Does AI decide the value or priority of an issue?

No. AI asks contextual questions and suggests a score from your answers. You review the explanation and confirm a score, enter your own, or skip the dimension. Confirmed scores feed the existing ranking; a human still decides what to publish to Linear.

Will ten people scoring the same issue see the same questions?

For the same issue version, dimension, and scoring setup within an organization, yes: the starting question and three choices are reused. Individual answers, follow-ups, and score suggestions are not pooled. Different evidence can lead to different scores.

When does a starting question change?

A new issue version or relevant scoring-guidance change creates a fresh question version. Changes to the model or question-generation instructions also refresh it. Different issues or organizations do not share cached starting questions.

Does sharing questions remove all AI costs?

No. It avoids generating the same starting question repeatedly. Each person's assessment and any follow-up still use AI separately. This improves consistency and reduces repeated work without replacing individual judgment.

What is sent to the AI provider?

After you consent, CoJii sends issue details, the relevant team scoring guidance, and the answers you choose to submit to OpenAI for suggestions. This is not an on-device model. Shared starting questions do not use personal answers. Other participants' answers and scores are not inputs to your assessment. Direct scoring is available without AI consent.

Is confidence assigned by AI?

No. Contributors choose their own confidence for their scores. Confidence influences the weighting of human responses; it is not another AI vote or a separate Strategic Alignment score.

AI-Guided Scoring: Questions First, Human Judgment Always

The hard part of prioritization is not choosing between 7 and 8. It is understanding what an issue changes, who benefits, and what evidence supports your judgment. CoJii starts with those questions.

Questions are the starting point, not an extra mode

Question-led scoring is the main scoring experience across CoJii. Instead of interpreting a rigid definition of every number, answer a short, issue-specific question about the dimension you are scoring. AI translates the evidence you provide into a suggested score and a brief explanation. You remain responsible for accepting or changing it.

The underlying dimensions and ranking system stay in place: Impact, Customer Signal, Strategic Alignment, Readiness, and Effort. You evaluate the dimensions enabled for your team or assigned to your scoring link or session. Contributors set their own confidence separately.

From an issue on your board to a confirmed score

Open the issue and a scoring dimension. Review the issue details and choose to continue with guided questions after reading the AI disclosure.
Answer what you know. Pick one of three contextual choices or write your own answer. There are no scores attached to the choices.
Clarify only when needed. AI can ask one follow-up if your answer does not provide enough evidence. It is not an open-ended chat.
Review the suggestion. Read the proposed score and its reasoning. It is a draft, not a submitted vote.
Confirm, change, or skip. Confirm a score you agree with, enter a score directly, or choose “I don't know — skip.” An unknown is not automatically a low score.

Direct scoring is a shortcut within this experience, not a competing scoring mode. It remains available without AI consent, including when AI is unavailable.

Example: “Remember board filters”

Imagine an issue to preserve someone's board filters between visits. For Impact, a starting question might be: “How many users are currently affected by rebuilding their board filters on each visit?”

Answer with what you actually know: who uses the filters, how often the problem occurs, and whether it is a convenience or a serious workflow interruption. AI should not invent usage figures or revenue to fill the gaps. If you cannot judge the impact, skip it rather than treating the missing evidence as proof of low value.

Effort gets its own question about the implementation work. A useful feature is not necessarily easy to build, and a short issue title does not establish engineering effort. Existing effort formats remain available; the questions help you reason before confirming an estimate.

Same starting question does not mean the same score

Within an organization, people evaluating the same issue version, dimension, and scoring setup receive the same starting question and choices. CoJii generates that starting point once and reuses it, including when several people arrive at the same time.

Your assessment uses your own answer and any evidence you have already provided in this scoring flow. Your follow-up and suggested score can therefore differ from someone else's. Other participants' answers and scores are not used to steer your suggestion. Confirmed scores still participate in the normal team results.

Changes to the issue or relevant scoring guidance create a fresh starting-question version. Reuse does not make AI free: personal assessments and follow-ups still require AI requests. It avoids paying to regenerate the same starting question for every contributor.

How it fits the workflows you already use

Manager scoring: review and confirm draft dimension scores, then apply the drafts to the issue. This sets manager scores; it does not add an extra crowd response.
Individual ad-hoc links: answer the requested dimensions, confirm your scores, choose your confidence, and submit through the existing scoring form.
Campaigns: use the same question-led flow for the issues and dimensions in the campaign, without needing a Linear account.
Live sessions: answer independently for the current issue, confirm your scores and confidence, and use the session's normal ready and results steps.

AI never adds an extra vote, chooses your confidence, or publishes a priority to Linear. A manager's direct score remains distinct from contributor responses.

What sharing means for privacy and access

AI suggestions use OpenAI, not a local model. After consent, issue details, relevant team scoring guidance, and the answers you submit are sent to the provider. Only submit information your organization permits you to send. You can use direct scoring without agreeing to AI processing.

The shared starting question is based on issue context and team guidance, not personal answers. Shared question storage is separated by organization and issue identity and does not contain personal answers or individual score suggestions. Reuse never bypasses normal link, session, assigned-dimension, or organization access checks. A closed scoring link cannot retrieve its cached question.

Use the score to support a decision

Better questions make the inputs easier to explain; they do not make the evidence certain. Review disagreement, gaps, and confidence before committing to a priority. For the wider process, see the Linear prioritization guide.