Backstory Tiered 141 Customers in 3 Days. It Used to Take Five Teams a Quarter.
EntrepreneurshipBackstory re-tiered its entire customer base in three days. The same exercise previously required five teams working a full quarter. The speed is notable, but the part worth understanding is what the AI found that the human teams had been getting wrong.
What Happened
Backstory, a SaaS company in the data and AI space, documented the full process on SaaStr: they built a five-level AI maturity scoring system and ran it across 141 customers in roughly 20 minutes per full analysis cycle, using Claude and an AI workflow tool called Cowork.
The data sources are what made this different from previous attempts. Four connectors — Amplitude, Atlassian/Jira, an internal MCP server, and Slack — replaced the manual data pulls that previously required cross-functional requests to product, BI, and finance teams. The most replicable insight: internal Slack channels containing account team discussions provided signal that had never been captured in structured systems before. That unstructured, “often more accurate” private dialogue got factored into the scoring model.
The output was four customer tiers: Tier A (8 accounts with 10x growth potential across all metrics), Tier B (21 accounts with longer-path 10x potential), Tier C (accounts at strategic junctures requiring intervention), and Tier D (likely to churn without evolution).
Why It Matters
The acceleration demonstrates what changes when AI can ingest and cross-reference multiple data sources simultaneously — not just the speed, but the scope. The previous manual process required five teams because no single team had access to all the relevant signals. Finance had billing data. Product had usage data. Sales had CRM context. Engineering had support ticket patterns. None of them had each other’s data in a usable format.
An AI workflow with four properly configured connectors has all of it at once, and can run the full analysis in 20 minutes rather than a quarter.
The critical finding embedded in the process: “a high volume of feature requests” initially read as a negative signal in the model — customers asking for missing functionality must be unsatisfied. When Backstory tested the assumption against the full dataset, high feature request volume correlated with highest adoption, not lowest. The human teams had been perpetuating a bias. The AI model caught it when given enough data to test the assumption properly.
The So What
The Backstory case is a clear signal that account segmentation and customer tiering are AI-ready workflows for most SaaS companies in 2026. The barrier isn’t the AI — it’s connecting the data sources. If your CRM, product analytics, and support systems are in silos with manual export processes between them, you’re operating the five-team quarterly model by default. The practical question isn’t whether to automate this but which connectors to build first. Start with the data source your team currently argues about the most — that’s the one with the most bias baked in.
Source: SaaStr
Content created with AI assistance and reviewed for accuracy.
Join the conversation
Stack Insiders is our free community for readers who want to go deeper — share resources, ask questions, and connect with others across every vertical we cover.
Join Stack Insiders →