Skip to main content

​
The rule

AI in Weivo writes about your data and proposes things for you to approve. It never casts a vote, never edits a rating, and never decides a priority. Every number on every output comes from stakeholder ratings, coverage and effort that a person entered or confirmed.

​
Three surfaces

​
The assistant

The Assistant button in the header, or the round button in the bottom right of the screen, opens the assistant. It knows the current project: its planning mode, vocabulary, needs, stakeholders and weights, features and their efforts, and what is already linked on Coverage. Ask it to set up a project from a description, to review whether the weights look balanced, to find needs nothing delivers, or to draft the missing half of a backlog. When it has concrete items to suggest, it returns cards: needs with a category, stakeholders with weights that sum to 100 alongside the existing roster, features with an effort and, in a Weighted project, coverage scores for the needs they deliver. Each card has Apply; Apply all applies a whole batch. Nothing is created until you apply it, and the assistant is told what became of its earlier proposals so it does not repeat them. Coverage scores it proposes arrive on the grid as unconfirmed suggestions, drawn with a dashed edge and counted above the grid. They count toward the ranking at once, and they stay marked until you keep or change them. See Coverage. The Let AI set it up button on a new project’s Home tab opens the assistant with a ready prompt. On a demo project the assistant answers questions but cannot apply anything, and says so.

​
Explain this conflict

On the Analysis tab, open a Mixed or Disputed feature’s Details and click Explain this conflict. The AI describes the disagreement from the current ratings: who rated the relevant needs how, how their weights pulled the score, and what is worth checking. Withdrawn ratings are never quoted.

​
The report brief

Generate Summary on the Decision Report writes a brief in short bulleted sections (three in a Weighted project, two in a Direct one) from the numbers on the page, under rules that forbid inventing totals, links or people. Regenerate rewrites it and republishes the copy every share link serves. See Decision Report.

​
Vocabulary and context

Every prompt uses your project’s vocabulary and the strategic goal and success signal from Project Settings, so an assistant working on a portfolio talks about objectives and programs, not needs and features.

​
Allowance

AI is included on every plan and is not metered per plan. Each AI feature has a daily fair-use allowance; if you reach it, the app tells you which feature is paused and when it resets.

​
Your data

AI features call the Anthropic API. Anthropic does not use API inputs to train its models. What is sent is the project content the feature needs: for the assistant, the current project state and your messages; for a conflict explanation, the ratings behind that feature; for the brief, the report context. Nothing from other projects or accounts is included. See the privacy policy and sub-processors.