Teamspective and the EU AI Act
The EU AI Act's rules for high-risk AI at work apply from 2 December 2027. This page shows where the regulation stands, how Teamspective is preparing as the provider of its AI features, and what organizations that use Teamspective need to do as deployers.
Regulation timeline
The Commission's guidance on what counts as high-risk is still a draft. The high-risk rules apply from December 2027.
Be cautious of any vendor that claims full high-risk compliance today. No provider can complete a conformity assessment yet.
See the regulation timeline →Working position
Lea's most common use cases are not high-risk.
Some uses, such as evaluating a named employee, may be high-risk under Annex III, point 4(b). We are identifying them and will restrict them or meet the provider obligations for them before December 2027.
See our working position →Who does what
Duties for high-risk use start in December 2027.
Today: AI literacy for the people who use Lea, using Lea within its intended purpose, using no AI tool for a prohibited practice, and recording the AI Act in your internal policies.
See who does what →How Lea is built
Lea informs people. It does not decide about them.
No decisions about people and no rankings of individuals. Lea sees only what the user can already see, names the data behind its answers, and is always presented as AI.
Admins decide who uses Lea and with which instructions. Organizations that require it can keep Lea's processing within the EU, and customer data is not used to train AI models.
See how Lea is built →Regulation timeline
The Act entered into force in 2024 and applies in stages.
The AI Act enters into force
No obligations yet. The staged dates below start counting from here.
Prohibited practices and AI literacy
Banned uses include inferring the emotions of employees from biometric data such as voice or facial expressions (Article 5(1)(f)). Providers and deployers take measures to support the AI literacy of their staff (Article 4).
Draft guidelines on high-risk classification
The European Commission publishes draft guidance on Article 6, with examples for workplace AI. The final guidelines are still pending.
High-risk rules postponed to December 2027
The Digital Omnibus amendment (Regulation (EU) 2026/1744) moves the start of the high-risk rules for workplace AI from 2 August 2026 to 2 December 2027.
Transparency obligations
People must be told when they interact with an AI system (Article 50).
First standard and transparency guidelines
EN 18286:2026 on quality management systems is cited in the Official Journal on 14 October, and the Commission's guidelines on Article 50 follow on 12 November. Standards for the other high-risk requirements are still in preparation.
High-risk rules for workplace AI apply
Provider and deployer obligations for the Annex III systems, including employment and workers management (point 4).
Working position
We will publish the formal classification once the Commission adopts its final guidelines, and in any case before 2 December 2027.
| Feature | What the AI does | Working assessment |
|---|---|---|
| Lea, leadership guidance | Helps managers understand team-level engagement and feedback, prepare for conversations and develop their own leadership. | Not high-risk Lea's intended purpose. Leadership guidance is not a purpose listed in Annex III. |
| Lea, use cases about named employees | Evaluating a named employee's performance or behaviour, or producing input to decisions on promotion, termination, pay or task allocation. | May be high-risk Covered by Annex III, point 4(b). We are identifying these use cases and will either keep them outside Lea's intended purpose or meet the provider obligations for them before 2 December 2027. |
| Engagement comment summaries and insights | Summarize and group survey comments at team level, behind anonymity thresholds. | Not high-risk Aggregate results about teams. No evaluation of individuals. |
| Comment sentiment | Marks written comments as positive, negative or neutral. | Not prohibited, not high-risk Based on written text, not biometric data. |
| Writing help and translation | Helps people word feedback, role descriptions and surveys, and translates text. | Not high-risk Improves text a person wrote. Transparency rules apply. |
Who does what
Teamspective is the provider of its AI features. The organization that uses them is the deployer. Each has its own duties.
