Feedback - Draft with AI FAQ's

Employees can now request help when writing feedback to their colleagues by using the Draft with AI feature.

Employees can who prepare their draft feedback on the platform and use Draft with AI request the AI Tool to review and return a reworded version of it (“Suggested Feedback”), which the Employee can then accept, edit or discard.  The Suggested Feedback returned by the AI tool aims to elaborate upon, and improve the spelling and grammar of, the Employee’s draft feedback.

This FAQ explains how the Generative AI works and the measures that Clear Review has adopted to ensure its safe and lawful use.

1. What is Generative AI?

Generative AI refers to deep-learning models that can generate high-quality text, images, and other content based on the data they were trained on.

1. Why has Clear Review integrated the AI Tool on its platform?

Some Employees may find it difficult to write feedback about their colleagues.  The AI Tool is intended to help Employees write their feedback about their colleagues in a creative and friendly way, and its use is entirely optional.  It will only provide Suggested Feedback if the Employee requests help from the AI Tool.

1. How has Clear Review implemented AI?

The AI Tool introduces a new AI feature called ‘Draft with AI’ on the Clear Review platform.  If Employees click on it, it will analyse their Draft Feedback and use the AI Tool to re-word it and return alternative Suggested Feedback that, for example, improves spelling and grammar, or more clearly explains or expands upon the Employee's Draft Feedback. More precisely:

• Employees can type their Draft Feedback on the platform and will have the option to click the ‘Draft with AI’ button.
• On clicking ‘Draft with AI’, the AI Tool will analyse the text and provide alternative, reworded Suggested Feedback.
• The Suggested Feedback will be displayed to the Employees.  They can choose to use the Suggested Feedback as is, or to edit or discard it and use their original text.
• The AI Tool does not make any decisions about, or profile, the Employee or their colleague – it simply provides Suggested Feedback based only on a review of the language of the Draft Feedback originally entered by the Employee. 

1. Does the AI Tool process personal data?

Yes.  When an Employee enters feedback about a colleague onto the platform, that feedback will necessarily constitute personal data about the Employee and their colleague (because, for example, it will represent the opinion of the Employee, and may describe the working relationship between the Employee and their colleague). 

Consequently, the Draft Feedback that Employees enter, and the Suggested Feedback returned by the AI Tool, will represent personal data about the Employee and their colleagues. 

This personal data is at all times protected and processed in accordance with the Data Processing Addendum / Schedule between the customer and OneAdvanced.

1. Does the AI Tool process sensitive data^[1]?

Clear Review does not encourage or require Employees to submit sensitive data when providing feedback about their colleagues, and the AI Tool does not infer any sensitive data about Employees or their colleagues when processing their Draft Feedback and returning Suggested Feedback. 

The AI Tool only re-writes the Draft Feedback initially provided by the Employee to provide Suggested Feedback.

1. Is Clear Review a controller^[2] or processor^[3] of the personal data processed within Draft Feedback or Suggested Feedback by the AI Tool?

Clear Review processes all personal data as a processor. 

This means that it processes any personal data that may be contained within Draft Feedback and Suggested Feedback only on behalf of its customers and subject at all times to the Data Processing Addendum / Schedule between the customer and OneAdvanced.

1. Who is the Controller of the personal data processed within Draft Feedback and Suggested Feedback by the AI Tool?

Clear Review customers are the controller of their Employees’ personal data, including personal data their Employees submit to the Clear Review platform.

1. Does Clear Review use any third party to deliver the AI Tool’s functionality?

Yes, Clear Review has integrated with the Microsoft Azure OpenAI service.  More precisely:

• the generative AI model used by the AI Tool is ChatGPT-35 Turbo (the “Model”);
• Clear Review utilises the Model from Microsoft and runs it on Clear Review’s dedicated instance of Azure OpenAI
• There is no sharing of either Draft Feedback or Suggested Feedback with OpenAI – all processing of Draft Feedback and Suggested Feedback is undertaken solely by Clear Review and Microsoft.

1. What type of processing activities are carried out by Microsoft?

When Employees click the ‘Draft with AI’ button, the Clear Review platform sends the Draft Feedback to Clear Review instance of Azure OpenAI, where the Model is deployed to process the Draft Feedback, and apply content and abuse filtering (see Questions #17 “Is there a risk that the AI Tool may return Suggested Feedback that is inaccurate or deceptive?” and # 18 “Is there a risk that the AI Tool may suggest Suggested Feedback that is harmful?”).

Any draft or suggested feedback are retained by Microsoft for 30 days only. This is only for the purpose of abuse monitoring, whereby any content which is flagged as harmful will reviewed by Microsoft employees in a secure manner. (Note: Equally, the human reviewers employed by Microsoft are located in the same region where the Azure instance running the Model is located. See question #18 “Where are the Draft Feedback and suggested feedback processed?)

1. Does Clear Review have an agreement with Microsoft that complies with GDPR?

Yes, Clear Review has put in place contractual terms with Microsoft to ensure that they process Draft Feedback and Suggested Feedback as a data processor and provide sufficient guarantees to implement appropriate technical and organisational measures, in such a manner that their activities will meet the requirements of GDPR.

