There are enough prompt lists for customer success managers now that counting them is not the point. Vitally organises 23 prompts by CS motion and does it well. Thinkific runs a longer list. What none of them cover is the part of a CSM's week that is not a motion at all but a seam: the handoff from the account executive who sold the deal, and the moment a renewal conversation turns into an expansion conversation. Those two seams are where accounts quietly go wrong, and they are where a prompt earns its keep. This page covers the standard motions and then spends real time on the seams.
What CSMs use AI well for, and the three things it still gets wrong
AI is good at three CSM jobs: reading a large pile of account history and telling you what changed, turning messy notes into something a customer executive will read, and drafting the difficult message you have been avoiding since Tuesday. It is reliably bad at three others. It will invent product capabilities if you do not tell it what you actually sell. It reads sentiment from text and calls it health, which is not the same thing as usage data. And it will produce a confident renewal risk score from evidence that does not support one, which is worse than no score at all.
The fix in every case is the same and it is boring: paste the evidence, name what you sell, and ask the model to mark anything it is inferring. Every prompt below assumes you have done that.
Onboarding and first-90-days prompts
Kickoff plan: Here is the closed-won summary and the notes from the sales cycle. Draft a 90-day onboarding plan with three milestones, the customer-side owner for each, and the single measurable outcome that would let this customer say the purchase was justified. Flag anything the sales notes promised that we have not confirmed we can deliver.
Stakeholder map: From these emails and call notes, list everyone at the account, their role in the purchase, and who has not appeared since the contract was signed. Rank the disappearances by risk.
Health check and risk-signal prompts
Signal read: Here is 90 days of support tickets, usage summary and email threads for this account. List the five signals that most suggest this account is drifting, quote the evidence for each, and separately list what would have to be true for the account to be fine despite those signals.
Quiet account: This account has been silent for six weeks. Based on the history, give me three plausible explanations ranked by likelihood, and the one message most likely to get a truthful answer rather than a polite one.
The post-sale catalogue lives in Account Management and Customer Growth.
QBR and executive business review prep prompts
QBR build: Here is the account's usage data, support history and the goals they stated at kickoff. Build a QBR that opens with their outcome rather than our feature list, states plainly where we underdelivered, and ends with one decision I am asking the executive sponsor to make. Keep it to seven slides.
Executive translation: Rewrite this QBR summary for someone who has never used the product and has eight minutes. Cut anything that is activity rather than outcome.
Renewal conversation prompts
Renewal case: Using the value delivered this year and what the customer said they wanted at kickoff, write the renewal business case in their language. Include the honest counter-argument their finance team will make and what I should say to it.
Timeline notices: Build the 120, 90, 60 and 30-day renewal touchpoints for this account, each with a different purpose, so the sequence does not read as four reminders.
Renewal micro-tasks are catalogued in Post-Sale and Renewal Micro-Work.
Expansion and upsell prompts
Expansion evidence: From this usage data, identify the two teams at the account showing behaviour that our next tier would serve better. For each, state the evidence, the likely internal champion, and the reason this would be a bad time to raise it.
Land and grow: Draft the message introducing an adjacent team, written so my current champion looks good for making the introduction rather than being asked for a favour.
Escalation and churn-save prompts
Escalation brief: Here is the full thread. Write the internal escalation: what happened, what the customer believes happened, what we owe them, and the single decision I need from leadership today. No hedging language.
Save plan: This account has given notice. Build a save plan with three options at different cost levels, and for each, the honest probability it works and what we lose if it does not.
The handoff seam: AE to CSM, and CSM to support
The AE-to-CSM handoff fails in a predictable way. The AE knows why the deal was bought, and that knowledge lives in their head, their notes and a CRM field nobody filled in. The CSM inherits a contract and a logo. Three months later the customer references a promise the CSM has never heard of.
The prompt that fixes it runs on the AE's material, not the CSM's: Here is the full sales cycle — call notes, emails, the proposal and the mutual action plan. Produce a handoff brief with five sections: what problem they are buying to solve, in their words; the success metric they named; every commitment we made, quoted, with who made it; the stakeholders and their positions; and the three things that would most surprise a CSM who only read the contract.
Run it before the handoff call, not instead of one. The second seam is CSM to support: Summarise this account's open issues for a support engineer who has no history here, ordered by customer impact rather than ticket age, and note which ones the customer has already been told twice about. Cross-team coordination prompts sit in Cross-Functional Collaboration Micro-Tasks.
The renewal to expansion transition
The other seam is a timing question, and getting it wrong costs both conversations. Raising expansion before the renewal is secure makes the renewal feel conditional. Raising it after the paperwork is signed wastes the only moment in the year when the customer is actively assessing value.
The prompt: Given this account's health, the renewal date and what we have delivered, tell me whether the expansion conversation should come before, during or after the renewal, and give the strongest argument for the option you did not pick. That last clause matters more than the recommendation. A CSM who has heard the counter-argument makes a better call than one who has been told what to do.
The context you have to paste in, or none of these work
Every prompt above assumes four things in the window: what you actually sell, the customer's stated goal in their own words, the evidence you are reasoning from, and an instruction to mark inference as inference. Without those, the output is a well-written guess. With them, it is a draft you can edit in four minutes instead of writing in forty. For the manager-side view of the same work, see AI prompts for sales managers; the full category structure is on the taxonomy page. Start with a free prompt, or Browse the library →.
Frequently asked questions
What are the best AI prompts for customer success managers?
The ones organised by account motion rather than by count: onboarding, health checks, QBR prep, renewal, expansion and escalation. The best of them share a structure — you paste the evidence, name what you sell, and ask the model to separate what it read from what it inferred. Prompts that ask for a plan without evidence produce generic output no customer will recognise as being about them.
Can ChatGPT write a QBR?
It can write a strong QBR draft if you give it the usage data, the goals the customer stated at kickoff and the support history. It cannot decide what the customer cares about, and it will pad with activity metrics if you do not tell it to lead with outcomes. Ask it to name where you underdelivered; the QBRs that build trust are the ones that do not read as marketing.
How do I use AI to spot churn risk in an account?
Paste 90 days of tickets, usage and correspondence and ask for the signals that suggest drift, with the evidence quoted for each, then ask separately what would have to be true for the account to be fine. The second question is the useful one. Text sentiment is not health data, so treat any score the model produces as a prompt for your own judgement rather than a verdict.
Is Claude or ChatGPT better for customer success work?
Claude handles long pasted histories well, which suits QBR prep and account reviews where you are reasoning across a year of material. ChatGPT connects to more tools, which matters if you want the data pulled rather than pasted. Most CSMs use whichever their company approved, so write prompts that carry their own context and work in either.