Skip to content
Sparqbox Docs

What the AI first reviewer does

The AI first reviewer is one reviewer on every idea, alongside the human coordinators. It scores ideas before humans see them. It never decides outcomes.

What the AI does

Section titled “What the AI does”

For every new idea, the AI receives:

  • The idea's title, description, and problem.
  • The expected impact field if filled.
  • The category name.
  • The list of criteria for that category, with their names, questions, and weights.

It returns:

  • One score, 1 to 5, per criterion.
  • An optional comment per criterion.
  • A short summary of two to three sentences explaining the overall assessment.

The scores feed into the same weighted-score formula human reviewers use. The summary is shown on the idea's detail page. The AI writes in plain language: short sentences, everyday words, no filler. The warmth of the tone follows the Response style setting in Settings → AI (default: casual).

What the AI never does

Section titled “What the AI never does”

The AI is a recommender, not a decision-maker. Specifically, the AI does not:

  • Calculate totals. Weighted scores are computed deterministically by the system, never by the AI.
  • Change idea status. Status transitions are explicit human or automatic-rule actions.
  • Assign reviewers. Reviewer assignment is a coordinator or admin action.
  • Write the final feedback. Auto-feedback is generated deterministically; admin feedback is written by an admin. The AI's summary is included as context, not as the verdict.
  • Block the workflow. If the AI can't score for any reason, the idea continues without it.

How the AI's score appears in the UI

Section titled “How the AI's score appears in the UI”

The AI's column on the scorecard is rendered with a gradient background so it's visually distinct from human columns. The reviewer pill on its row shows "AI first reviewer".

Per-criterion AI scores appear next to human scores. The AI's per-criterion comments are shown when you expand a row.

If the AI fails

Section titled “If the AI fails”

Sometimes the AI call doesn't complete. Examples: the AI service is temporarily unavailable, your workspace hit a rate limit, or the request timed out.

When that happens:

  • The AI's reviewer row stays at Pending.
  • The idea continues. Human reviewers can still score it and the idea can still reach scored once they finish.
  • The AI's score is simply absent from the weighted total. The decision is driven by human scores alone.
  • There is no automatic retry. If you want the AI to look at the idea again, an admin can re-trigger from the idea's detail page (when the option is available).

The product page displays a small note explaining that the AI didn't complete; reviewers continue normally.

What's next

Section titled “What's next”

AI settings

Workspace settings covers the AI tab where context and behaviour are configured.