hub Professional ($90 a seat a month annual, $1,500 setup; customer success workspace with health scores; its AI account summary is a handoff beta). Customer success platforms with AI, price on request: Gainsight, Vitally, Planhat. Deck builders: Gamma Pro ($18 a seat a month annual), Rollstack Standard (from $833 a month, billed yearly). People: Stealth Agents (from $1,600 a month, full time). Note: Gainsight's assistant reads only the 50 most recent timeline entries per account (release notes, September 2025).