Intelligence Learning
Use learned evidence and confidence as reviewable Funnel inputs without making predictions the authority for customer or billing actions.
Intelligence learning studies product events, account facts, billing outcomes, and reviewed attribution to find patterns that may improve Funnel targeting and prioritization.
It is a contextual capability, not a separate product destination. You encounter it while setting up data, investigating an account, reviewing a signal, or configuring intelligent Funnel eligibility.
What learning can do
- Summarize patterns that tend to precede Grow, Save, or Convert outcomes.
- Rank accounts or evidence for review.
- Suggest bounded targeting inputs, timing, cooldown, suppression, or follow-up defaults.
- Explain confidence, freshness, sample size, caveats, and missing instrumentation.
Deterministic facts and explicit product events can directly control Funnel eligibility. Predictions, sentiment, and derived KPIs are initially ranking and explanation inputs. A workspace should let them control eligibility only when validation, confidence thresholds, fallbacks, and reporting are in place.
Confidence tolerance
Use the workspace or Funnel confidence tolerance to choose how much learned uncertainty is acceptable. The interface may express this as Low, Medium, or High tolerance while preserving the underlying score and confidence evidence.
A default keeps setup moving; operators should not have to confirm every suggested mapping or threshold. Low-confidence or stale evidence must fail safely to a deterministic path or no render, never to an unsupported commercial claim.
Human and system authority
Models can prepare reviewable suggestions and bounded defaults. They cannot:
- Publish a Funnel.
- Enable an external destination.
- Decide that a browser user has billing authority.
- Mark a commercial action successful.
- Override verified billing, entitlement, identity, or outcome truth.
Where to review evidence
Accountsshows current intelligence in customer context.Signalsshows scored revenue evidence and delivery history.Eventsshows the underlying product-event stream, mappings, and data quality.Funnelsuses approved intelligence fields in eligibility and path selection.Resultsshows whether targeted journeys produced verified outcomes.