§ 01 · The Report

What Deloitte actually surveyed, and why I read it.

A confession, before we go further. I am a designer. The contents of a Deloitte industry report are not, traditionally, the kind of reading material I take to bed with a cup of tea. The reason I have spent a month with this particular report is that what restaurants are doing with AI has begun to filter into the menu-design conversation in ways I cannot ignore. Voice AI in drive-thrus is rewriting how the spoken menu works. Computer vision is starting to monitor whether the plated dish looks anything like its photograph. Generative AI is producing the background graphics on a noticeable share of new menu templates being uploaded to the marketplaces I write about. The technology is arriving in the dining room, and the people I write for — small restaurant owners, cafe operators, the people who design their own materials — deserve a careful read of what the industry leaders are actually saying.

Deloitte’s State of AI in Restaurants Survey ran in the fourth quarter of 2024 and polled 375 restaurant executives across 11 countries. The sample skews large — 93 per cent of respondents work at organisations with at least 1,000 employees, and more than a third come from companies with 2,000 or more locations. That is worth registering. The view you are about to read is the view from the C-suite of the big chains, not from the independent bistro down the road. That biases the findings in specific directions. It also makes the data useful in a particular way: it tells you, with some precision, what the largest restaurant operators in the world are about to do, which determines what the rest of the industry will be reacting to over the next three to five years.

The report’s top-line finding is straightforward. Eight in ten restaurant executives plan to increase their AI investment in the next fiscal year, with nine per cent expecting to increase it significantly. Only two per cent forecast a decrease. The expected benefits, in order of how often executives cited them, are improved customer experience, smoother operations, more impactful loyalty programmes, and improvements to procurement and supply chain. The technology is, by the industry’s own internal account, being treated less as a strategic experiment and more as standard infrastructure. That shift in framing matters, and the rest of this piece is mostly about why.

§ 02 · The Three Waves

Adoption is happening in waves. We are firmly in the first one.

The most useful structural framing in the report is that AI adoption in restaurants is unfolding in three distinct waves rather than as a single front. This matches what I see when I talk to restaurant operators — the technology is not arriving everywhere at once. It is arriving in customer experience first, then in operations, then in product development and food preparation. Understanding this sequence is the difference between a smart investment in 2026 and an expensive one.

The first wave is already well underway. Sixty-three per cent of respondents are using AI in customer experience daily, and another 26 per cent are running pilots. That is roughly nine in ten organisations either using or testing AI to influence the moment of customer contact — the recommendation engine in the ordering kiosk, the chatbot on the website, the personalisation logic in the loyalty app, the voice AI in the drive-thru. Fifty-five per cent are using AI for inventory management every day, with another 25 per cent piloting. These are the two areas where the return on AI investment is now plainly visible to operators, which is why they are also the two areas the survey shows the most aggressive forward commitment.

§ 03 · The Second Wave

Loyalty and employee experience are next, and the timing is significant.

The second wave the report identifies covers customer loyalty programmes and employee experience. Combined daily-use and pilot rates hover near 70 per cent for both applications, with substantial planning behind that. If current plans come to fruition, the report notes, AI use in customer loyalty and employee experience could eventually surpass that in inventory management. That is the sentence I underlined twice.

It is worth pausing on what AI in loyalty actually means. Most loyalty programmes today use machine learning to figure out which customers are about to stop coming and which offers will bring them back. The next generation, which the survey shows is being actively built, will personalise the actual menu the customer sees in the app to their previous order history, dietary preferences, time of day, and probable mood inferred from their recent visit pattern. That is a non-trivial shift. A loyalty programme that adjusts your visible menu is no longer a discount programme. It is, functionally, the chain restaurant’s version of having a personal waiter who knows what you usually order.

Employee experience is the application I think will receive the least attention publicly and the most attention internally over the next two years. The report’s data shows brand owners reporting that this is where they are seeing meaningful operational impact — AI for shift scheduling, AI for training and onboarding, AI for handling the routine questions that previously took up so much of a duty manager’s time. None of this generates a press release. All of it shows up on the bottom line. If you wanted to predict where AI investment quietly compounds in this industry, my bet would be here.

The third wave covers food preparation and new product development — computer vision for plating consistency, machine learning for menu engineering, AI for flavour compound analysis when developing new dishes. Daily use is under 50 per cent today for both. But these two areas had the highest planning-and-development readings of anywhere in the survey, which means the slope is steep. By 2028, I would expect the wave numbers to look quite different.

