The EU AI Act: What you need to do before August 2026

The European Union's Artificial Intelligence Act (EU AI Act) is the first official AI regulation to be introduced globally. The EU AI Act became legally binding in 2024, and compliance requirements have been phased in since. 

August 2, 2026, marks the primary deadline, when the majority of the act's comprehensive obligations, governance rules, and requirements will take full legal effect. Exceptions could still apply for some high-risks systems, due to the EU's 'Digital Omnibus' agreement, which was reached politically on 7 May 2026, but has not yet been formally adopted. If this is not enacted before 2 August 2026, the original timeline applies as written. In short: watch this space.

The act has direct implications for CX and failure to comply will be costly: for the most serious violations penalties have been set at seven percent of global turnover or €35 million, whichever is higher. For context, violations of the GDPR are capped at €20 million, or up to four percent of global turnover. 

Ahead of the August 2026 deadline, this article explains: 

• Three key changes that will impact CX in 2026 and 2027.
• What the Act states and what it means for CX.
• The impact of AI disclosure on service capacity and workforce management (WFM).

For further reading on how the EU AI Act will impact customer experience, CX Network has these additional resources: 

• A comprehensive guide to navigating the EU AI Act
• The impact of the European Union's AI Act on customer experience
• How to use agentic AI in line with the EU AI Act
• A guide to procuring trustworthy AI solutions
  
  

Don't miss any news, updates or insider tips from CX Network by getting them delivered to your inbox. Sign up to our newsletter and join our community of experts. 

Learn More

3 Ways the EU AI Act will change CX in 2026

There are three specific elements of the next phase of the EU AI Act that will directly impact CX: 
1.    Transparency
2.    Emotion recognition
3.    AI literacy 

Transparency

What the Act states: The Act's exact wording states that all humans must be "informed that they are interacting with an AI system", unless it is obvious to a "reasonably well-informed, observant and circumspect [person], taking into account the circumstances and the context of use". 

While the transparency – and accompanying, obligatory alternatives – will likely be welcomed by customers, developers have spent years refining AI systems to operate in a way that is natural, conversational, and human-like.  

What this means for CX: This part of the Act has the potential to change the economics of AI deployment, as well as tightening the rules around it.

In short, organizations must fully disclose how, where and when a customer's interactions are being handled by AI, from chatbots in service to decision-making algorithms behind the scenes. This disclosure must appear both "in the moment" and in company policy – the Act states generic declarations of AI use will not suffice. This rule applies from August 2026. 

On how to disclose AI use in a customer-centric way, Eelko Lommers, director of CX, design and innovation for zooplus, says that every interaction presents an opportunity to build human connection. As such, he says notifications about the use of AI should be designed into the experience itself. 

"You must be explicit but completely natural at the same time. I recommend using consistent indicators, like a clear "Common Icon" or conversational welcome message that sets the right context immediately," he explains. "But I see this as just the start."

Lommers adds: "Soon, agentic interactions will be embedded in all customer experiences – from search to navigation, from content streams to customer support – as a tool to reduce friction. This is why it is important that these interactions and conversations are designed with great care, to deliver value, to be engaging, but always in service of what a customer needs. 

"Agentic content, whether you engage in conversation, or use generated visuals, video or text, like all content should show who the author is. And the author or creator could be an agent, which makes agentic interactions a key element of your content strategy," he adds. 

The Act doesn't stop at disclosure of AI use – humans must also be given the option to bypass AI and interact human-to-human if they wish. On the one hand, offering this choice has always been regarded as best practice. Lommers says: "Limiting options increases friction and feels like a lack of autonomy." 

However, regulating for this means those who work in contact center and workforce management now face new challenges in terms of capacity management. There's more on this below. 

Laurence Buchanan, leader of EY Studio+ says transparency should be clear and concise. "Customers should be told up front, in plain language, that they are interacting with AI and given clear and easy options to switch to alternative channels. Arguably this is good practice regardless of regulation as transparency, combined with a seamless interaction channel mix is foundational to building trust."

AI is not confined to service. When it is influencing a decision that affects the customer – such as a loan assessment, a claims decision, or a product recommendation – Sue Duris, principal consultant at M4 Communications, says disclosure must be explicit and detailed, both for technical compliance reasons and customer trust. "The customer needs to understand what the AI did, how it arrived at the decision, and that a human review is available," she explains. 

Emotion recognition

What the Act states: So long as the system in question is not being used for law enforcement, "deployers of an emotion recognition system or a biometric categorization system" must inform those impacted of the system's operation and process the personal data in accordance with GDPR. This rule has applied to workplace and educational settings since 2025, and will this year extend to the dozens of systems that track customer emotion. However, full compliance has been deffered. While organizations must disclose the use of these systems from August 2026, they have until December 2027 to meet the full high-risk compliance regime.

What this means for CX: The tracking of customer sentiment and emotion in CX is still permitted, but these tools fall into the "high risk" category, which brings mandatory obligations to: 

• Comply with data governance requirements, which feeds into the GDPR;
• Ensure robust cyber security measures;
• Build transparency into customer interactions;
• And ensure continuous human oversight.

Furthermore, high risk systems must be registered in the EU database before entering into service, and this rule applies to both providers and deployers. Organizations must also undertake a "conformity assessment", demonstrating their system meets the Act's requirements. This technical documentation can be completed as a self-assessment, but for some high-risk systems it must also be third-party verified. 

