AI Assists and Reviewable Guidance
Use AI-generated signal explanations and drafts as reviewable support, not autonomous action.
Prevenue can generate guidance for signal review and prediction explanation. AI assists are designed to help a human operator understand evidence and prepare next-step copy.
AI assists do not send campaigns, create CRM tasks, execute lifecycle actions, or replace human review.
Where AI assists appear
AI-assisted guidance can appear on:
- Signal detail pages.
- Prediction and Insights discovery views.
- Mapping or data-quality analysis workflows when enabled.
Availability depends on workspace data, permissions, and whether the related run has produced enough context.
Signal guidance
On a signal detail page, select Analyze or Refresh in the Guidance panel.
The guidance can include:
- Signal explanation.
- Confidence rationale.
- Source summary.
- Caveats.
- Next best action.
- Destination context.
- Slack draft.
- External email draft.
- Sales note.
- Support note.
If signal evidence changed after guidance was generated, the panel can mark the guidance stale. Refresh before using drafts.
Review decisions
Generated guidance can be:
- Saved with edits.
- Marked ready.
- Dismissed.
Use these decisions to keep a review trail. Do not send external messages without checking the evidence and editing the draft for context.
Prediction explanations
Insights can show prediction explanations for propensity scores and discovery runs. These explanations may include why now, operator action, confidence limits, top drivers, evidence references, and caveats.
Use these explanations to understand model or heuristic context. They are not guarantees.
Good review practice
Before using AI guidance:
- Read the signal reason and deterministic recommendation.
- Review the evidence rows.
- Check account intelligence and readiness caveats.
- Confirm the destination is appropriate.
- Edit drafts for tone, accuracy, and customer context.
- Mark ready only when the guidance reflects the evidence.
Data safety
Do not paste secrets, raw transcripts, raw feedback, logs, prompts, full request or response bodies, auth headers, cookies, or customer-sensitive free text into AI review fields.
Prevenue guidance should be grounded in bounded event properties, evidence summaries, account context, and stored signal data.