§ 04 · The Technology Stack

Chatbots are leading. Generative AI, despite the noise, is barely deployed.

When the report disaggregates the AI conversation by underlying technology rather than by use case, an interesting and slightly surprising picture emerges. Chatbots top the list of daily-use technologies, with 60 per cent of respondents reporting daily deployment and another 27 per cent in pilot. Machine learning is being used daily by 54 per cent. Intelligent automation and natural language processing round out the top four. These are well-established AI technologies with a decade or more of commercial deployment behind them. The restaurant industry is, contrary to the headline narrative, not pioneering the technology — it is finally adopting it at scale.

The next tier is more experimental. Conversational voice AI, computer vision, and deep learning show more respondents in the pilot phase than in daily use, indicating these technologies are being actively tested but have not yet crossed the threshold into routine deployment. Voice AI specifically is being trialled in kiosks and drive-thrus — the order-taking automation that occasionally goes viral when a chain experiments with it. Computer vision is being tested for order accuracy and food anomaly detection.

And then, at the bottom of the daily-use list, something I want to flag specifically: generative AI is deployed daily by only nine per cent of respondents. Nine per cent. After roughly three years of the most aggressive technology hype cycle of my professional lifetime, the technology that occupies most of the column inches and most of the LinkedIn discourse is the technology fewest restaurants are actually using in production. More respondents reported planning and development around generative AI than current deployment, which is the corporate way of saying “we are looking at it, but we have not committed to it.” That gap between hype and deployment is, I think, the most useful single finding in the entire survey. It tells you that this industry, for all its public enthusiasm about AI, is making a deliberate distinction between the AI that earns its keep and the AI that earns its mentions.

§ 05 · The Readiness Gap

Nobody feels prepared for what they are about to spend on.

This is the part of the report I found most striking. Eight in ten executives plan to increase their AI investment in the coming year. Roughly two in ten feel their organisation has the risk and governance in place to handle that investment responsibly. Less than three in ten feel prepared from a technology infrastructure standpoint. Less than three in ten feel they have the talent. Strategy is the only area where most respondents agree their companies are adequately prepared, and even there, almost 40 per cent say they do not have a strong strategy in place. That is a substantial gap between the speed of capital deployment and the readiness of the organisations doing the deploying.

I am not, generally, a person who worries about whether corporations are spending their money sensibly. They will or they will not, and the market will correct them either way. What concerns me as a writer for the menu-design audience is the downstream effect of this investment-readiness gap. When a large restaurant chain deploys an AI-driven personalised menu system without adequate governance, the failure modes show up in the customer experience — menus that recommend dishes the customer is allergic to, loyalty programmes that quietly raise prices on regulars who never check, voice AI in drive-thrus that misunderstand half the orders. The chain is large enough to absorb the reputational damage. The independent restaurant down the road, which copies the trend two years later without the underlying technical infrastructure, is not.

The geographical split in readiness is interesting. Asian companies report higher readiness across nearly every dimension — strategy, operations, infrastructure, talent — than their US or European counterparts. This is consistent with what the wider technology press reports about the maturity of restaurant AI in the larger Asian markets, particularly in China and South Korea. American restaurants are catching up. European restaurants are noticeably behind. That ordering is worth keeping in mind when reading any global AI restaurant story.

§ 06 · The Challenges

What is actually holding the industry back. (It is not what the press says.)

When asked to identify the top challenges to AI implementation, the surveyed executives gave a slightly counterintuitive set of answers. The biggest constraint reported is the difficulty in identifying the right use cases. Not a shortage of ideas. A shortage of ideas that scale and demonstrably create business value. This is the kind of problem that gets glossed over in the public AI discourse, which is dominated by demonstrations of what AI can do rather than analyses of which uses repay the investment. The chief technology officer of a 200-location restaurant chain does not need to be convinced that AI can write a recipe. She needs to know whether the AI-written recipe is going to sell more covers than the human-written one, and whether the additional sales offset the cost of the system, and whether the operational complexity is worth the marginal improvement.

Risk management is the next biggest concern, and the one operators (as distinct from brands) rate as their number-one worry. The risks the survey specifically flags include intellectual property leakage, customer data misuse, regulatory non-compliance, and the absence of a clear governance model for AI decisions. None of these are restaurant-specific problems. All of them are problems any organisation deploying AI at scale runs into, and most of which are unsolved at the corporate level across the wider economy. The restaurant industry is, in this sense, just experiencing the same growing pains as banking, healthcare, and retail. It happens to be more visible because customers interact with restaurants several times a week.

