The problem
Pipeline value is usually the optimistic sum of everything open, or a sum weighted by a stage percentage someone typed into the CRM years ago. Neither reflects what actually happened to deals like these.
The list of whom to call is usually the same guess: whoever has gone quiet longest, not whoever is most worth unsticking.
How it works
_predict of won for each open deal, conditioned on what is known before the outcome (stage, blocker, whether there is a champion), with $why for the factors. The weighted pipeline multiplies each deal by that probability. Stalled deals, with no touch in fourteen days, are ranked by it, and their people are found through the contacts link in one query.
With little history the probabilities are weak and bunch together, as they do in the screenshot. That is calibration working: a confident number from a handful of deals would be the defect.
For the full architecture, see the technology overview. For the broader narrative across multiple use cases, read The Predictive Application.
See it live
This use case runs in Company AI, a reference application you run yourself. There is no hosted demo, so the link opens its illustrated tour and source on GitHub: Company AI tour on GitHub →
