The EU AI Act Deadline Puts Customer-Facing AI Under the CX Microscope

The EU AI Act has reached a major milestone for customer experience teams. As of the beginning of August, Article 50’s transparency requirements apply to certain AI systems, including those that interact directly with people. Under the rules, customers must be informed when they're interacting with an AI system, unless that fact is already objectively obvious. The European Commission’s guidance also addresses labeling AI-generated content and other transparency requirements. For CX leaders, that puts chatbots, voice assistants and AI agents under a new level of scrutiny. But compliance may require more than adding a disclosure to a chatbot.

Antony Passemard, VP of Customer Strategy at Cresta, sees the regulation as a positive development for customer trust. "Overall, I think it's a great Article," Passemard said, arguing that it aligns with the idea of using AI to enhance the human experience rather than replace it. The first challenge for CX teams, then, is knowing exactly where AI is entering the customer journey and what customers can actually see.

Disclosure becomes part of CX design

The requirement may sound straightforward: tell customers when they're talking to AI. Passemard agrees that the basic disclosure shouldn't be particularly difficult to implement. "We have no issues with being explicit that it's an AI agent serving the end customer. In fact, we recommend it," he noted. Cresta recommends that customers be explicit about AI use rather than attempting to make an automated interaction pass as human.

But the more interesting question for CX teams is how that disclosure is delivered. Colleen Jones, president at Content Science, argues that transparency needs to be treated as part of the experience itself. "The test should be whether a reasonable customer can tell, without having to stop and investigate, that they’re interacting with AI," Jones asserted. That means CX leaders may need to think beyond the legal wording of a disclosure and consider its placement, language, and timing. A buried privacy notice may technically contain the information customers need, but it does little to help someone understand who, or what, is actually handling their request.

The human-in-the-loop question gets complicated

Article 50 also gives CX organizations a reason to examine the boundary between AI assisting employees and AI communicating directly with customers. Passemard said Cresta's Agent Assist functionality generally isn't affected in the same way when AI is supporting a human agent, because the human remains in the customer interaction. But when the technology can generate and send messages without human intervention, the situation changes. "In that case, our customers would have to disclose as part of that message that the content has been AI-generated," he explained.

That distinction matters as customer service teams move toward increasingly autonomous AI. A system might begin as an internal recommendation engine and gradually gain the ability to communicate directly with customers. For CX leaders, compliance therefore can't be reduced to a list of bots. Teams need visibility into what AI can actually do, when humans are involved, and when an automated system has the authority to act or communicate on its own.

AI governance moves closer to the customer

That visibility problem extends beyond contact center and customer experience platforms. Cory Maffeo, founder and CEO of Agentive Integrations, believes that organizations need to understand the AI systems embedded throughout their technology stack, not simply the tools they intentionally purchased as AI. "Do you know every automated channel a customer can reach in your business? Most owners miss one," he pointed out.

This is particularly relevant for CX teams, where AI can be embedded inside CRM systems, analytics tools, workforce platforms, and other software without necessarily being the headline feature. And as voice AI becomes more convincing, transparency becomes even harder to treat as a checkbox. Passemard points to AI-generated voice as an area that could require technical solutions, including machine-readable watermarking.

Alexandros Koronakis, senior director and head of government, policy and regulatory affairs, EMEA at Twilio, similarly frames the rules as an operational shift rather than simply a legal deadline. The transparency obligations are "an opportunity to redefine the customer experience," he said, particularly as companies look to build digital trust without introducing unnecessary friction.

For CX leaders, that may ultimately be the bigger takeaway from the August 2nd deadline. The challenge isn't simply informing customers that AI is involved. It's how organizations can make that transparency feel like a natural part of the experience while giving customers confidence about what happens next. As AI becomes harder to distinguish from human interaction, knowing when to say "you're talking to AI" may become just as important to CX design as knowing when to say "how can I help?"