How do I use AI calibration tools in Leapsome Reviews?
Leapsome's AI calibration tools are available to calibration committee members from the calibration board. Once enabled by an admin, AI-generated pre-reads compile employee data automatically before each calibration session. During calibration, AI flags surface directly on employee cards and the employee context panel provides a full evidence view with AI rating suggestions, all from within the calibration board or heatmap.
Together, these tools give calibration committees the context they need to make consistent, evidence-based rating decisions without manual data gathering, and create an auditable record of every adjustment made.
Table of Contents
- What are AI calibration pre-reads?
- How to enable AI calibration pre-reads
- What is the employee context panel?
- FAQ
What are AI calibration pre-reads?
AI calibration pre-reads are automatically generated summaries compiled for each calibration committee group at the start of a calibration cycle. They give committee members a structured overview of employee data, including potential flags such as rating inflation signals, evidence mismatches, and cross-team alignment issues, before the live calibration session begins.
Pre-reads reduce the time committees spend manually gathering and summarizing employee data, and help surface patterns that might otherwise be missed. They are clearly labeled 'AI-generated' with a disclaimer, and are not applied automatically to any calibration outcome.
Pre-reads are generated per calibration committee group. If a committee member belongs to multiple groups, they receive a separate pre-read for each. Pre-reads include data from the last three completed review cycles where available.
Once pre-reads are generated, committee members receive:
- A task in the Home Dashboard checklist, labelled 'Review AI calibration pre-read'
- A banner on the calibration board showing the number of flagged employees
The banner appears in both the box view and the heatmap and clicking the banner opens a modal with the full pre-read details.
How to enable AI calibration pre-reads
To enable AI calibration pre-reads for a review cycle:
- Open the review cycle and go to the 'Calibrations' tab
- Ensure calibrations are enabled for the cycle
- Locate the setting 'Generate AI calibration pre-reads' and enable it
- Save your changes
Pre-reads are generated automatically on the calibration start date defined in the cycle's timeline settings. They cannot be triggered manually before that date.
To generate AI calibration pre-reads during the calibration session, click on 'Generate pre-read' button.
What is the employee context panel?
The employee context panel is a right-hand sidebar that opens when a committee member clicks an employee card in the box view or a cell in the heatmap. It provides a complete evidence view for that employee without leaving the calibration board.
The panel contains the following sections:
- Header: Employee name, job title, level, and tenure, with a link to the full employee profile
- Scores: All calibration-relevant question scores for the current cycle
- AI score suggestions: Each suggestion shows the question, the current score, the suggested score, and a brief AI rationale. Committee members can 'Accept' or 'Reject' each suggestion. Accepting a suggestion updates the calibration score and auto-generates a calibration note. Both accepted and rejected suggestions collapse to a compact state to keep the panel scannable.
- Evidence snapshot: Goals and OKRs linked to the calibration questions (with completion percentages), and feedback snippets relevant to those questions. Scoped to the X and Y axis dimensions only.
- Past performance reviews: A chronological list of all previous review cycles the employee participated in, with scores for calibration-relevant questions and links to the full reviews.
- Calibration history and notes: A unified log of all calibration activity for this employee in the current cycle, including score changes (with who made them, when, and any attached rationale) and manually added notes. An 'Add note' option is visible directly in this section.
When opened from the heatmap on a specific question, the evidence snapshot section automatically scrolls to that question's data.
Please note: AI suggestions in the panel are separate from the rationale modal triggered by rating changes. The panel surfaces suggestions from the pre-read analysis that can be accepted or rejected directly. The rationale modal is triggered by any manual rating change made via drag-and-drop or direct edit, whether or not a pre-read suggestion was involved.
FAQ
1. Can I generate pre-reads for some calibration groups but not others?
No. The 'Generate AI calibration pre-reads' setting applies to all calibration committee groups in the cycle. It cannot be scoped to individual groups.
2. What data is used to generate calibration pre-reads?
Pre-reads draw on data from the current review cycle and up to the last three completed cycles. This includes review scores, goal and OKR completion data, and feedback where visibility settings permit. The exact data scope matches the committee member's existing access rights — the pre-read does not surface data the committee member could not already view.
3. Are AI suggestions automatically applied to calibration scores?
No. All AI suggestions must be explicitly accepted by a committee member before any score is updated. Suggestions are shown as recommendations only and never applied automatically.
4. Is the rationale mandatory when adjusting a rating?
No. Providing a rationale is optional. The modal must be dismissed, but the rationale field can be left empty. If no rationale is provided, the rating change is still saved; only the attached note is absent from the calibration history.
5. Who can see the calibration history and notes in the employee context panel?
Calibration history and notes follow the same access rules as the calibration board and heatmap. Committee members active in the current cycle can view and add notes. Admins and managers with access to calibration analytics can view the history in closed cycles.
6. Can the AI calibration features be disabled without affecting the rest of the review cycle?
Yes. The 'Generate AI calibration pre-reads' setting can be left disabled in any cycle where you prefer to run calibration without AI pre-reads. The calibration board, box view, heatmap, and rationale modal continue to function without pre-reads enabled. If the workspace-level AI feature is disabled, all AI calibration functionality is turned off globally.