Acme · 47% faster onboarding research

The challenge

Acme’s customer success team was the bottleneck on every new enterprise rollout. Each new account required two days of manual research — pulling together prior tickets, contract terms, integration history, and product usage from four separate systems. The team was drowning, and the company was leaving expansion revenue on the table because onboarding could not keep up with sales.

The call

Two paths were on the table: hire a research analyst per region (the safe, expensive option), or build a retrieval system the existing team could query in seconds. We recommended building, on the condition that the team that would use it would help shape it.

The technical bet was that hybrid retrieval — semantic search over unstructured ticket and email data, structured queries over usage and contract metadata — would be enough to answer 80% of the questions a researcher was asking. The remaining 20% would still need a human, and that was fine.

What we shipped

A small RAG platform sitting in front of Acme’s existing data warehouse and ticket store, with a thin web interface designed for the success team’s actual workflow — not a generic chat box. Sources cited inline. A feedback loop that captured which answers were useful so we could improve retrieval over the engagement.

The hand-off included a working session in week seven where the team’s senior researcher made the next set of changes themselves, with us in the room. They have shipped twelve improvements since.

The approach we took here is the one we describe in how to measure AI ROI — define the baseline first, then pick the system that moves it.

The outcome

• 47% reduction in onboarding research time measured across the first 60 days post-launch.
• The CS team absorbed two new enterprise rollouts per quarter without adding headcount.
• The team owns the system. The last commit from us was three months ago.