May 27, 2026
I've had three remarkably similar conversations with hospitality CEOs in the last week.
All of them were about to go down a path that was going to lead to 18 months of disappointment and frustration.
In each case, they'd been playing with Claude or ChatGPT and had been blown away by its obvious potential.
In pretty short order, they'd gone from being sceptical about the AI hype to recognising that it's going to be transformative.
These CEOs recognised they can't afford to be left behind on AI and they want to get moving.
But they were all struggling with where to start and how to do it right. And they were all frustrated that people in their head office seemed reluctant to engage with AI and were carrying on like nothing has changed.
Having spoken to these CEOs at length, I thought I'd write up my advice and perspective in the hope other people find it helpful.
So here's what I think are the 10 things hospitality CEOs need to understand about AI.
Mustafa Suleyman, CEO of Microsoft AI and a co-founder of DeepMind, gave an interview to the Financial Times in February where he's quoted as saying:
"Most white collar work, where you're sitting at a computer — a lawyer, an accountant, a project manager, a marketing person — will have most of its tasks fully automated by AI within the next 12 to 18 months."
That was February. So it's now 12–15 months until most of the tasks currently performed by people sitting at a computer in your support office will be done by AI.
This will be a bigger transformation to the way companies work than mobile. Bigger than cloud. Bigger than the internet.
If you're leading a multi-site hospitality business, this isn't something you can delegate to your IT team and check in on next quarter. This is the number one, business-critical challenge facing your company.
However, we're right at the start of this thing. It's just that the technology is moving so fast that a change that would ordinarily take decades to play out may happen in months or years.
At the start of this year, Marc Andreessen said he didn't think the real AI boom had even started yet.
The underlying technology is transformative, but it's not distributed yet. Only a handful of people are using it to its real potential, and we're at the infancy of understanding how to apply it in real business beyond software engineering.
That means hospitality CEOs still have a window of opportunity to be the first mover and gain a genuine competitive advantage. It also means they should be deeply sceptical of anyone telling them they've got it all figured out.
Over the last six months, coding agents have made extraordinary leaps. AI-authored code now accounts for more than a quarter of all production code, and in controlled trials developers have completed coding tasks anywhere from 26% to 56% faster. Generative AI is genuinely transformative for software engineering. That's why you're seeing technology advance so quickly. It's also why you're seeing big tech companies lay off thousands of engineers.
But hospitality isn't software engineering. And this is where I see CEOs making their first major mistake.
Software engineering is a closed problem. You write code, you run a test, it passes or it fails. The training data is vast: billions of lines of open-source code, comprehensive documentation, decades of best practice.
In contrast, the processes and workflows inside a hospitality company are almost entirely undocumented. The problems are open-ended and often don't have a right answer. And the work that creates value tends to require the input of multiple people working across multiple functions.
The most popular use case for AI at the moment in the real world is the "copilot" approach. That could be Microsoft Copilot itself, or another application that's added a copilot feature to help you get more value from the app.
Copilots can be great. But if I was being unkind I'd say they were the "fancy spellcheck, write me an email, produce a presentation for me" mode of generative AI. Genuinely helpful. Saves you time. Not going to change the world.
The risk of treating AI as a personal productivity tool in this context is that you produce an enormous volume of valueless work, and it could seriously damage your organisation.
Imagine your company as a pinball machine for a minute, and the ball represents work moving between different people. Now if one or two of those people — the flippers — suddenly start being 10x or 100x more productive, they'll start firing off more and more work.
What happens then is that you get a bullwhip effect. That work all comes crashing down onto some other poor soul who can't or won't use AI and is now getting hit with 10x more work than they can manage. You'll get bottlenecks and you'll burn people out.
If the underlying workflow is inefficient, or the work isn't adding value, improving individual productivity just creates mountains of useless work. It's performance theatre. Everyone sends 100 times more emails, boasts about how productive they are, and creates zero business value.
As the cost of producing work moves to zero, being purposeful about what you decide to do becomes incredibly important.
At the moment too many CEOs want to use AI to accelerate the pace of work. The real opportunity for hospitality is to use AI to re-engineer or get rid of work.
CEOs have a natural inclination to look for quick wins, and that's a good instinct. But without a long-term vision of how this works, you can build yourself into a cul-de-sac.
It's very easy to build an agent. Anyone can do it.
The first one you deploy will be really well received.
So will the second, third, fourth and fifth.
But when you start operating with 10, 100, 1,000 agents, what starts to matter more and more isn't whether an individual agent does its job, but whether it can be part of a broader, coherent system — where work gets handed over between agents and where they operate with a shared context.
If you haven't designed for that system ahead of time, and you just decide to go after individual quick wins, you're going to reach a breaking point. The system of work will break down, and then you'll have to rip it all out and rebuild from the ground up — this time with the software architecture needed to make it work.
That's an expensive and demoralising experience that's best avoided.
So it helps to build iteratively with one eye on where you're heading.
Here's my view about what the future looks like. I think the future for multi-site hospitality businesses is thousands of people working alongside thousands of agents.
AI isn't going to replace frontline team members. Their roles aren't disappearing. Ideally they're just going to be augmented by agents that make their lives easier and let them focus on what really matters: producing a great customer experience.
But at the multi-site and head office level, things will change very dramatically.
In my mind there are two scenarios.
The first is that hospitality businesses dramatically reduce their head office headcount as they learn how to operate large estates with much, much smaller support teams.
