Skip to content
Sparqbox Docs

Reading AI scores as a reviewer

The AI gives you a starting point, not a verdict. This page is for coordinators who want to use the AI's score well.

Score independently first

Section titled “Score independently first”

Before you look at the AI column, score the idea on your own.

Reading the AI's score before you score yourself anchors your judgment. Even when you think you're disagreeing, you tend to disagree relative to what the AI said, not relative to the idea on its own merits. Scoring blind is a five-minute discipline that pays off.

When you've finished your own scorecard, then look at the AI column.

What the AI knows, and what it doesn't

Section titled “What the AI knows, and what it doesn't”

The AI sees the same idea text and the same criteria you do. By default, it doesn't see who submitted the idea (most workspaces leave the Anonymise submitter setting on).

What the AI doesn't have, unless your admin has filled it in:

  • Your organisation's history with similar ideas.
  • The internal context of who's championing what.
  • Specific knowledge about your customers, products, or market.

If the AI's scores feel generic, ask your admin to fill in Settings → AI. Organisation background, industry and market, products and services, and strategic priorities all feed into the AI prompt and noticeably sharpen the output.

The AI's weight in the final score

Section titled “The AI's weight in the final score”

How much the AI counts depends on a workspace setting called AI reviewer weight, in Settings → AI.

  • Default is 1.00. The AI counts equally with the human average.
  • At 0.50, the AI counts half as much as the average human.
  • At 0.00, the AI is excluded from the total entirely. Its scores still appear; they don't enter the average.

For the math, see How scoring math works.

When you and the AI disagree

Section titled “When you and the AI disagree”

A sharp gap between your score and the AI's score is itself useful. It usually means one of three things:

  • The criterion is ambiguous, you and the AI are reading the question differently. Worth raising with the admin who owns the rubric.
  • The idea doesn't say enough, one of you is filling in gaps with assumptions. Asking the submitter for clarification often resolves this.
  • The AI is missing context, the idea makes sense in light of internal information the AI doesn't have. Your score wins; consider adding the missing context to Settings → AI so the AI is better next time.

Don't try to reconcile every gap. The point is to surface them.

What's next

Section titled “What's next”

Score an idea

Scoring an idea is the procedural reference for coordinators.