When you allow customers to interact with an AI system, you are often legally required to disclose this. The EU AI Act and European transparency standards require that users be informed in a timely and clear manner when they are interacting with an automated system, so that they can make an informed choice. In this article, you’ll find practical answers to the most frequently asked questions about AI transparency in customer interactions, ranging from legal obligations to the best way to word your disclosure.
Are companies legally required to report this?
Yes, in most cases, companies are legally required to inform customers when they are interacting with an AI system. The EU AI Act (Regulation (EU) 2024/1689), which entered into force on August 1, 2024, imposes transparency obligations on both providers and controllers of AI systems, particularly in customer-facing applications.
Specifically, the law requires that AI systems designed to communicate with people—such as chatbots and virtual assistants—clearly disclose this to the user. This also applies to systems that detect emotions or categorize individuals. The obligation to inform customers about AI interactions falls under the “limited risk” category in the AI Act’s risk classification, but that does not mean it is non-binding.
In addition to the AI Act, the GDPR and sector-specific regulations also play a role. Furthermore, if you use AI in decisions that directly affect customers—such as assessing an application or handling a complaint—the customer may, under Article 86 of the AI Act, request an explanation of the factors that determined that decision. The compliance requirements for high-risk systems will become enforceable on August 2, 2026, but the obligation regarding AI literacy and the prohibitions on manipulative AI practices have been in effect since February 2, 2025.
In short: don’t wait for enforcement. The transparency requirement is already a legal reality.
When is the best time to inform customers about AI?
The best time to inform customers about AI is at the beginning of the interaction, before the substantive part of the conversation begins. This allows the customer to make an informed decision about how they want to proceed with the conversation and helps prevent them from feeling misled afterward.
In practice, there are two logical points in time:
- During the first interaction: As soon as a customer opens a chat, initiates a phone call, or sends a message, let them know right away that they are communicating with an AI system. This doesn’t have to be a lengthy explanation; a short, clear sentence is sufficient.
- When switching channels: When a customer is transferred from a human agent to an AI system—or vice versa—a new notification should be sent. Customers expect consistency and are likely to abandon the process if they feel they are being transferred without their knowledge.
What you want to avoid is burying the notice in the terms and conditions or somewhere at the bottom of a page. Transparency only works if the customer actually sees and understands the information. A notice that doesn’t appear until after three rounds of messages is too late and can undermine trust.
How do you word an AI notification without scaring customers away?
A good AI message is brief, friendly, and emphasizes the benefits to the customer. There’s no need to point out that it’s “just” a robot. Instead, focus on speed, availability, and the option to always be transferred to a human.
Here are a few practical tips for effective wording:
- Be direct but positive: “You’re now chatting with our AI assistant. It will help you quickly, even outside of business hours.” This is transparent without sounding alarming.
- Avoid technical jargon: Words like “automated system” or “machine learning interface” can raise suspicion. “Digital assistant” or “AI assistant” work better.
- Always offer an alternative: Make it clear right away that the customer can ask to speak with a human representative at any time. This makes it much easier for them to reach out.
- Keep it short: Two sentences are enough. No one reads a long disclaimer, and it’s more likely to raise suspicion than inspire trust.
The tone of your message sets the tone for the entire conversation. A warm, clear introduction makes customers more receptive to the AI interaction that follows.
How do customers react when they find out it’s an AI?
Customers’ reactions to AI notifications depend heavily on how the notification is delivered and what the customer expects. When the notification is friendly and transparent, and the AI then provides quick and accurate assistance, most customers react positively or neutrally. Resistance arises mainly when the AI fails to meet expectations.
What we see in practice is that customers accept AI in customer interactions as long as a few basic conditions are met. They want to know that a human is available if needed. They don’t want to have to repeat their story when they’re transferred. And they want the AI to actually answer their question, rather than just referring them elsewhere.
Customers who react negatively to AI generally do so not because it’s AI, but because they’ve had bad experiences in the past with chatbots that couldn’t solve their problems. Transparency combined with a well-functioning system is therefore the key. When customers realize that the AI is actually helping them, their resistance quickly diminishes.
When should an AI conversation be handed over to a human?
An AI conversation must be transferred to a human agent as soon as the situation falls outside the AI’s competence or mandate. These are the most common situations in which a transfer is necessary.
- Emotional or sensitive situations: When a customer is frustrated, sad, or angry, a human response is more valuable than an automated reply.
- Complex or unusual questions: If the AI cannot answer the question or provide an appropriate solution, transferring the call is the only correct option.
- Explicit customer preference: When a customer requests a human agent, the transfer must take place immediately and smoothly.
- High-risk decisions: For decisions that have direct consequences for the customer—such as rejecting an application or amending a contract—human oversight is not only desirable but, in some cases, also required by law.
The transfer itself is just as important as the moment of transfer. Make sure the human agent receives the full context of the conversation so the customer doesn’t have to repeat their story. This is one of the most commonly cited frustrations in customer interactions and can be completely prevented with the right technical setup.
What makes an AI notification credible to customers?
An AI notification is credible when it is consistent, is followed by behavior that lives up to its promise, and makes the customer feel that their interests are the top priority. Credibility isn’t built with words alone, but through the entire customer experience that follows.
Specifically, this means:
- Consistency across channels: If you mention on your website that you use AI but don’t mention it in the app or over the phone, it undermines trust.
- AI does what you promise: If you say the assistant will help quickly, it has to do just that. An AI that keeps customers running in circles not only harms itself but also undermines trust in your organization as a whole.
- Human oversight is visible: When customers know that people are monitoring the interaction and that they can always escalate the issue, they feel safer during an AI interaction.
