Writing.
Two series: notes for founders, and how AI and consolidation are changing hospitality technology.
For founders
Notes for founders and investors on building, shipping and diligencing technology businesses.
-
Turning off training doesn't mean your AI tool keeps nothing.
Anything you paste or upload into a cloud AI tool sits on someone else's servers for a while. Four questions to ask before you do.
Read the article -
The time between a security fix and an attack is now measured in hours.
AI found two critical PHP flaws that sat unnoticed for a decade. Attackers have the same tools. What founders should have ready for the next one.
Read the article -
Moving to another cloud region is only easy if you've already worked out how.
A power cut in one Google Cloud London zone hit customers in all three. Moving region only works if you set it up beforehand.
Read the article -
Anthropic outgrew git. The same problem shows up at much smaller scale.
Anthropic's merges went from 2,500 to 30,000 a day and git couldn't keep up. A one-person project hits a smaller version of the same wall.
Read the article -
From demo to first pilots: four things that break.
The demo works. The question is what happens when ten sites are live and using it for real. The same four problems come up nearly every time.
Read the article -
Fractional CTO or technical co-founder?
A co-founder was never a fractional role, and the two get confused a lot. If you're still working out what to build, you need a co-founder.
Read the article -
The code was not what failed on Saturday night.
An AI voice product's provider went down for 94 minutes on a Saturday evening. Nobody had built monitoring, alerting or a fallback.
Read the article -
How much of this codebase was written by AI is the wrong diligence question.
Nobody can answer it, and the percentage tells you nothing about the risk. Does anyone understand this? Can it be changed without breaking?
Read the article -
A working product no longer proves anyone understood how it was built.
Researchers found Moltbook's database key in the site's own JavaScript. The tools that produce the code don't hand you the practice that goes with it.
Read the article
On the shift
A weekly series. It follows AI, consolidation and pricing through one industry in detail, because that is where the pattern is easiest to see.
-
The FTC says cancelling Toast can mean repaying its loan in one go.
The FTC alleges restaurants with a Toast Capital loan must repay it in full if they cancel or switch POS. Agree exit terms before you sign.
Read the article -
Consolidation is squeezing out hospitality's independents, and operators will pay for it.
Why I think consolidation in hospitality technology leaves operators with less choice, weaker leverage and higher bills.
Read the article -
Who decides which AI agents can book your restaurant.
Agents are getting in either way. The rules on who gets in are being agreed between platforms and AI companies, and restaurants aren't party to them.
Read the article -
Why hotels still make you sign for a drink.
Every drink at the bar came with a chit and a pen. All of that could be done at the terminal. So why is it still rare?
Read the article -
Burger King pulls back on drive-thru AI.
Burger King is pulling back on AI voice ordering in its drive-thrus. Full service restaurants face their own version of this on the phones.
Read the article -
Where the answer comes from when an agent asks for a table.
It either comes from a copy of your availability, or it comes direct from your diary. Whoever the agent asks is the one who learns what the guest wanted.
Read the article -
Marriott is testing ads on OpenAI's ad platform.
Marriott is testing ads on OpenAI's ad platform. An independent restaurant will pay through its booking platform, with the platform's margin on top.
Read the article -
Walmart can run a restaurant delivery service without charging the restaurant.
Walmart can charge the restaurant nothing because it isn't trying to make money on the restaurant. DoorDash and Uber will have to price against that.
Read the article -
83% of restaurant locations never appear in AI recommendations at all.
83% of restaurant locations never appear in AI-generated recommendations at all. They're not ranked lower. They're simply absent.
Read the article -
Google now holds the whole hotel transaction, from first question to payment.
The hotel carries the cost and the service obligation, and Google keeps the understanding of the guest.
Read the article -
Six-figure cheques to switch reservation platform tell you what your guest book is worth.
Three months' fees buys a subscriber. Six figures buys a guest history for whoever owns the platform.
Read the article -
Your POS decision and your reservation decision are no longer independent.
Toast with Resy, Square with OpenTable. Your POS decision and your reservation decision are no longer independent.
Read the article -
ChatGPT started taking bookings, and whether yours can be booked was decided years ago.
Whether a restaurant is bookable in ChatGPT is decided by which reservation system it uses. You don't rank lower, you're simply not bookable at all.
