π Want to know how to prepare your evaluation step by step with AI? Then read this article.
Learned uses AI to prepare draft versions of reviews based on collected feedback from various sources within the platform, such as 1:1 notes, captured goals and manually submitted observations via email or other integrations. These automatic summaries help managers and employees draft reviews faster and more focused, without losing quality or nuance.
How it works:
The AI functionality within Learned generates a draft review based on previously collected and recorded data. Information from multiple sources is combined to provide as complete and representative a picture of the employee as possible.
When a Learned user prepares a review, they are given the choice to start manually or use AI. When the AI option is selected, several steps are automatically performed. First, the questions in the review form are analysed. At the same time, the system checks whether relevant insights and/or a historical review are available in Learned.
The available insights are carefully analyzed to find relevant information that helps answer the questions. Based on this information, the system always performs a sentiment analysis. When relevant insights are available, this analysis is used to determine the score.
If a previous review is available, this is taken as the starting point by default and the sentiment analysis is used to determine whether the score should be higher or lower.
π‘: Where possible, the scores are always elaborated and explained by AI based on the available information, so that the user gains insight into how the review score was established.
The manager or employee always retains the option to edit this draft text before it is finalised.
The AI uses input from the following sources:
Insights from 1-1s and goal updates in Learned.
Notes made on the Insights page in Learned.
Insights forwarded from external systems, such as: Mail, Microsoft Teams or Slack.
Relevant information from historical reviews.
βοΈ: To use the AI responses, you must have collected at least 3 insights.
If the system is unable to find any suggestions, this will be indicated in the form's fill-in view.
Privacy & data processing in generated reviews
When generating these draft evaluations, Learned strictly adheres to the applicable privacy legislation (AVG/GDPR). The processing is carefully designed with the following safeguards:
Only data within the Learned platform: Information explicitly entered or shared by users within the Learned platform is included in the AI analysis.
Manageable and transparent data processing The AI generates only a draft proposal. Users always retain full control: they can approve, modify or delete feedback before anything is finally saved or shared.
Data minimisation: Only relevant parts of feedback (such as observations, trends or recurring themes) are used to structure the evaluation. Not all raw data is taken verbatim.
Security and storage: All data is stored and processed within Learned's secure infrastructure. Access is restricted to authorised users within the organisation.
Processor agreement and AVG/GDPR compliance: Processing is covered by the existing processor contract with Learned. Organisations retain control over which functionalities are active and who has access to the generated content.
Bias, transparency & consent in AI applications
When using AI in staff review, care is essential. Learned is aware of the risks around bias, the black box effect and the need for active consent. Therefore, additional measures have been taken to make the use of AI fair, insightful and voluntary.
Preventing bias and discrimination
AI models - just like people - can contain unintentional biases, for example based on language use, gender or cultural differences. To prevent this:
The AI is continuously trained on diverse and representative datasets.
Output is never applied automatically: users always assess and edit the generated text themselves.
No sensitive personal data or signals that could lead to discrimination are explicitly used.