- No hidden agenda: Don’t use AI to keep customers away from human employees. Customers can sense this, and it damages your reputation.
Ultimately, credibility is the result of honesty plus quality. A transparent notification from a well-functioning AI builds trust. A transparent notification from a poorly functioning AI only increases disappointment.
How Pegamento Helps Ensure AI Transparency in Customer Interactions
At Pegamento, we understand that AI transparency is not only a legal requirement but also an opportunity to strengthen your customers’ trust. We help organizations use AI responsibly and effectively in their customer interactions, with solutions built from proven modules that eliminate the need for costly customization.
What we can do for you:
- Intelligent AI assistants that introduce themselves properly and seamlessly hand off to a human agent when necessary
- An omnichannel setup where customer context is always carried along, so customers never have to repeat their story
- Agentic AI that goes beyond simply following instructions: our Agentic AI for customer service takes the initiative on its own and acts proactively, marking an evolution from executive bots to self-thinking assistants
- Everything under one roof: from implementation to management and support, without having to manage multiple vendors
- Compliance-ready solutions that take into account the requirements of the EU AI Act and related regulations
Would you like to know how to implement AI transparency in your organization? Contact us, and we’d be happy to help you figure it out.
Frequently Asked Questions
Geldt de transparantieplicht ook voor AI die intern wordt gebruikt, of alleen voor klantgerichte systemen?
De transparantieverplichtingen uit de EU AI Act richten zich primair op AI-systemen die direct met mensen communiceren, zoals chatbots en virtuele assistenten in klantcontact. Voor intern gebruik gelden andere regels, afhankelijk van de risicocategorie van het systeem. Denk aan AI die wordt ingezet bij personeelsbeslissingen of prestatiebeoordeling — die vallen vaak onder de hoog-risico categorie en kennen eigen verplichtingen. Raadpleeg bij twijfel een juridisch adviseur die gespecialiseerd is in AI-regelgeving.
Wat zijn de gevolgen als je de AI-transparantieplicht niet naleeft?
Niet-naleving van de EU AI Act kan leiden tot aanzienlijke boetes: tot 15 miljoen euro of 3% van de wereldwijde jaaromzet voor overtredingen van de transparantieverplichtingen, afhankelijk van welk bedrag hoger is. Daarnaast loop je reputatieschade op als klanten ontdekken dat ze ongemerkt met AI communiceerden. De handhavingsbevoegdheden worden gefaseerd ingevoerd, maar de verplichtingen zelf zijn al van kracht — wachten op actieve handhaving is dan ook een risico dat je beter kunt vermijden.
Hoe pas je de AI-melding aan voor verschillende kanalen, zoals telefoon versus chat?
De kern van de melding blijft hetzelfde — duidelijk, vriendelijk en aan het begin van het contact — maar de vorm past je aan het kanaal aan. In een chat volstaat een korte openingsboodschap zoals: ‘Hallo! Je spreekt met onze AI-assistent.’ Bij telefonisch contact werkt een ingesproken begroeting het beste, bij voorkeur in een natuurlijke stem die aansluit bij de toon van je merk. Zorg er in alle gevallen voor dat de melding niet verloren gaat in een lange welkomsttekst of IVR-menu met meerdere opties.
Moet je de AI-melding ook opnemen in je privacyverklaring of algemene voorwaarden?
Ja, het is aan te raden om het gebruik van AI ook te vermelden in je privacyverklaring, zeker wanneer de AI persoonsgegevens verwerkt of betrokken is bij geautomatiseerde besluitvorming. Dit is een aanvullende verplichting naast de actieve melding aan het begin van het klantcontact — de privacyverklaring vervangt die directe melding niet. Onder de AVG/GDPR ben je bovendien verplicht transparant te zijn over hoe je persoonsgegevens verwerkt, inclusief de rol die AI daarin speelt.
Hoe meet je of je AI-melding effectief is en goed wordt ontvangen door klanten?
Een eenvoudige manier is het monitoren van het aantal klanten dat direct na de melding afhaakt of vraagt om een menselijke medewerker — een hoog percentage kan wijzen op een melding die onnodig alarmerend overkomt. Daarnaast kun je klanttevredenheidsscores (CSAT) vergelijken tussen AI-gesprekken en menselijke gesprekken, en gerichte vragen opnemen in je klanttevredenheidsonderzoek over de ervaren transparantie. A/B-testen met verschillende formuleringen geeft je concrete data over welke aanpak het beste werkt voor jouw doelgroep.
Wat doe je als een klant principieel bezwaar maakt tegen communicatie met AI?
Wanneer een klant expliciet aangeeft niet met een AI te willen communiceren, moet je die voorkeur respecteren en direct doorschakelen naar een menselijke medewerker. Dit is niet alleen een kwestie van klantvriendelijkheid, maar sluit ook aan bij de geest van de EU AI Act, die menselijke autonomie en keuzevrijheid centraal stelt. Zorg er technisch voor dat deze overdracht soepel verloopt en dat de medewerker de beschikbare gesprekscontext ontvangt, zodat de klant niet opnieuw zijn verhaal hoeft te doen.
Hoe ga je om met meertalige klanten als het gaat om de AI-melding?
De AI-melding moet begrijpelijk zijn voor de klant, wat betekent dat je de melding idealiter aanbiedt in de taal van het kanaal of de voorkeurstaal van de klant. Bij omnichannel systemen kun je dit automatiseren op basis van taaldetectie of de eerder opgegeven klantvoorkeur. Een melding in een taal die de klant niet begrijpt, voldoet niet aan de transparantieverplichting — begrijpelijkheid is een expliciet vereiste onder de EU AI Act.
