What Does the EU AI Act Mean for Workforce Management and Quality Monitoring?

The EU AI Act classifies AI used to allocate tasks based on individual behavior, or to monitor and evaluate the performance of workers, as high-risk, and that description fits a growing share of the tools contact centers use to score and schedule agents. The obligations for these employment uses were originally due to apply on August 2, 2026. The Digital Omnibus on AI, published in the Official Journal of the European Union on July 24, 2026 as Regulation (EU) 2026/1744, moved them to December 2, 2027, which gives contact centers just over a year to prepare.

One rule that matters to contact centers is already in force. Since February 2, 2025, AI systems that infer the emotions of employees in the workplace have been prohibited, with narrow exceptions for medical and safety reasons. Any contact center running speech or sentiment analytics in the EU should be checking its configuration against that rule today, while the high-risk obligations give it time to plan the rest.

This article is general information about a complex regulation. It is not legal advice, and each organization should confirm its obligations with qualified counsel.

Which contact center tools count as high-risk under the EU AI Act?

AI tools count as high-risk under the EU AI Act when they are intended to be used to make decisions affecting the terms of work-related relationships, including promotion and termination, or to allocate tasks based on individual behavior or personal traits, or to monitor and evaluate the performance and behavior of workers. That language comes from point 4(b) of Annex III.

In contact center terms, the tools closest to that definition are the ones that assess individual agents. AI-driven quality scoring of an agent's interactions evaluates a named person, and so do performance analytics that rank agents against each other or feed into promotion decisions. Routing or scheduling logic that assigns work based on an individual agent's observed behavior also sits close to the definition. Forecasting that predicts aggregate contact volume by interval sits further away, because it models demand across the operation and makes no assessment of any individual.

The Act also includes an exemption that makes classification a case-by-case exercise. Under Article 6(3), an Annex III system is not treated as high-risk if it poses no significant risk and performs a narrow procedural task, improves the result of a previously completed human activity, detects patterns without replacing human assessment, or performs a preparatory task. That exemption never applies to a system that profiles individuals, so tools that score individual agents are unlikely to qualify. A provider that concludes its system falls outside the high-risk category must document that assessment before the system goes to market and register the system in the EU database. National authorities can request that documentation, and customers can reasonably ask a provider for it as well.

Is emotion detection on contact center calls allowed under the EU AI Act?

Inferring contact center agents' emotions from their voice or facial expressions is prohibited under the EU AI Act, while emotion detection aimed at customers is permitted as long as it leaves the agents out. Article 5(1)(f) bans AI systems that infer the emotions of people in the workplace, except for medical or safety reasons, and that ban has applied since February 2, 2025.

The European Commission's non-binding guidelines on prohibited practices, published on February 4, 2025, use a call center to illustrate exactly where the line sits. Using webcams or voice recognition to track employees' emotions, such as anger, is prohibited. Using voice recognition to track customers' emotions is allowed, provided the system does not track the employees' emotions at the same time. The guidelines also make clear that monitoring employee emotions for wellbeing or motivation purposes does not qualify as a medical reason.

For contact centers this turns into a concrete configuration question. Speech and sentiment analytics that score both sides of a call need to be set up so they do not infer the agent's emotions, which in practice usually means limiting emotion detection to the customer's side of the call. That is a check worth running now, with the analytics provider involved, because the penalties for prohibited practices are the steepest in the Act.

Customer-facing emotion recognition carries obligations of its own. Emotion recognition is listed as a high-risk use in Annex III, so the high-risk rules will reach those systems in December 2027, and since August 2, 2026, deployers have had to inform the people exposed to it.

What will employers have to do when the high-risk rules apply?

When the high-risk rules apply, employers that use these systems become "deployers" under Article 26, with duties centered on human oversight and transparency toward workers. Deployers must use the system according to the provider's instructions for use and assign human oversight to people who have the competence, training, authority, and support to exercise it. Where the deployer controls the input data, that data must be relevant and sufficiently representative for the system's purpose.

The operational duties are specific. Deployers must monitor the system in operation and report risks and serious incidents to the provider and the relevant authorities. They must also keep the logs the system generates, to the extent those logs are under their control, for at least six months. Before putting a high-risk system into service at the workplace, they must inform workers' representatives and the affected workers that they will be subject to it. People who are subject to decisions made or assisted by an Annex III system must also be told that the system is being used.

The Act's reach extends past companies headquartered in Europe. It applies to deployers established or located in the EU, and to providers and deployers outside the EU where the system's output is used in the EU. For non-compliance with deployer obligations, Article 99 sets fines of up to 15 million euros or 3 percent of worldwide annual turnover, whichever is higher, with the lower of the two applying to SMEs and small mid-caps. A global contact center operation with EU agents or EU sites should assume these rules are part of its planning.

How should contact centers use the time before December 2027?

Contact centers should use the time to inventory every AI capability that touches agents and classify each one by what it actually does. The useful question for each tool is whether it assesses or directs an individual agent, or whether it models the operation as a whole. Quality scoring, performance analytics, coaching recommendations, and behavior-based routing usually belong in the first group and deserve the closest review.

From there, the work is mostly documentation and design. Ask each provider for its instructions for use and its documented assessment of whether the system is high-risk, then review the logging and oversight controls the system offers against the duties above. Decide who in the operation will hold human oversight for each system and train them for it. Plan how agents and their representatives will be informed, and line that process up with existing data protection impact assessments under GDPR, which Article 26 explicitly connects to.

This is where Aspect's approach to AI in workforce operations is designed to help. Aspect Intelligence produces explainable recommendations that expose their drivers and keeps audit trails of what was recommended and what action was taken. Configurable approval workflows keep final decisions with planners and supervisors. An AI Trust bundle, including model cards, lineage, test evidence, and security scan reports, is available for customers to review. Each organization's obligations depend on how it configures and uses its systems, so its own legal review remains essential, and the right provider documentation makes that review faster and better informed.

What this means for your workforce strategy

The EU AI Act puts AI that evaluates and directs workers under a clear set of expectations built around human oversight and transparency. The December 2027 date is a deadline for compliance, and the preparation work starts much earlier, because inventorying tools and redesigning oversight take time in a large operation. The emotion recognition ban is already live, which makes speech analytics the first place to look.

The contact centers that come through this smoothly will treat it as an extension of good workforce practice. Explainable recommendations and clear human accountability were already the foundation of AI that supervisors and agents trust, and the regulation now makes them a requirement.