Customer success is a portfolio of relationships you can only manage if you see the signals early. An onboarding sequence decides whether an account ever reaches value. A QBR needs a story built from usage data, not a recap of features. A renewal that sneaks up with declining usage is a renewal you are about to lose. And the spreadsheet that would tell you which accounts are slipping sits in an export you keep meaning to actually read. The accounts are the job. The reading and writing around them is what decides whether they renew.
An AI agent fits because the work is part writing, part data, and built on things you already know about your book. It runs real Python on an uploaded usage CSV to tell you who is genuinely at risk, weighted by ARR, instead of you eyeballing rows. It searches the live web for what is happening at a customer's company before a QBR. It drafts onboarding nudges, renewal outreach, and escalation summaries in a warm, concrete voice that leads with the customer's outcome instead of your feature list. The math is exact and the drafts start from your context, not a blank page.
Set your book's facts and the health signals you care about in Memory once, save your health check as a Skill, and put the at-risk analysis on a Monday schedule, and the prompts below cover onboarding, QBRs, renewals, and escalations. You own the relationships. The agent does the spreadsheet and the first drafts.
Capabilities this leans on
Do this first. Every email, QBR, and risk read below comes back in your voice and against the health signals you actually track.
Remember these facts about my work: I'm a customer success manager at Cleartide, a B2B SaaS workflow tool, managing about 40 mid-market accounts worth $8k to $60k ARR each. My job is onboarding, adoption, renewals, and expansion. Renewals are annual and I own the QBR for each account. Our voice with customers is warm, concrete, and proactive: we lead with their outcomes, never with our features. The health signals I care about are weekly active users, feature adoption depth, time-to-first-value, and support ticket trend. I never send a renewal email that reads like a collections notice.
Draft the sequence that gets an account to value, and the nudge when adoption stalls.
Write a 4-email onboarding sequence for a new Cleartide account, the Meridian Logistics team, 25 seats: day 0 welcome with the one thing to set up first, day 3 invite-your-team nudge, day 7 first-workflow walkthrough, day 14 a check-in that ties usage to their goal of cutting manual handoffs. Each under 120 words, warm and concrete, one CTA each.
Their admin set it up but only 6 of 25 seats are active after two weeks. Write a friendly adoption nudge to the admin that names the gap, offers a 20-minute team training, and frames it around the outcome they bought, not our features. Under 90 words.
What you get: A full onboarding sequence and an adoption nudge that move accounts toward value instead of leaving them to drift.
Web search reads the customer's world; the uploaded usage tells you their story.
I have a QBR with Meridian Logistics next week. Search for any recent company news, funding, or leadership changes I should know going in, and give me 3 talking points that connect what's happening at their business to how they use Cleartide.
I'm uploading their last two quarters of usage as a CSV. Summarize the story it tells: adoption trend, most and least used features, and where they're leaving value on the table. Turn it into a 6-slide QBR outline: where they started, what they've achieved, where the gaps are, the plan for next quarter, the expansion opportunity, and the ask.
What you get: A QBR built on the customer's data and their business context, ready to shape into slides instead of starting cold.
Upload the CSV; real Python flags who is at risk, weighted by ARR, not eyeballed.
I'm uploading a CSV with one row per account: account_name, arr, seats, weekly_active_users, features_used, last_login_days_ago, open_tickets, renewal_date. Using real analysis, flag every account at risk, defining at-risk as logins dropping off, fewer than 30% of seats active, or a renewal inside 90 days with declining usage. List them ranked by ARR at risk with the signal that flagged each.
For the 5 highest-ARR at-risk accounts, write me a one-paragraph save plan each: the likely root cause from the data and the specific next action with a timeline.
Chart weekly active users against renewal date for the at-risk group so I can see which renewals are walking in cold.
What you get: A ranked, ARR-weighted at-risk list from real math on your export, plus a concrete save plan for the accounts that matter most.
Get renewal emails that lead with value, even when the account is shaky.
Meridian Logistics renews in 60 days, usage is healthy, and they've maxed their 25 seats with a waitlist internally. Write the renewal-and-expansion email: lead with the outcomes they've gotten this year, propose moving to the 40-seat tier, and offer a quick call. Warm and concrete, never a collections-notice tone. Under 130 words.
Now a different one: an account that renews in 45 days with flat usage and one unresolved support issue. Write a renewal email that addresses the issue head-on, reaffirms the value, and asks for the call, without sounding nervous about the renewal.
What you get: Renewal and expansion emails that read as confident and customer-first, whether the account is thriving or wobbling.
Turn your judgment about account health into a repeatable Skill.
Write a health-check playbook I can run on any account: the signals to pull, the green, yellow, and red thresholds for each, and the recommended play for each color (the outreach, the internal flag, the next step). Make it specific to a B2B SaaS workflow tool.
Save it as a Skill called 'Account health check' so I can run it by pasting one account's numbers and get back a status and the next action.
What you get: A consistent health playbook saved as a Skill, so any account gets a status and a next step in one paste.
Turn a messy support thread into a clear internal summary and a steady customer update.
An account, Brightline Freight, is escalating: two failed syncs this week, a frustrated admin, and a renewal in 30 days. Here's the support thread (attached). Write a clear internal escalation summary for my manager and the product team: what happened, the customer impact, the renewal risk, what's been tried, and what I need from each team by when.
Now write the customer-facing update to the admin: honest about the issue, specific about the fix and timeline, and steady. Under 100 words, no corporate filler.
What you get: An internal escalation summary that gets the right help fast and a customer update that holds the relationship steady.
Scheduled tasks re-run the analysis and surface renewals before they go cold.
Every Monday at 7am, remind me to upload the latest usage export, then re-run the at-risk analysis and send me a digest: new accounts that crossed into at-risk this week, anyone who recovered, and the renewals inside 60 days I haven't touched.
On the 1st of each month, draft me a summary of the accounts renewing next month with a one-line health read and a recommended play for each.
What you get: A weekly at-risk digest and a monthly renewal forecast that land on their own, so nothing slips while you're heads-down.
Customer support
On-brand replies, reusable macros, thread summaries, and a daily triage digest.
SaaS founders
Landing copy, changelogs, lifecycle emails, support macros, and churn analysis.
Sales & outreach
Research prospects, personalize cold email at volume, and turn calls into CRM notes.
Open the Agent, paste any prompt above, and change the details to fit your business.