← Back to Support

AI Capabilities Overview

How AI Processes Signals

Every signal that enters MindBacklog is automatically analyzed by AI before appearing in your Intelligence Hub. The AI handles the heavy lifting so you can focus on decisions, not data entry. Once enough signals on a related theme accumulate, the nightly clustering worker drafts them into actionable Insights — distilled briefs ready for triage.

At a high level, the AI performs several key tasks on each incoming signal:

  • Understands the content — Reads and summarizes the signal, identifying what the user is saying and how they feel about it.
  • Filters out noise — Automatically separates relevant feedback from spam and off-topic content using your Product Summary as context.
  • Classifies the signal — Categorizes it as a feature request, bug report, praise, complaint, question, or general feedback.
  • Connects the dots — Links the signal to existing features in your backlog, building quantitative evidence for prioritization.
💡
Compounding Intelligence

The more signals your AI processes, the better it understands your product, your users' language, and recurring patterns. Classification and signal-to-feature linking become increasingly accurate over time.

Signal Classification

Six categories: Feature Request, Bug Report, Praise, Complaint, Question, and General Feedback. Classification accuracy improves as your Product Summary and signal history grow.

AI Priority Score

Calculated per feature based on multiple weighted factors:

  • Signal Count — Number of linked customer signals (quantitative demand)
  • Strategic Alignment — How well the feature maps to active OKRs
  • Feature Completeness — How well-defined the feature is (description, PRD, stories)
  • Signal Quality — Sentiment and diversity of linked feedback

Scores are normalized 0–100. Higher = stronger evidence to prioritize.

Content Generation

FeatureDescriptionCredits
Generate DescriptionAI writes a feature description from the title + product context1
Generate PRDFull Product Requirements Document with sections for goals, user stories, requirements, constraints10
Generate StoriesBreaks feature into user stories with acceptance criteria and story points2
Estimate EffortAI suggests T-shirt size (XS–XL) with reasoning1
Product SummarySynthesizes context sources into a master profileIncluded

Three-tier search across the platform:

  1. Exact match — Keyword matching on titles, keys, descriptions
  2. Broadened match — Expanded keyword variations
  3. Semantic search — AI-powered meaning-based search that understands intent, not just keywords

Available across the Intelligence Hub's Raw Feed (signals), the Roadmap (features), Ask Mind (chat queries), and feature linking.

A nightly worker reads every recent signal's embedding and runs HDBSCAN clustering to group semantically related items into persistent themes. Each new cluster is titled by an LLM; existing themes get matched and updated cheaply (no re-titling unless the signal mix changes meaningfully). The result powers two surfaces:

  • Intelligence Hub → Trending tab — full theme list, sortable by Volume (30-day signal count) or Growth (week-over-week velocity), with click-through drill-down and one-click promotion to a Radar feature.
  • Dashboard → Trending Themes card — top 5 themes with NEW / ↑x% / ↓x% / 🔥 spiking badges.

Themes dormant for 60 days are auto-archived. The cost stays steady because re-titling only fires when the cluster's signal makeup drifts (Jaccard similarity threshold).

AI Daily Brief

Auto-generated summary delivered on the Dashboard each day. Covers signal trends, progress updates, key items needing attention, and strategic highlights. Free on all plans, no credits consumed.

Complete Credit Cost Reference

ActionStarter CostGrowth/Scale
Signal Processing (per signal)1 creditUnlimited
Generate Description1 creditUnlimited
Generate PRD10 creditsUnlimited
Generate Stories2 creditsUnlimited
Estimate Effort1 creditUnlimited
AI Priority Score1 creditUnlimited
Ask Mind Message1 creditUnlimited
Daily BriefFreeFree