Scoring and AI

Work with the right context

Build a scoring policy around explicit fit and engagement signals. Give each contribution a clear reason, a limit and an expiry window, so repeated tracking events cannot overwhelm genuine buyer context. Preserve the policy version that produced the score.

Keep the next action clear

Optional AI suggestions use an approved set of inputs and retain the model version alongside their explanation. Keep their contribution within the agreed boundary. The sales team still records its qualification decision after reviewing the prospect and the evidence.

Coordinate the workflow

Approve scoring model. Marketing Operations submits immutable rule and optional model versions with permitted features, point caps and expiry windows. Admin independently approves the prepared version before its effective time. Protected feature exclusions are enforced before any model request; model suggestions do not make qualification decisions.

Calculate lead priority. Evaluate the approved policy against an exact lead revision. Count each source event once within its version and time window; expiry and reversals recompute the score. Commit a run only if its input revision is current. Optional model output is recorded with version and explanation, bounded by the approved contribution policy. Emit a threshold alert once per qualifying transition.

Review funnel quality. Compare qualification and acknowledged conversion outcomes with the policy version that prioritized each lead. Distinguish missing outcomes from rejected leads. Marketing Operations proposes a new rule version rather than changing historic score explanations; validate it against representative records before approval.

Explainable lead scoring. A scored lead displays the rules and evidence contributing to its rank. Repeated, expired or reversed evidence cannot inflate the score.

AI-assisted prioritization. A model suggestion is traceable to its version and scoring run. Reject disallowed features, unapproved models and automatic SQL decisions.

Explain why a lead moved up the queue

A recent product inquiry may contribute more to priority than an old event registration. The scoring run records the applicable rule version and the inputs used for the current lead. When the team opens the explanation, it should see those contributions and any expired or reversed signals. A higher total without its source context is insufficient for a useful follow-up decision.

AI-assisted priority uses the approved model and its permitted inputs. The model’s contribution is bounded, and its explanation remains separate from the deterministic rules. Changing a model or its feature list requires a reviewed policy version before it can influence new runs. The sales team can therefore distinguish a changed prospect from a changed scoring policy.