How do I enable and configure the AI Review Assistant in Leapsome?
The AI Review Assistant is enabled through 'Settings' > 'Employees' > 'Feature Access Matrix' > 'AI features'. As a super-admin, you can toggle the Review Assistant on for your organization and configure role-level access for different roles from the same location.
The AI Review Assistant replaces the traditional review questionnaire experience with a guided conversation. Instead of answering static questions one by one, reviewers engage in a dialogue that draws on rich context from across Leapsome, goals, past reviews, peer feedback, and employee tenure, and uses that context to prompt more specific, evidence-based feedback. The result is higher-quality, more consistent reviews with less reviewer fatigue.
As an admin, you control which features are enabled, who can access them, and how visibility settings govern what context the assistant can surface.
Table of Contents
- What the AI Review Assistant does
- How to enable the Review Assistant
- What context does the Review Assistant use?
- How the look-back window works
- What reviewers experience
- How the Review Assistant protects data privacy
- FAQ
What the AI Review Assistant does
The Review Assistant is a conversational AI embedded directly in the review questionnaire. When a reviewer opens an assessment, they can choose to start a guided brainstorming conversation with the assistant instead of writing free text directly.
During the conversation, the assistant:
- Breaks the questionnaire into a dialogue, asking one topic at a time
- Surfaces relevant context automatically (goals, past feedback, employee info, peer reviews)
- Prompts the reviewer to be more specific or evidence-based when answers are vague
- Generates a draft assessment based on the conversation, which the reviewer can edit before submitting
The assistant never submits responses automatically. Reviewers always review and confirm drafts before they are saved.
How to enable the Review Assistant
To enable the AI Review Assistant for your organization:
- Go to 'Settings' > 'Employees' > 'Feature Access Matrix'
- Locate the 'AI features' section
- Enable the 'Reviews - assessment assistant'
- Configure role-level access: choose which roles (admins, managers, employees) can use the assistant
- Save your changes
To enable all AI features at once, toggle 'Enable all AI features' at the top of the 'AI features' section.
Please note: The Review Assistant is only visible on non-completed assessments that are in edit mode. Reviewers will not see the assistant option on submitted or locked assessments.
What context does the Review Assistant use?
The assistant draws on multiple data sources automatically when a reviewer starts a conversation. Understanding which sources are used and how visibility rules apply is important for configuring your review cycles correctly.
Past reviews and performance documents
The assistant accesses native Leapsome reviews from previous cycles. It also reads from the 'Past reviews' document category, automatically attaching uploaded documents (such as PDF reviews from before your Leapsome migration) that are either newer than one year or contextually relevant to the current review.
This means reviewers have full historical context even for employees whose performance history predates Leapsome. Documents are only surfaced if they are visible to the reviewer based on your document visibility settings.
Goal progress and qualitative comments
The assistant surfaces relevant comments from the reviewee's goal progress updates. For each relevant comment, it includes the goal name, progress state, and date, giving reviewers the qualitative context behind each goal without manual searching.
Only comments within the selected look-back window and visible to the reviewer based on goal visibility settings are surfaced.
Employee profile: role, level, and tenure
The assistant reads the reviewee's current job title, level, last title or level change date, and hire date. It uses this information to calibrate its guidance, generating role-appropriate expectations and asking questions that fit where the employee is in their career.
For example, if someone was recently promoted, the assistant frames its coaching around early expectations for the new role rather than treating them the same as a long-tenured employee.
Peer and self-review insights
When peer and self-review responses are visible to the reviewer based on your cycle's visibility configuration, the assistant incorporates them:
- Self-review: The assistant surfaces self-identified strengths and growth areas as conversation prompts
- Peer feedback: The assistant aggregates themes from anonymous peer responses and uses them to corroborate or challenge the reviewer's perspective
Critical admin note: the assistant only uses data that is already visible to the reviewer per your cycle's visibility settings. If you have configured peer feedback to be hidden from managers until a certain stage, the assistant respects that configuration. Anonymous peer feedback is referenced thematically and is never individually attributed.
How the look-back window works
At the start of a review conversation, reviewers see the default context window and can adjust it before the conversation begins.
| Option | Description |
|---|---|
| Since last review (default) | Context since the last completed review cycle for this person |
| Last 6 months | Context from the past 6 months |
| Last 12 months | Context from the past 12 months |
The selected window applies to all context retrieved in that session, including goals, meeting data, and documents. It is not saved as a permanent preference.
Please note:
- For employees hired less than 6 months ago, the window defaults to their hire date and shorter options are hidden
- If no data exists in the selected range, the assistant notifies the reviewer and suggests expanding the window
What reviewers experience
Once enabled, reviewers see a 'Start brainstorming' button at the top of their review questionnaire. Clicking it opens a chat interface.
The assistant guides the conversation question by question, encouraging specific and evidence-based feedback. Reviewers can track which questions still need input. When ready, they click 'Generate draft' to produce a full assessment draft, which they can edit, accept sections of, or reject entirely before final submission.
If a reviewer steps away mid-conversation, the session auto-saves. They can resume by clicking 'Resume chat with Leapy' when they return.
The assistant also supports voice interaction: reviewers can dictate their responses using the microphone icon, and can have assistant messages read aloud using the speaker icon. Voice dictation requires microphone access in the browser.
How the Review Assistant protects data privacy
The Review Assistant operates within your existing Leapsome access controls:
- The assistant only surfaces data the reviewer is already permitted to see based on your organization's visibility settings, goal permissions, and document access rules
- Anonymous peer feedback is never individually attributed, the assistant references themes only
- For more GDPR and data privacy details, please read here.
FAQ
1. Can I enable the Review Assistant for some roles but not others?
Yes. In 'Settings' > 'Employees' > 'Feature Access Matrix' > 'AI features', you can configure which roles, admins, managers, and employees, have access to the Review Assistant independently.
2. Does the Review Assistant change how review questionnaires are configured?
No. The assistant works alongside your existing questionnaire templates. It uses the questions in your template as the basis for its guided conversation but does not alter the template itself or how responses are stored.
3. Can reviewers skip the assistant and write directly?
Yes. The assistant is optional. Reviewers who prefer to write directly can do so, the standard questionnaire experience remains available.
4. What happens if a reviewer's direct report has no historical data in Leapsome?
If no data exists in the selected look-back window, the assistant notifies the reviewer and suggests expanding the window. The assistant can still provide general guidance based on the employee's role and tenure, even without historical review data.
5. Does the assistant use data from ongoing (incomplete) reviews in the same cycle?
Peer and self-review data from the current cycle is only surfaced if it is already visible to the reviewer per your cycle's visibility settings. Responses not yet shared or still in progress are not surfaced.
6. Is voice input transcribed and stored?
Voice input is transcribed in the browser and appears as text in the chat. It is treated the same as typed input — the transcription is not separately stored.