The Kitchen That Thinks It Knows Best

This is the second in a short series on AI in commercial foodservice. The first looked at whether AI can design a commercial kitchen. This one looks at what the technology does once the kitchen is built and what consultants need to understand about it.

Mention AI or automation in hospitality and what do you picture? Most people still picture robot waiters. Some have no idea it exists within kitchens at all and see it simply as the thing that’s made everyone’s LinkedIn posts sound the same. I’ve been guilty of this too.

In reality, most AI in foodservice is less controversial than the headlines and more useful than the memes. Sensors, dashboards, scheduling systems, waste tracking software, connected equipment. None of this is a thing of the future. Many larger or more systemised operations are already using parts of it, often without calling it AI at all.

The anxiety I encounter most is that it’s going to take over. That human skill and interaction will be slowly automated out of existence. This blog looks at what actually exists, what it does well, and where it falls short. But the more interesting question underneath all of it is this: what happens when hospitality starts being managed solely like a system rather than experienced as a human environment?

The Kitchen That Knows Everything

There is a site manager I met who oversees three hundred kitchens. He told me about a waste monitoring system his organisation had rolled out. Scales with AI cameras built in, filming everything that lands on them. The system tells him that site forty-seven wasted a measurable quantity of cucumber on Tuesday.

He didn’t tell me this to celebrate the technology. He told me to make a point about what he does with that information. He doesn’t send a memo. He doesn’t adjust the ordering algorithm and move on. He gets in the car, or picks up the phone, and has a conversation.

Because what the data tells him is what happened. What he needs to know is why. Was the chef exhausted? Was the menu wrong? Did someone over-order because they were covering for a colleague and unfamiliar with the site?

The system cannot tell him any of that. That still requires a human.

That distinction between data and understanding is the most important thing I can say about AI in the commercial kitchen.

We Have Been Here Before

This is not the first time a piece of equipment has been accused of threatening the profession.

The combi oven, now so embedded in professional kitchens that many chefs would sooner give up their knives, was initially regarded with deep suspicion. It was going to automate cooking. It was going to deskill the kitchen. It didn’t. It became the tool they can’t live without. The one that handles the repetitive, the precise, and the time-sensitive, and frees up the chef to focus on the things that actually require a cook.

Bee Wilson makes a similar point in Consider the Fork, tracing kitchen tools through history as a series of moments where new technology was feared, dismissed, and then absorbed into the everyday (including even the fork itself). The pattern of anxiety followed by normalisation is remarkably predictable. Recognising it should make us more measured about what AI will and won’t do.

What AI in Foodservice Looks Like In Reality

A brief tour of the landscape: what exists, where it is genuinely useful, and why consultants increasingly need to understand it at design and specification stage.

This is not a comprehensive list. Increasingly, some form of automation, connectivity, data analysis or AI-assisted functionality is appearing across almost every category of commercial catering equipment. The lines between automation, smart controls and AI are also becoming increasingly blurred. What follows is simply a snapshot of some of the technologies currently shaping operational kitchens and hospitality spaces.

Waste monitoring Systems

Camera-and-scale systems log food waste at the point of disposal. The data is often revelatory. It can also be misleading as it does sometimes get it wrong, and the technology still needs refinement. More immediately, it changes how consultants think about the waste station which is no longer an afterthought. Scale placement, sightlines, and staff workflow need to be considered at design stage, not retrofitted.

Connected Cooking Equipment

Some modern combi ovens can adjust cooking parameters in real time using sensors, store and distribute recipes across multiple sites, and allow remote monitoring of performance and HACCP data. This is largely mature technology and has been developing for years. Similar concerns once surrounded combi ovens themselves, with suggestions they would replace chefs. In reality, they became another professional tool: improving consistency, reducing manual intervention, and changing the role rather than removing it.

Much of what is currently described as “AI” in the commercial kitchen is often better understood as connected equipment, automation, or data-led control systems. The language has become blurred: partly technical, partly marketing.

Traditional combi logic already uses responsive automation extensively. Sensors monitor humidity, temperature, probe readings, load size and cooking conditions, allowing the oven to adjust steam levels, fan speed or cooking time automatically. Connected platforms then build on this further through recipe distribution, remote monitoring, predictive maintenance and operational analytics across multiple sites.

Equipment is also increasingly able to self-monitor performance, predict maintenance requirements, and alert service teams before faults become operational failures. That changes not only maintenance strategy and repair costs, but also the relationship between operators and manufacturers themselves.

Increasingly, connected systems are interacting with wider building infrastructure too, coordinating energy use, extraction demand and operational loads in ways that affect both sustainability strategy and utility planning.

Genuine AI-led systems are now beginning to emerge in some areas of foodservice technology, particularly through computer vision and camera recognition. Certain systems can identify products, analyse browning, recognise tray positioning or optimise cooking processes using image recognition and learned datasets.

Specification decisions are no longer purely physical or mechanical. Consultants are now also assessing software ecosystems, connectivity requirements, data ownership, interoperability and long-term support models alongside traditional utility and operational considerations.

Demand-Controlled Ventilation

Traditional kitchen ventilation runs at fixed capacity throughout service regardless of cooking load. Demand-controlled systems modulate fan speed based on actual activity, with meaningful reductions in energy consumption. On projects with sustainability targets or constrained electrical capacity, this belongs in the briefing conversation early. It affects utility strategy, running costs, and for BREEAM or LEED assessed projects, feeds into broader energy modelling. The consultant’s job is to know it exists before the M&E engineer raises it.

