I was on a shuttle bus in Chicago last month, headed from my hotel downtown to the National Restaurant Show, when I began eavesdropping on a group of people sitting behind me. They were making plans for a steak dinner that night, and it had become a bit of a juggling act.
They were cross-referencing prices, menus and reservation times, trying to find a place everyone could agree on, while also triangulating the distance between the various options and their hotel. The whole process involved at least two phones and took what felt like the entire half-hour ride to McCormick Place.
It was maybe the first time I truly thought to myself: “AI could probably have done this.”
And it most definitely could have.
In January, OpenAI, the maker of ChatGPT, unveiled a tool called Operator that can perform online tasks, including booking a restaurant reservation. It can also order food from delivery apps, or find a recipe and then order the ingredients to make it.
It’s just one of a growing army of AI “agents” that aim to help people automate and optimize their daily lives. And they could have big implications for how people use restaurants.
Imagine those people on the shuttle had used an AI agent to find a steakhouse and book their reservation. They get some time back to do other things. And a restaurant that the AI deemed to be a good fit gets some business. There’s a chance it’s a different restaurant than what the group would have picked on their own by Googling. In the future, restaurants may need to think about how to optimize themselves to be found by AI, rather than SEO.
Google, unsurprisingly, is developing its own AI agents, such as Project Mariner, which can handle web-based tasks like restaurant reservations. If these agents work, they could push Search to the wayside.
Now let’s say this group uses an agent to order dinner from a delivery service instead. The service, and the restaurant, still get the business, but it’s originating from a different place. This could start to take share from other ordering modes, like mobile apps and restaurant websites, eventually making those channels obsolete. I’d expect tech companies that specialize in those products to start looking for ways to integrate agentic AI into their software.
To be clear, these tools are still in their early stages. In demos of OpenAI’s Operator, the agent seems to move about as fast as a human browsing the web, if not slower. The user may still need to log in to websites or enter payment information, so it’s not entirely hands-free. Also: It’s only available to people who shell out $200 a month for a ChatGPT Pro subscription, so it’s far from being mainstream.
At the same time, its current abilities represent just the tip of the iceberg in terms of how AI agents could reshape restaurant ordering. Beyond automating rote tasks, agents will soon be able to learn customers’ likes and dislikes, make recommendations, and even go ahead and buy things for them.
That’s the essence of cutting-edge new initiatives from credit card giants Visa and MasterCard, both of which have rolled out new programs that claim to allow AI to shop for you.
With Visa’s Intelligent Commerce, for instance, customers can share their spending behavior and set guidelines and spending limits for an AI agent that is deputized to make purchases.
Jack Forestell, the company’s chief product and strategy officer, described it as the next step in the evolution of shopping, from physical stores to websites and mobile apps. “Now, with Visa Intelligent Commerce, AI agents can find, shop and buy for consumers based on their pre-selected preferences,” he said in a statement.
Mastercard’s Agent Pay has a similar premise. In a press release, the company gave the example of a “soon-to-be-30-year-old” using the agent to help plan a birthday party, with outfits, accessories and ambience all curated—and purchased—by the AI. (There’s no mention of food, but what kind of party doesn’t have that?)
It’s fascinating to think about how this form of agentic AI might work in a restaurant context. Will agents be able to order your usual for lunch without being asked? Make reservations ahead of your anniversary? Buy you a coffee when it knows you were out late? Suggest a nice healthy salad after a weekend junk food binge?
If the answer to those questions is yes, it only raises more questions about how AI agents might impact restaurants. For one, they would seem to be naturally biased toward takeout and delivery, rather than in-person dining, which could pave the way for even more off-premise business.
They would also appear to be skewed by people’s existing patterns and behaviors. Will agents be able to help us discover new restaurants or cuisines, or anticipate how our tastes will evolve over time? Or will they just continue feeding us what we’ve enjoyed in the past?
And another issue not to be overlooked: Who owns the data on orders mediated by AI agents?
One thing that seems clear is that the way we discover restaurants will change. Today, consumers Google “wings near me” or scroll through delivery apps. Tomorrow, an AI agent may do that sifting for them. How will it decide?
For now, we have a lot more questions than answers. But because of the pace at which AI is moving, the answers may get here sooner than we think. Marc Lore, the billionaire founder of the food delivery app Wonder, has claimed that AI is already choosing 85% to 90% of his meals. His goal with Wonder is to bring that same technology to the masses.
Then again, restaurants have always had a higher level of immunity to the latest technology trends. At its core, it is a people-driven industry, and it takes a lot of pride in that. (Though it has been surprisingly quick to embrace AI.)
Like the tech that came before it, AI agents will probably make some things more convenient for consumers, like picking a restaurant or placing an order. It may overtake some existing technologies. But there will be parts of the dining experience it will never be able to replace.
Like, say, arguing with your family about which Chicago steakhouse to go to for dinner.