AI literacy

What the Act states: This rule is clear cut: staff must be appropriately trained before using AI. The wording reads: "Providers and deployers of AI systems shall take measures to ensure, to their best extent, a sufficient level of AI literacy of their staff and other persons dealing with the operation and use of AI systems on their behalf." The required level of AI literacy must account for technical knowledge, experience, education and training. It must also cover "the context the AI systems are to be used in, and consider the persons or groups of persons on whom the AI systems are to be used". Obligations have been in force since February 2025 under Article 4. This year's raises the bar on what AI literacy needs to deliver in practice. 

What this means for CX: When CX Network conducted its annual research into the state of CX in 2026, the need to drive data literacy across the organization emerged as the sixth most important strategic CX aim for practitioners, selected by 35 percent of respondents around the world in a list of 14 choices.

When asked about the changes they believe generative and agentic AI will bring to their organization's CX, 36 percent selected the requirement for teams to upskill from a list of nine choices.

Duris views the rules around AI literacy as an inflection point of enterprise AI use. "Most organizations treat AI skills as an L&D line item, a one-time training," she says. "The Act effectively requires AI literacy to be infrastructure – embedded in how teams work, not a box ticked at onboarding. That framing shifts responsibility from HR to leadership."

All those who work with, procure, or design AI for CX must understand how the systems work and, crucially, their capabilities and limitations, including how decisions are made. "Not technical expertise but enough to ask the right questions of vendors and internal teams," Duris explains.

• Hard Skills: Buchanan says these should cover "AI governance, risk and bias management, data oversight and business process management". Duris adds that hard skills must also include documentation and audit trail management and regulatory knowledge, such as understanding the Act's obligation by role, i.e. provider vs deployer, "and how they interact with GDPR, sector-specific regulation, and Consumer Duty in financial services". The ability to create and maintain logs can also be classed as a hard skills. 
• Soft skills: Duris says the key soft skills should include critical thinking, ethical reasoning, cross-functional communication, customer empathy, and escalation judgement. Buchanan adds that service design and people and change leadership should also feature. 

Skills must also be tested. 

"These should be tested through scenario-based exercises, ongoing monitoring of AI performance, CX measurement and cross-functional reviews," Buchanan says. "Ultimately, compliance is less about technology and more about trust – embedding transparency, choice, and accountability into the customer experience."

Elaborating on the need for scenario-based training, Duris says this should "present real situations and assess judgment, not just recall". She adds: "A customer service agent needs different AI literacy than a product manager or a compliance officer. Testing should reflect actual job responsibilities."

Testing should also be periodic. "AI capabilities and regulations change rapidly," Duris says. "Annual reassessment is the minimum, with assessment required more frequently for high-risk deployments." 

Finally, practical exercises can also support ongoing education and the application of new skills. Explaining skills should be demonstrated rather than declared, Duris says practical exercises could include "reviewing AI outputs for errors or bias, completing a risk classification exercise, or documenting a human oversight decision". 

What happens to service planning when customers don't want to interact with AI? 

As outlined, transparency must be reinforced with options: customers who do not want to interact with AI must have the option to continue their interaction in another way. 

From the customer's perspective, this offers choice. From the practitioner's perspective, the projected ROI on those expensive AI tools – based on the number of contacts that can be resolved without requiring a human agent – no longer works. 

Furthermore, service demand and agent capacity are meticulously planned. If an organization can no longer predict how many customers queries can be resolved by AI, the ability to meet demand – and keep metrics such as AHT and cost to serve under control – is diminished. 
Buchanan says this means organizations must plan for "a sustained level of human-served and assisted demand". 

He adds: "This creates complexity where volumes shift dynamically between AI and human channels. It also challenges business case assumptions around cost to serve. 

"However, leading CX and service organizations have always been strong at keeping a finger in the pulse of customer interactions, spotting the friction points in a customer journey – along with their implications – and designing those out of the experience," Buchanan continues. 

Duris says one way to do this is to model "two distinct service streams" to cover AI-handled interactions and human-handled interactions. In practice, this includes creating separate SLAs and routing rules for human-only service queues.

"If a significant percentage of customers opt out of AI, the human agent demand increases substantially," she says. In terms of headcount planning, this means "staffing models built around AI deflection assumptions need contingency planning for higher than expected human contact volumes".

She says peak demand modelling is also required as opt-out rates may spike after AI incidents or negative media coverage, creating unpredictable demand surges
When it comes to WFM, Duris says agent scheduling needs to account for a base level of human-only volume that can't be deflected. Training may also be needed as agents handling opt-out customers may be talking to customers who are "already frustrated or distrustful of AI, or to address more advanced questions". 

Although there are ways to manage unpredictable demand, Duris says a steady stream of customer opting to bypass AI is actually a "VoC signal organizations aren't tracking yet". She explains: "A rising opt-out rate is a canary in the coal mine for broader AI trust issues, before it shows up in CSAT or churn. CX leaders should be tracking and reporting this as a governance metric, not just a capacity planning variable."

However, as the quality of AI continues to improve while enhancing the efficiency of customer journeys, Lommers believes that many customers may still choose to engage with agentic solutions. "Customers will choose to engage because it will give them agency. Because we allow them to choose. This way we move past mediocrity and turn compliance into engagement and trust," he explains. 

Quick links 

• What high-performing companies do to improve CX
• How to tackle 5 challenges facing CX in 2026
• Is synthetic data the future of customer listening?

CX HORIZONS: The state of CX in 2026

Download our report to learn the results of CX Network's research into the top CX trends,  challenges, spending plans, and customer behaviors driving CX in 2026. Hear from thought leaders on what you should be doing now to ensure your organization is on track to become a strategic growth engine.

Download Now