Talent shortage is the third major concern, particularly for brand owners building the underlying systems. This one I expect to ease over the next three to five years as the supply of AI engineers and data scientists in the restaurant sector specifically catches up with demand. The bottleneck is not unique to restaurants. It is simply that restaurants pay less than financial services and have to wait their turn.

What is conspicuously not on the executives’ list of top concerns: lack of executive commitment, choosing the right technologies, and computing infrastructure. The leadership conversation about whether to invest in AI is over. The conversation about how to invest, on what, and with what safeguards has barely begun.

§ 07 · The Menu Question

What this all means for the menu in your hand.

I want to pull the analysis back to the territory I actually know about, which is the menu — the physical or digital object the customer is holding when they decide what to eat. The Deloitte report does not address menu design directly, but its findings have specific and substantial implications for menus, and I want to spell them out.

First. The static printed menu is not going away. The chain restaurant industry is rapidly investing in personalised digital ordering, but the printed menu remains the touchpoint at the table for the entire mid-market and the bulk of independent restaurants. None of the trends in the report change that for the foreseeable future. The cafe owner in Hackney is not, in 2026 or 2028, going to abandon her printed menu in favour of an AI-generated one.

Second. The expectations customers bring to the printed menu are quietly being reshaped by their experiences with the digital one. If a customer is used to ordering from a chain where the loyalty app remembers their previous orders, knows their dietary restrictions, and suggests dishes within their usual price range, they arrive at the independent bistro with a different baseline of what a menu can do. The independent bistro cannot match the personalisation. But the independent bistro can compensate with menu design that is distinctive, well-written, and treats the customer as an intelligent adult rather than as a data point. The premium on craft, in other words, rises as the alternative becomes more algorithmic.

Third. The AI tools that genuinely affect small-restaurant menu design are not the ones generating attention in the report. They are the practical, lower-glamour applications — AI image generators for decorative graphics, AI writing assistance for cleaning up menu descriptions, AI translation for menus in multiple languages, AI layout assistance for testing different arrangements quickly. The high-end applications — computer vision for plating consistency, predictive ordering, voice AI — are operationally interesting but do not touch the menu itself. A small restaurant should pay attention to the boring applications and ignore the exciting ones.

Fourth. The chain restaurants are about to make a substantial number of well-funded mistakes, and the failure stories will dominate the trade press for the next two years. This is not pessimism; it is what large enterprises do when they deploy capital at speed against organisational readiness gaps the report itself documents. The independent restaurants that watch these mistakes carefully, learn from them, and avoid making the same ones at smaller scale will be the ones who emerge from this period with stronger businesses. The independent restaurants that ignore the trend entirely will not.

§ 08 · A Closing Thought

What I am still thinking about, a month after reading.

The most useful thing I took away from a month with this report is a clearer sense of which bets are being made, by whom, with how much conviction, and with what level of preparation. The headline story — restaurants are spending more on AI — is accurate but boring. The interesting story underneath is more textured. It is a story of investment racing ahead of readiness, of well-established AI technologies being adopted at scale while the speculative ones generate noise without proportional deployment, of geographical divergence with Asia ahead and Europe behind, and of a clear sequence of where the value is appearing first.

The thing I keep returning to is the gap between the 82 per cent of executives planning to spend more and the 20 per cent who feel ready on risk and governance. That gap is, almost by definition, where the next round of cautionary tales will originate. Chain restaurants will deploy AI systems that misfire in customer-facing ways. The misfires will generate press coverage. The press coverage will create a public narrative that AI in restaurants is dangerous, unreliable, or untrustworthy — which it sometimes will be, in specific deployments, while remaining genuinely useful in others. Navigating that narrative without overcorrecting in either direction is the small task ahead for anyone writing about this space.

What I am most hopeful about — and I mean this sincerely, not as the obligatory upbeat closer of an industry report — is that the smallest businesses, the cafe owners and the bistro proprietors and the village-pub landlords, may well end up benefiting most from this transition. The AI tools they adopt are cheaper, simpler, and lower-risk than the enterprise stacks the chains are buying. The competitive advantage they retain — care, taste, judgement, the welcome at the door — is precisely the thing the chains are quietly admitting they cannot replicate at scale. The chains have the data and the technology. The small operator has the craft. Both will survive. They will become more distinct from each other, not less.

Margot Ellery

Editor · Printable Menu Lab