Generic AI can write fast. Brands don't need fast.
What hospitality brands need is language that stays consistent under pressure — across menus, internal briefs, and guest-facing communication. A note on brand-tuned tools, and what they're really for.
The first time I asked a general-purpose chatbot to write me a holiday menu insert for one of the restaurants, the result took eleven seconds and was the kind of copy that, three years ago, I would have paid a junior writer two days for. It was competent. It was clean. It was also, in a precise and deflating way, exactly the same as the copy three other restaurants in our category were probably going to publish the same week.
That is the problem with fast. Fast is generic. The defaults of a generic model are the average of every menu, every welcome card, every press release the model has ever read. The average is fluent. The average is also forgettable. Hospitality brands cannot afford forgettable. The whole proposition is the opposite of forgettable — a feeling of having been spoken to specifically, by people who actually know what they are doing.
What brands actually need from these tools
Not faster output. Output that holds its line under pressure.
The pressure is the operating reality of running a restaurant or a hospitality brand at scale. A menu insert is not a one-off. It is the latest in a long series of touchpoints — the website, the reservation confirmation, the in-room card, the response to a difficult Google review, the apology note for a kitchen mistake, the holiday newsletter. If those touchpoints sound like they were written by twelve different people who have never met, the brand is not landing. The customer notices. They cannot tell you what they noticed; they can tell you the place “doesn’t feel as tight as it used to.”
The unsexy promise of a brand-tuned tool is consistency. The same voice on a Tuesday morning press release as on a Saturday night incident apology. Not because a person is policing it. Because the tool was given enough of the brand’s actual writing to learn what the brand sounds like when it is working — and refuses to write the average.
The discipline the tool requires
The mistake most operators make with these tools is to treat them as writers and judge them on speed. They are not writers. They are an instrument, and like any instrument they require a setup pass and a hand on the controls.
The setup is the part most operators skip. Twenty pages of the brand’s existing writing, ranked from “this is exactly us” to “this is not us at all”. A short paragraph on what the brand will not say. A list of words the brand uses and a list of words it does not. The names of the people the brand actually addresses. None of this is glamorous. All of it is the difference between a tool that is a slightly faster version of the average, and a tool that sounds like the brand on a good day.
The hand on the controls is the editor. The tool produces a draft. A person who knows the brand reads it and decides whether it is shippable. Most of the time, with the setup done correctly, the answer is yes with one or two small edits. Without the setup, the answer is “rewrite it from scratch”, and the operator has lost an hour they could have spent on the floor.
What this is not
This is not an argument that AI is going to replace the writer in a hospitality brand. The writer is the person doing the setup, doing the editing, and deciding what the brand will and will not say in the months ahead. That work is more important now, not less. The tool simply removes the inefficiency of the writer producing every routine sentence by hand.
It is also not an argument for personalization in the algorithmic sense — the kind that addresses a guest by their first name and recommends a wine based on their last three orders. That is a separate conversation, and frankly a more delicate one. Personalization done badly is creepy. Personalization done at the level the technology currently allows is mostly done badly. I would put a brand-tuned writing tool into a serious operation tomorrow. I would not put algorithmic personalization into a guest-facing experience without months of careful work.
Where I have seen this land
The places it has landed best have been the ones where the writing was already a problem. A multi-unit operator whose unit-level managers were all writing their own social posts, with predictable variance in tone. A heritage brand whose internal team was producing newsletters that no longer sounded like the brand because the brand had drifted from its own language and stopped noticing. A founder-led restaurant whose founder was, frankly, the bottleneck on every customer-facing line of writing the business sent.
In all three cases the tool did not replace the writer. It replaced the variance. The tone of every public touchpoint converged. The customer noticed in the way they always do — not in a comment, but in a sense of the place feeling tighter.
The fast version of this is fine for marketing copy that nobody will ever look at twice. The version of this that matters in hospitality is the slow setup, the disciplined editor, and the brand voice that finally stops drifting on the days the founder is not in the room. Generic AI can write fast. Brands don’t need fast. They need their own voice, repeatable.