Will your customers talk to an AI phone agent? What diners want

Many of your restaurant's customers are open to an AI phone agent, and what they ask in return is simple: say it's an AI, and let them reach a person. In a Harris Poll survey for TouchBistro, 41% of U.S. diners said they were comfortable ordering by phone from an AI, and 32% said they were not [1].

The rest of the research agrees: diners want transparency and a way out. In the European Union, telling callers they're talking to an AI is now also the law. Here are the surveys, the rule, and a greeting that covers both.

Do customers like AI phone ordering?

Many do, though far from all. TouchBistro's 2025 U.S. Diner Trends Report, based on a Harris Poll survey of 1,500 U.S. diners in October 2024, asked how comfortable people are with specific uses of AI [1]:

Comfort is rising. In PAR Technology's 2025 survey, 41% of consumers said they would choose a restaurant that avoided AI altogether; in its June 2026 survey of 1,000 U.S. adults, average discomfort with restaurant AI had fallen to 27% [2]. HungerRush, which surveyed 1,000 U.S. diners with Dynata about AI in quick-service and fast-casual restaurants, found that 54% are comfortable with restaurants using AI to improve convenience, accuracy and speed [3].

Few restaurants have taken the step, though: 26% of operators use AI tools of some kind, but only 6% use AI for customer orders, according to the National Restaurant Association's State of the Restaurant Industry 2026 report [4]. For many callers, yours may be the first AI they order from.

One caveat: these are stated attitudes, and TouchBistro, PAR and HungerRush all sell restaurant technology. Read the numbers as a trend, not a guarantee.

What makes diners trust an AI on the phone?

Two conditions keep coming back: knowing it's an AI, and being able to reach a person.

• Transparency first. 43% of U.S. consumers rank transparency as the top factor in trusting AI at restaurants, nearly three times any other factor [2].
• A human within reach. 87% of diners say being able to connect with a human staff member is important [3].

The takeaway: diners don't reject AI, they reject being fooled or trapped. An agent that says it's an AI from the first sentence, and hands off to your team when needed, gives them exactly what they ask for.

What does the EU AI Act require from an AI phone agent?

If your restaurant operates in the European Union, disclosure is no longer optional. Article 50 of the EU AI Act requires AI systems that interact directly with people to be designed so that those people “are informed that they are interacting with an AI system” [5]. The European Commission lists AI voice assistants used in customer support among the systems covered [6].

• When: “in a clear and distinguishable manner at the latest at the time of the first interaction” [5]. On the phone, that means in the greeting.
• Since when: August 2, 2026 (Article 113) [5]. The EU's Digital Omnibus on AI, in force since July 27, 2026, doesn't delay it; only the machine-readable marking of AI-generated content gets extra time, until December 2, 2026, for systems already on the market [6] [7].
• Who: the duty sits with the provider, the company that develops the agent and markets it under its own name. The restaurant using it is a “deployer.” If you have a custom agent built and put it into service under your own name, you can become the provider [5].
• Fines: up to €15 million or 3% of worldwide annual turnover, whichever is higher; for small and medium-sized businesses, whichever is lower [5].

Can an agent skip the disclosure when it's “obvious”?

The law exempts cases where the AI is obvious “from the point of view of a natural person who is reasonably well-informed, observant and circumspect” [5]. Don't build on it. The Commission's July 20, 2026 guidelines, non-binding but written for enforcers too, say the exception must be read narrowly, for cases with almost no doubt left; that a human-sounding voice makes AI less obvious; and that it can't be relied on when the public includes vulnerable people such as children or the elderly [6]. A restaurant line takes calls from everyone.

European diners agree with the law. In France, 40% of consumers say they would feel deceived if a company told them only afterward that they had been talking to an AI, and 32% say it is unacceptable and want to know from the start; only 27% say it doesn't matter (Ipsos bva, 1,000 people, November 2025) [8].

This is a summary, not legal advice. For your specific situation, ask your vendor or a lawyer.

What should an AI phone agent say in the first 5 seconds?

One sentence does it, if it has four parts: your restaurant's name, the words “AI” or “artificial intelligence,” what the agent can do, and the way out. For example:

“Hi, you've reached [Restaurant Name]. I'm the restaurant's AI voice assistant. I can take your order for pickup or delivery, and if you'd rather have someone from the team call you back, just say so.”

What doesn't count, according to the Commission's guidelines [6]:

• the word “assistant” on its own, which is too vague;
• a disclosure buried in your terms or on your website;
• a beep or jingle by itself (fine as an extra, not as the disclosure);
• a persona that suggests a human, like a first name with no mention of AI.

If a caller asks “Am I talking to a robot?”, the agent says yes, plainly: the guidelines expect an AI to disclose its nature whenever it's asked [6]. For an order call of a few minutes, one clear announcement at the start is usually enough; a reminder helps if the call drags on or changes course [6].

That's the approach Soleir takes: the agent says it's an AI at the start of every call.

How do you keep diners' trust with an AI phone agent?

These habits work everywhere, in the EU or not, and answer the two complaints surveys keep surfacing: AI that hides, and AI that blocks.

1. Announce it. In the greeting, with the word “AI,” and never deny it later.
2. Keep a way out. Transfer to staff when someone is free; otherwise, take a message for a quick callback. For an upset caller, a tricky case or an off-menu request, the agent hands over.
3. Never make things up. A price, a wait time, an ingredient: if it isn't in the menu data, the agent says so instead of guessing. That matters most for allergens, where a guess can put a guest in danger.
4. Pass on what it doesn't know. The open question travels with the order, so your team can call back or note it on the ticket.
5. Respect personal data. Ask only for what the order needs (name, phone number, delivery address), say if calls are recorded, and know where the data goes. In the EU, the AI Act adds to data protection rules; it doesn't replace them [6].

Then test it yourself. Call your own number at a quiet hour and again during the rush. Listen to the greeting, ask “Are you a robot?”, ask for a person, throw in a tricky menu question. Five minutes will tell you whether your agent is honest. Still deciding whether an agent is worth it? Start with what missed calls really cost.