Teamspective, as provider
- No prohibited practices, including no emotion recognition Art. 5
- People are told they are using AI Art. 50
- AI literacy for our own staff Art. 4
- Document Lea's classification and register it where required Art. 6(4), 49(2)
- Publish Lea's intended purpose and the uses outside it Art. 13
- Risk management system Art. 9
- Data governance and technical documentation Art. 10–11
- Logging of Lea's operation, kept for at least six months Art. 12, 26(6)
- Instructions for use for deployers Art. 13
- Human-oversight design, accuracy and robustness Art. 14–15
- Quality management, conformity assessment and EU registration Art. 17, 43, 49
Your organization, as deployer
- Support AI literacy for the people who use Lea, such as admins and managers Art. 4
- Do not use any AI tool for a prohibited practice Art. 5
- Use Lea within its intended purpose Art. 25(1)(c)
- Use Lea as our instructions for use describe Art. 26(1)
- Name the people responsible for human oversight Art. 26(2)
- Monitor use and report serious incidents Art. 26(5)
- Inform employees and their representatives before Lea is used with them Art. 26(7)
How Lea is built
No decisions about people
Lea makes and triggers no decisions on hiring, promotion, termination, pay or task allocation. Those stay with people.
No rankings or scores of individuals
Lea does not rank, score or compare employees, and does not characterize individuals from their participation patterns.
Same access as the app
Lea answers only with data the user can already see in Teamspective. Access rights and anonymity thresholds are checked on every request.
Observations for human review
Lea frames findings as observations to explore, names the data it used, and offers recommendations as starting points.
Admins stay in control
Administrators decide whether Lea is on, for whom, and with which instructions. A discussion host can switch off Lea's suggested actions. An organization can use Teamspective without its AI features.
Clear that it is AI
Lea is presented as an AI agent wherever people use it: the web app, Slack and Microsoft Teams. Organizations that require it can keep Lea's processing within the EU.
Frequently asked
What does the EU AI Act say about HR tools?
Annex III, point 4
AI used in recruitment, to make decisions about promotion, termination or task allocation, or to monitor and evaluate the performance and behaviour of workers is listed as high-risk.
Article 6(3) exemption
A listed system is not high-risk when it only performs a narrow or preparatory task and does not materially influence a decision. The exemption never applies to a system that profiles individuals.
Human review does not change the class
The Commission's draft guidance says that a person reviewing the output does not by itself take a system out of the high-risk category.
Text is not biometric data
Analysing the sentiment of written survey comments is not prohibited emotion recognition. Analysing voice or facial expressions at work is.
How should we prepare our internal policies?
The AI Act is a regulatory requirement your policies should identify, even before its high-risk rules apply.
- Requirements register (A.5.31)
- List the AI Act, its application date of 2 December 2027 for high-risk systems, and the parts that already apply (Articles 4, 5 and 50).
- Supplier assessment (A.5.19–5.23)
- Record Teamspective's position from this page: provider role, working classification of Lea, controls in place, and the plan to December 2027.
- Scope of use
- Record which AI features you have enabled and for whom, for example Lea for administrators only during a preview.
- Planned actions
- AI literacy for current users now. Keep Lea within its intended purpose, and inform employees and their representatives before a wide rollout.
Is Teamspective "EU AI Act compliant"?
We meet the parts of the Act that apply today: no prohibited practices, transparency, and AI literacy. The high-risk obligations come next, and we are preparing for them as described above.
Is Lea a high-risk AI system?
Not as a whole. Lea's purpose is leadership guidance, which the Act does not list as high-risk. What matters is the use: evaluating a named employee's performance, or feeding a decision about their promotion, termination, pay or tasks, falls under Annex III, point 4(b). We are identifying those use cases and will either restrict them or meet the high-risk obligations for them before 2 December 2027. The Commission's draft guidance says human review alone does not change a classification, so we do not rely on "a person decides" as the argument.
Does Teamspective use emotion recognition?
No. We do not analyse voice, video, facial expressions or any other biometric data. Sentiment marking of written survey comments is text analysis, which the Act does not treat as emotion recognition.
Is our data used to train AI models?
No. Customer data, inputs and outputs are not used to train AI models. Our data processing agreement lists the subprocessors involved.

