Rebuilding an Account-Tiering Model Used to Take Five Teams a Quarter. One Person Did It in Cowork in Four Days.

SaaStr recapped a session from Haya Kamola, head of customer success at a company called Backstory, on rebuilding an account-tiering framework for 141 accounts using AI — a project that previously required her team plus four others working a full quarter, compressed to three to four days. The team first defined a “golden customer” profile through account-team judgment before touching any data, then built signals that had never been measured repeatably: a five-level AI-maturity score per account generated by a prompt run systematically against CRM fields, public company signals like AI-forward product launches, and full account conversation history pulled from emails, meetings, and Slack. A cross-functional data pull that used to require product, BI, finance, and other teams was replaced with four connectors — Amplitude for usage, Atlassian/Jira for feature requests, Backstory’s own MCP server for conversation history, and Slack for internal account-team dialogue — plus one manual CSV export from Salesforce. The whole analysis ran as an ordered workflow in Claude via Cowork, taking about 20 minutes per full run; it took four rounds of reweighting to go from roughly eight candidate signals down to four scoring buckets (growth potential, AI maturity/velocity, engagement level, account health), and the process caught a real error along the way — high feature-request volume had been scored as a negative engagement signal, when the data showed it actually correlated with the healthiest, highest-adopting accounts.

The result fits a pattern SaaStr’s own operation has been documenting from the inside: a July post from Jason Lemkin describes running SaaStr itself with just 3 human employees and more than 20 AI agents in production, growing revenue 47% year-over-year with over $1 million in agent-attributed revenue — while still arguing traditional B2B software remains necessary for shared logic across agents, a canonical system of record, and compliance auditability. Backstory’s case is a specific instance of the same shift: not agents replacing analysis, but agents collapsing analysis that used to require five teams into a workflow one person can run and rerun.