The seven operating truths of AI-native companies (and what's left in the hands of an Italian hotel or tourism destination)

There is a genre of article, now recurring, that describes companies where artificial intelligence is not one more tool but the nervous system of the entire organisation. McKinsey has just published a particularly dense one, based on interviews with fifteen companies defined as "AI-native" — from fintech startups to agritech platforms — and has distilled it into seven "operating truths" that supposedly distinguish those who get real results from those who stop at the first chatbot they install.

I'll admit my first reaction, reading it, was one of distance. Italian tourism, made up overwhelmingly of micro and small family businesses, destinations with razor-thin budgets and teams that often amount to a single person doing everything, seems to inhabit a different planet from startups that have "agents with their own names and Slack handles" or CEOs who build production integrations while riding the metro. But precisely for that reason I wanted to read it all the way through with critical attention, because sometimes it is the most marked distances that reveal which principles remain valid even at very different scales — and which ones should instead be left where they are, meaning in a context that is not ours.

First truth: treat agents like colleagues

McKinsey observes that the most advanced companies no longer talk about AI as a "co-pilot" but as a genuine co-worker, with assigned tasks, progressive autonomy and the capacity to work twenty-four hours a day. For a small independent hotel, this language sounds almost science-fictional, but the underlying principle makes sense: AI starts to deliver real results not when you use it to write an email a little faster, but when you hand it a defined slice of a process that used to belong to a person, with a clear perimeter. In my own concrete experience, this translates, for instance, into managing review responses during low season (which I still check, adjust and personalise before sending) or into first drafts of seasonal newsletters — tasks that can be delegated with supervision, not into managing the relationship with the guest, which remains and must remain human.

Second truth: build only what is truly yours

The criterion McKinsey attributes to the companies interviewed is blunt: you build internally only what creates a defensible advantage based on data, expertise or intellectual property that an off-the-shelf tool cannot replicate — everything else you buy. Here the principle transfers perfectly to our sector, perhaps even more so than for a tech startup. The data and knowledge that a hotel or a destination actually own — the story of the territory, the relationships with local producers, the way a guest is welcomed — are the only thing no competing software will ever be able to replicate. Booking technology, revenue engines, analytics tools: you buy them, choose them, evaluate them, but it is not worth building them in-house. The competitive advantage of a small property never lies in the tool; it lies in what the tool cannot copy.

Third truth: curate the knowledge layer

This is, in my view, the most concretely useful part of the entire article for anyone working in our sector. McKinsey reports that AI-native companies do not necessarily insist on a single centralised source of truth, but build lightweight connectors that make any data queryable wherever it lives — whether that is an Excel sheet, a chat, a shared document. And they warn that this knowledge layer deteriorates faster than you might expect: if data becomes stale, agents will keep delivering confident but wrong answers, and trust evaporates. For an independent hotel or a destination, this translates into a very practical question I often put to the clients I work with: does the information about your property, scattered across your website, OTAs, Google Business Profile, social channels and printed materials, all say the same thing? Is it up to date? If an AI response engine finds different versions of the same check-in time or the same price, the problem is not technological — it is one of discipline in information management, and it is work you can start today, without buying a thing.

Fourth truth: a composable, governed architecture

This is the point where the distance between the companies interviewed and our sector widens again. McKinsey talks about a modular architecture in which tools integrate with one another according to clear governance rules, defined permissions and traceable actions. That is a sensible conversation for a company that develops its own systems; much less so for a property that relies on a PMS, a booking engine and a couple of marketing tools purchased from external vendors. The practical advice I find myself giving here, also based on the work I do with the properties I support, is more modest but equally important: before adding a new AI tool, always ask yourself what it needs to talk to, who manages it and who has access to what. The chaos never comes from the first tool — it comes from the fifth tool that nobody connected to anything else.

Fifth truth: trust is built incrementally

One of the most interesting passages concerns progressive autonomy: you run a process manually until the friction demands otherwise, then you automate individual steps one at a time, not all at once. "Automate slowly, do it by hand until the pain forces automation," says one of the founders interviewed — a principle I fully share, and one that holds, if anything, even more strongly for a small property where an unmanaged AI error immediately becomes visible to a real guest. I never hand an entire process that touches the guest to an automated tool until I have watched it work, with supervision, long enough to know its limits.

Sixth truth: scale requires a different organisational design

McKinsey observes that companies that truly scale AI reorganise small multidisciplinary teams around concrete business outcomes, instead of letting each function adopt its own tools in isolation. This is honestly the point where the distance from our reality is greatest, because in a hotel with a handful of rooms or a destination with two people in the office, scale simply does not exist in the same sense. But the underlying principle — that AI delivers results when the people involved share the same business objective rather than using disconnected tools — remains valid even for a small team, and is worth keeping in mind rather than chasing the idea of "reorganising" the way a scaleup would.

Seventh truth: adoption is a culture, not a rollout

The last truth is probably the most exportable of all, and it is the one I insist on most when I do training. The companies that have truly integrated AI did not impose it with a deadline; they turned it into a virtuous cycle made of four elements: leaders who use it first and show that openly, sharing the results obtained, concrete measurement of adoption, and hiring people who are already comfortable with these tools. In a small property the fourth element matters less, but the first three are just as decisive as in any scaleup — perhaps more so: if the owner or management never uses these tools in person, the rest of the team will rarely do so with conviction, and without conviction AI remains an isolated experiment that never produces real value.

What to take away

The real value of this report, read from our vantage point, is not the model as a whole — which remains designed for organisations with resources and velocity that independent Italian tourism simply does not have. The value lies in the cross-cutting principles that hold at any scale: build only what is genuinely distinctive, keep data clean and consistent before chasing the next new tool, give AI autonomy gradually rather than all at once, and above all be the first to use it — honestly, before asking anyone else on your team to do the same. No radical transformation is required to apply them. All it takes is the discipline to do it methodically, rather than out of fashion.

FAQ

Q: What are McKinsey's "seven operating truths" about AI-native companies?

A: They are principles identified by McKinsey from interviews with fifteen companies that use AI in a structural way, covering the treatment of agents as colleagues, the build-versus-buy decision, data curation, technical architecture, incrementally built trust, organisational design and adoption culture.

Q: Do these truths apply to a small independent hotel as well?

A: Only in part. Some principles — such as data consistency and gradual automation — are directly applicable at any scale. Others — such as reorganising into agentic teams — presuppose resources and dimensions that a small property typically does not have.

Q: Where should a small tourism property start if it wants to engage seriously with AI?

A: With the consistency and quality of its own data (opening hours, prices, descriptions) across all channels, and with the gradual automation of individual processes, verified with human supervision before expanding their autonomy.