1. Where are the Draft Feedback and Suggested Feedback being processed?

Clear Review has implemented an Azure OpenAI instance in the European Economic Area (EEA). Any Draft Feedback and Suggested Feedback processed by our AI tool is processed in the EEA. The human reviewers employed by Microsoft are located in the same region where the Azure instance running the Model is located.

[1] In these FAQs, we use the term “sensitive data” to refer to data that is deemed “special category data” under the GDPR – i.e. personal data revealing racial or ethnic origin, political opinions, religious or philosophical beliefs, or trade union membership, and genetic data, biometric data for the purpose of uniquely identifying a natural person, data concerning health or data concerning a natural person’s sex life or sexual orientation.  In these FAQs, the term “sensitive data” will also include personal data about any criminal convictions or offences relating to an individual.

[2] “Controller” means an entity that determines the purposes and means of the processing of personal data (i.e. why and how personal data are processed).

[3]“Processor” means an entity which processes personal data on behalf of a controller.

[4] Draft Feedback and Suggested Feedback is scanned for harmful responses and may be subject to manual review.  See FAQ 13 for more information.

1. Will the AI Tool be used to make automated decisions about any Employees?

No, the AI Tool is used exclusively to provide Suggested Feedback.  The AI Tool is a large language model that operates solely by analysing the Draft Feedback provided by the Employee and providing reworded Suggested Feedback.  In doing so, the AI Tool does not profile nor make any automated decisions about Employees or their colleagues.  Employees are able to choose whether to use the Suggested Feedback as is, or to edit or discard it.

1. Does Clear Review use the Draft Feedback and the Suggested Feedback for any other purpose?

Clear Review does NOT use Draft Feedback and Suggested Feedback for any other purpose^[4].  In particular, Clear Review does not use Draft Feedback or Suggested Feedback for any AI training purposes, or to improve Clear Review’s services or to develop new ones.  See also Question #11 “Are Draft Feedback and Suggested Feedback being used to train the AI Tool?”

1. Does Clear Review retain the Draft Feedback and the Suggested Feedback, and if so for how long?

Draft Feedback and Suggested Feedback are retained by the AI Tool for a maximum period of thirty (30) days.

1. Does Clear Review warn Employees that their Draft Feedback may be processed by the AI Tool?

Yes.  Clear Review provides a tooltip which pops up to inform Employees that, by clicking the ‘Draft with AI button’, their Draft Feedback will be processed by the AI Tool for the purpose of providing Suggested Feedback.

As the controller of their Employee’s personal data, customers have a responsibility to inform their Employees how their personal data will be processed (including when they use the Clear Review platform) and their rights under applicable data protection laws.  The purpose of these FAQs is to provide Clear Review customers with the information they need in order to fulfil their own transparency obligations towards their Employees.

1. Are the Draft Feedback and Suggested Feedback used to train the AI Tool?

No, the AI Tool is ‘stateless’.  This means that Employees’ feedback – whether Draft Feedback or Suggested Feedback – is NOT used to train, retrain, or improve the AI Tool.

1. Is there a risk that the AI Tool may return Suggested Feedback that is inaccurate or deceptive?

The Suggested Feedback is based solely on the Draft Feedback provided by the Employees.  The AI Tool provides Suggested Feedback by rewriting the Draft Feedback provided by the Employees.  In other words, the AI Tool does NOT provide any “new” knowledge or information about the subject of the feedback (i.e. the Employee’s colleague) in the Suggested Feedback it provides to the Employee.

Further, Employees who click the ‘Draft with AI’ button will have the opportunity to amend the Suggested Feedback provided by the AI Tool before it is submitted, to ensure it accurately reflects their own personal views and, if necessary, they can edit it to remove anything in the Suggested Response that they consider may be inaccurate or misleading.

1. Is there a risk that the AI Tool may suggest Suggested Feedback that is harmful?

The AI Tool includes safety features that filter Prompt Feedback and Generated Content in real time to flag content that may contain harmful language.  If Draft Feedback is flagged as potentially harmful, it will be subject to manual review by a Microsoft Employee.

The content filtering system integrated in the Azure OpenAI Service contains neural multi-class classification models aimed at detecting and filtering harmful content; the models cover four categories (hate, sexual, violence, and self-harm) across four severity levels (safe, low, medium, and high).

Any feedback flagged as harmful are not stored in the models or used to train, retrain and improve the models.

1. Is there a risk that the Suggested Feedback may reflect any bias?

There is always an inherent risk of bias in any HR performance review process where one Employees provide feedback about their colleagues.  However, the Clear Review AI Tool has been trained to identify and minimise the risk of bias.  Specifically, human reviewers have been involved in training the AI Tool to catch and correct biased behaviour during its training process.  For this reason, Suggested Feedback generated by the AI Tool helps to reduce bias that may exist in “normal” human feedback provided by Employees as part of an HR review process.

1. Is the Draft with AI Feedback feature available for my organisation?

   The Draft with AI Feedback feature is included as part of our Standard and Ultimate subscription.
   If you are a system administrator and interested in finding out more, please contact your account manager or alternatively, email cr.clarissa@oneadvanced.com