Read the article -
The guest still chooses. The operator still doesn't.
Preferred is just a label and not a rule. The guest still chooses. The operator still doesn't. There's just less screen for the choice to happen on.
Read the article -
A restaurant searching hundreds of millions of delivery customers becomes a controller of that data.
A restaurant receiving personal data about someone it has never served becomes a controller of it, and owes that person a notice.
Read the article -
Restaurants can take orders inside ChatGPT with no marketplace commission.
There is no marketplace commission on these orders. Square has opened an AI ordering channel at a fraction of the cost of the marketplace one.
Read the article -
Uber is buying Delivery Hero, and reservations are part of the prize.
Delivery, reservations and guest data are becoming one market, and two companies are positioning to own it.
Read the article -
Mews cut 15% of its staff and called the roles obsolete.
Mews plans to move beyond selling software and become a service provider. More of your business flows through fewer companies.
Read the article -
Four hotel groups answered the same question and did not all answer it the same way.
The hotel chains are choosing between building the front door and renting space behind someone else's. On the restaurant side, the platforms are choosing.
Read the article -
Google gave the protocol away, because the protocol was never the lock-in.
The protocol was never going to be the lock-in. Google doesn't need to own the standard. It owns the place the question gets asked.
Read the article -
The closed stacks only win if the rest of the market lets them.
The closed stacks only win if the rest of the market lets them. A standard fixes the technical layer and leaves the commercial one alone.
Read the article -
When an assistant picks the restaurant, loyalty moves to the assistant.
Loyalty doesn't disappear in that world so much as move. The guest becomes loyal to the assistant that keeps getting it right, not to the venue it picked.
Read the article -
One book instead of three iPads, but where do the bookings actually end up?
Owning your data and controlling where it lives are not the same thing, and the FAQ answers the first while staying quiet on the second.
Read the article -
A credit card company is becoming one of the biggest owners of restaurant reservations.
For a business the size of American Express the revenue from reservations is a rounding error. What it's buying is the dining data.
Read the article -
Two hotel CEOs said opposite things about AI and both were right.
AI can hollow out the back office and free up the front of house at the same time. Most operators will live through both.
Read the article -
A waitress remembered our order. No system prompted her.
AI amplifies what is already there. It doesn't install a service instinct that was never present.
Read the article -
Most AI advice tells operators to buy something. The best first move is the opposite.
The honest answer is to begin with what you already have, before you spend a penny on new tools or new vendors.
Read the article -
Building a daily intelligence brief for hospitality AI.
What I learned about cost, constraints, and the gap between AI demos and operational reality.
Read the article -
The pricing model your software vendor is moving toward.
These models aren't better or worse than flat subscriptions. It depends entirely on the numbers, and on whether the vendor can prove AI caused the result.
Read the article -
Shadow AI is an operator problem, not an IT housekeeping one.
This is now a board-level concern, not an IT housekeeping one. Do you know what AI tools your team is already using?
Read the article -
Being found by an AI is only half the problem.
Being discovered by an AI is only half the picture. The other half is being bookable through one.
Read the article -
One company now owns the delivery platform, the reservation system and the guest data.
One company now owns the delivery platform, the reservation system, and the guest data that flows through both. Deliveroo gets the guest.
Read the article -
You got the booking. But whose guest is it?
You got the booking, but what do you actually know about that guest? This is the question most operators haven't got to yet.
Read the article -
Is this actually AI, or software doing what software has always done?
The problem isn't the technology, it's paying an AI premium for something that was never AI in the first place.
Read the article -
The next trillion-dollar company won't sell software. It will sell the work.
Hospitality is exactly that market. The consolidation that's already happened in hospitality tech over the last decade is going to accelerate.
Read the article -
86% of operators are comfortable with AI. Only 24% use it for core tasks.
86% of operators are comfortable with AI but only 24% are actually using it for core operational tasks. Three questions to start with.
Read the article -
Google just became a restaurant booking platform in the UK.
Google owns the moment a diner decides where to eat. Discovery platforms have gone from being a destination to being plumbing.
Read the article -
Fifteen years in hospitality technology, and what I'm seeing now.
Operators are on the receiving end of this whether they're ready or not. That's why I'm writing this series.
Read the article