The second scenario is what's known as a Jevons paradox. The technology we expect to replace people actually increases demand for them, because they end up doing higher-value work. This is the scenario I expect to see in the People space. I've been saying to People Directors: "AI agents will be the People equivalent of Excel for Finance teams." When spreadsheets became widespread in the 1980s, the consensus was that lots of finance jobs would disappear. What actually happened is we ended up with roughly the same number of people working in the finance function, but they stopped doing bookkeeping and started doing financial planning, forecasting and analysis instead.
So the scenario you're planning for is this: how do you use AI so head office teams can build "products" that fit into a coherent architecture and are accessible to people on the frontline?
That means everything needs to come back to one user interface on a mobile device.
You need a way of managing who can access what.
And you need to understand which agents work for whom, when and why.
When the operating context is uncertain, it makes sense to align your strategy with the trends that are shaping the economy. In other words, steer your boat to sail with the wind.
I think there are three long-term trends hospitality CEOs should be mindful of when they're thinking through their AI strategy.
First, the economics are shifting. The cost of AI inference is falling and the cost of labour is rising. It's going to make more and more financial sense to replace white-collar manual work with agents.
Second, hyper-personalisation is becoming the baseline. AI will enable personalised experiences for customers and employees at a scale that was previously impossible. Within a few years, people will expect this as standard.
Third, the "Amazon imperative": everyone wants it quicker, cheaper and more convenient. To paraphrase Jeff Bezos: no customer will ever ask for slower delivery, less value or more inconvenience. In hospitality this applies to both frontline employees and customers, and it means you need your agents to know what people want before they do — so they can anticipate needs rather than react to them.
Resist the temptation to go for a moonshot with your first agent.
Instead, focus on using agents to get rid of all the boring, routine stuff that actually eats up everyone's time.
There are two reasons this makes sense. First, automating a well-understood, repetitive workflow with an agent is far easier to make work in the real world. And second, your organisation is about to go through a period of significant change and adjustment. The single most valuable thing you can do right now is create slack in the system by giving people time back. If you can use agents to automate the admin no one really wants to do, you can give people hours back every week. They can then reinvest that time in the work that will actually move the business forward, and they'll have more headspace to explore, play and learn.
The key to making AI agents work in the real world isn't the agents themselves. It's the software wrapper you build around them — what's commonly called the harness.
For hospitality companies, that harness has several core components:
A mobile-first distribution channel for agents. Your frontline team members need one place to interact with every agent. Not 10+ apps. One. And it needs to work on their phone and be fun and easy to use.
Identity and permissioning. The system needs to know who each user is and control what they can see and do via agents.
Long-term context memory and knowledge graphs. Agents need to remember previous interactions and understand the organisational context. Without memory, every interaction starts from scratch like Groundhog Day.
An integration layer. The agents need to connect into your existing tech stack — rota, payroll, PMS, POS, CRM and the rest — so they can take action. What's special about agents is that they can take action. They can do stuff. If you don't connect into the apps you use to run your business, it's not an agent, it's a chatbot.
Orchestration. Something that coordinates multiple agents so they work as a system rather than operating independently.
Observability and audit trails. Which agents are being used, by whom, how often, and whether they're delivering value. It's worth remembering: agents aren't static. Their performance can get better or worse over time.
You could build all this yourself. You might have to build bits of it yourself. But this is a significant amount of software infrastructure, so you probably want tech partners who can help you. (Full disclosure: this is exactly the problem we work on at Youda, so I'm not a neutral observer here.)
But there's one thing you really do need to take control of, and that's…
The core components of AI are the model, compute and data.
As a hospitality company, you're unlikely to be in the business of training your own model. You'll be using one from Anthropic, Google or OpenAI.
You're also unlikely to run the model on your own chips. So you can stop worrying about compute.
Which leaves data.
That's your source of competitive advantage.
All other things being equal, the more of your data you can put through the model, the quicker you'll learn and the better your agents perform. Every interaction, every workflow, every piece of feedback compounds into a smarter system.
My worry is that lots of hospitality companies are going to wake up to the fact that they don't own, or can't access, huge silos of their own data. There are horror stories about tech suppliers holding your data hostage, but the most common issue is that the standard of APIs across the hospitality tech space tends to be pretty poor.
This is the biggest long-term risk, and some companies are going to get a rude awakening in 18 months' time when they start trying to use AI seriously and realise it'll mean swapping out half their tech stack.
So get a handle on this now. Understand where your data lives, which providers hold it, and what your contractual rights are. Then get started on pulling it all into a company-owned data warehouse, and insist that every tech provider you work with builds in and out of your data architecture.
A beautifully designed agentic system on top of fragmented data is a car without fuel.
There's a concept in technology called Amara's Law: we tend to overestimate the short-term impact of technology and underestimate the long-term impact.
I think this is how it's going to play out in hospitality. Agentic AI will be completely transformative, but it won't happen overnight.
The implication of this isn't that you can take your time. It's that the challenges of getting agentic AI to work in the real world are significant and will take time to work through.
It's not easy to re-engineer something that's in motion. If you want an analogy, it's like trying to turn a combustion-engine car into an electric one while you're driving it down the motorway.
So start now. Act with urgency. Expect it to take time to see transformational results. And do your best to make sure every step is taking you in the right direction, because the companies that get this right will enjoy a huge competitive advantage.
If this has resonated with you, or made you think about how your organisation should be approaching AI, please get in touch at matt@youda.co or connect with me on LinkedIn. I love chatting about this stuff. I definitely don't have all the answers, but I might be able to act as a bit of a sounding board as you think through your strategy. And if you know another operator wrestling with the same questions, do send this their way.
Let’s have a chat.