Compliance and HACCP Logging

Probably the least glamorous application of connected kitchen technology and one of the most genuinely useful. Some equipment now logs temperature data automatically, creating a digital HACCP trail without staff having to record anything manually. For a school kitchen, a hospital, or any operation with serious food safety obligations and a stretched team, this is not a gimmick. It removes a daily administrative burden and creates a reliable audit trail that paper systems cannot match. Nobody is losing their job to a temperature probe. But somebody is getting an hour of their day back.

Forecasting & Scheduling

AI systems can analyse operational data and advise on staffing, ordering, prep levels, GP, and queue patterns. They can compare historical sales against weather forecasts, local events, and footfall trends. Useful in the right hands. But the system only knows what you tell it. It needs to know about the field trip, the football match, the road closure, the fact that the café manager was on leave.

We had a project recently where this went wrong in a way that looked entirely right on paper. A workplace operation with poor performance had implemented AI-driven reporting. The outputs were convincing, stated with confidence. But nothing was being connected to anything else. Poor GP wasn’t linked to waste. High staffing levels weren’t linked to poor design so when staffing was cut, the operation could no longer produce fresh food. Handmade cakes disappeared. Local suppliers were replaced by centralised distribution. Rising sales of drinks and packaged goods were read as customer preference, when customers actually wanted fresh food that was no longer available. Sales fell. Waste rose. The reports weren’t technically wrong. They were simply incomplete.

That is not a cautionary tale about AI. It is a cautionary tale about what happens when data is interpreted without operational understanding.

Where This Technology Earns Its Place

Here is the thing the broader conversation tends to skip over: this technology is not equally useful everywhere, and knowing where it belongs is arguably the most important judgement a consultant can make.

For a restaurant selling an experience (atmosphere, personality, the feeling of being looked after) the human element is not a problem to be optimised. It is the product.

For a school kitchen feeding eight hundred children a day on a fixed budget with a small team, the calculation is different. The margins are thin, the compliance obligations are real, the skill base is often stretched, and the consequences of misjudged prep or waste are felt immediately. The same is true of hospitals, large contract catering operations, and high-volume workplace sites. These are environments where pattern recognition, automated compliance logging, and demand-responsive systems do their best work. This isn’t because the human element doesn’t matter, but because the operational pressure is high enough that the technology genuinely helps carry the load.

The mistake isn’t using AI or automation in foodservice. The mistake is assuming it belongs everywhere in equal measure. A consultant who specifies the same technology approach for a heritage café and a hospital kitchen isn’t demonstrating expertise, they’re avoiding the harder question of what the site actually needs.

Front Of House: Where Optimisation Meets the Guest

I was at a meal abroad recently. We asked for the menu and were handed a QR code. There was a collective sigh around the table. Not outrage. Disappointment.

We were there to be somewhere together, and the first thing we were asked to do was all sit on our phones.

The food was extraordinary. All was forgiven. But that moment I remembered was being handed the QR code.

This is not an argument against QR codes. There are contexts where they are genuinely the right solution. Fast turnaround. High volume. Customers who actively want speed and convenience.

The problem is the assumption that because something can be optimised, it should be optimised everywhere.

Eating out is increasingly one of the few remaining occasions where people interact with each other without a screen between them. That has value, and not every site should be in a hurry to eliminate it.

Kiosk ordering follows similar logic. The psychological satisfaction of watching each step of your order progress on a screen is real. But context matters. In environments such as garden centres, hospital cafés or supermarket cafés, customers are often distracted, carrying bags, managing children, less digitally confident, or simply looking for a slower and more human interaction. In those settings, the technology does not disappear into the background; it risks becoming another layer of friction.

There is also the question of hospitality itself. A genuinely well-timed recommendation from a person “the sourdough today is exceptional” still feels fundamentally different from a machine-generated upsell prompt.

Customers also tend to forgive human error far more easily than technological failure. When a member of staff makes a mistake, people often respond with patience. When a self-service kiosk fails, frustration becomes immediate and oddly disproportionate. Nobody is angry with a person. They are angry with the entire premise.

The Point Where Optimisation Starts Fighting Hospitality

A kitchen can become operationally efficient while simultaneously becoming worse for staff, for customers, and eventually for the business.

Labour algorithms that cut staffing too hard. Waste monitoring that becomes punitive. Performance dashboards that create surveillance culture rather than operational support. Over-standardisation that removes the autonomy keeping talented people in the building. These are not arguments against technology. They are failures of judgement in how it is applied.

There are also larger questions sitting underneath all of this that deserve more space than an overview allows. What happens to skill development when consistency is automated? Who owns the operational data a connected kitchen generates – the operator, the manufacturer, the software platform? What are the implications when a kitchen becomes dependent on a proprietary ecosystem it can’t easily leave? These are conversations the industry is only beginning to have, and they will matter more as these systems become more embedded.

The Things AI Still Doesn’t Understand

The kitchen may know exactly how much cucumber was wasted on Tuesday.

It still cannot tell you whether the chef was exhausted, whether the atmosphere felt cold, whether the regular customer stopped coming after the menu changed, or whether the café lost the small details that made people feel welcome.

The future of foodservice isn’t human or machine. It is knowing which sites need which balance and having the operational literacy to specify accordingly.

Whether you think AI belongs in hospitality or not is almost beside the point. If you’re specifying commercial kitchens in 2026, you need to understand what these systems do, where they create value, and where they create problems. You don’t have to be an AI cheerleader. But you do have to know what you’re talking about.

Written by Giorgia Lardner