Sentinel, a churn &
revenue radar.

The method lived
in one head.

Boy Hijnen runs Vantum, where he reads recurring-revenue businesses for a living. He is a sales expert with a sharp method for spotting which accounts are about to slip and which are ready to grow. The trouble: that method lived in his head, and the data that proves it sat scattered across clients' CRMs, billing and spreadsheets. By the time an account looked risky, the renewal was already in danger.

He also had no technical background. Graph databases, schemas, scripting: none of it. So the real challenge was not only to capture his method, but to do it in a way he could own and run himself.

Moneyball,
for revenue.

Boy built his method over years: a universe of trackable metrics, grouped into drivers, that point to outcomes. Metrics, drivers, outcomes. It is the Moneyball idea applied to recurring revenue: instead of going on gut, you watch the data points that actually predict whether an account stays, grows, or leaves.

That method was the asset worth keeping. My job was to turn it from something only Boy could run into something a machine could run the same way, every time.

I guided.
He built it.

This was not a co-build. We met week by week and I guided the technical side while Boy was at the keyboard. We ran experiments together, and he went from zero understanding of graph databases to building and owning the whole thing.

We encoded his methodology as a knowledge graph in Neo4j, the methodology store: not just the metrics, but how to interpret them when applied to a real client's data. The scripting was generated with Claude, and the audit workflow runs a client's data against that graph. An AI queries the methodology graph to reason like an experienced sales expert, like Boy. And it runs 100% deterministically: the same inputs give the same read every time, which is what makes it trustworthy enough to act on.

Because Boy built it, he owns it and extends it himself. Every new signal he learns becomes a new metric the graph watches.

flowchart TD
  M["Boy's sales method as a Neo4j knowledge graph: metrics, drivers, outcomes"] --> B["Audit workflow, scripted with Claude"]
  A["Client data in: CRM, billing, sales, surveys (CSV)"] --> B
  B --> C["AI queries the graph to reason like Boy"]
  C --> D["Deterministic read: same inputs, same result, every run"]
  D --> E["Meaningful trends and a monthly to-do list: churn risk and expansion"]

The output is practical, not a dashboard nobody opens. Sentinel hands whoever owns client growth a monthly to-do list: which accounts are slipping and why, and which are ready to buy more. It catches churn early and surfaces expansion at the same time, so the conversation shifts from cutting costs to growing the customers you already have.

His expertise,
now it runs.

The method Boy spent years building now runs on its own. He leads his sales pitches with Sentinel, and the working version already returns real value to anyone who feeds it their data.