The report that could not name it

I read the January 2026 ADN Tourisme x AFDAS x Matrice report on AI in France's destination management organisations (DMOs). I read it as a trainer who in 2025 ran sixty sessions and worked with more than twelve hundred tourism and territorial professionals. Here is what I saw, what its authors saw, and what the format of their commission stopped them from seeing.

The report holds

Let me say this plainly. This is not a quick report. The method is serious, the interviews are dense, the verbatim quotes are chosen with care. The study was initiated by ADN Tourisme, the federation representing more than twelve hundred destination management organisations in France, around twelve thousand seven hundred professionals across the sector.^[1] It was carried out with AFDAS, the French national skills fund, as part of the Studies and Foresight on Employment and Skills programme, and delivered by the consulting firm Matrice from July to December 2025.

At the core, twenty-two in-depth qualitative interviews conducted between August and October 2025, across eight organisations, from local tourism offices to a regional agency, with departmental committees and a sector AI expert in between.^[1] The panel gives a picture of the entire tourism value chain, from the front office to the data functions. The people interviewed speak truthfully. Quotes are attributed to job functions, not weaponised.

The conceptual frame set out in the opening pages is precious. It carefully separates data, automation, algorithms, specialised AI and generative AI. That clarification, which takes up the early pages, defuses much of the routine confusion in conversations on the subject. Anyone about to speak about AI in a DMO would do well to read it first. The foresight method that follows, three scenarios built by morphological analysis on a 2028 horizon, does not go for easy futurology. It opens up reasoned bifurcations and traces their operational implications.

That is what makes the analysis worth reading. Not a bad report. A good report that misses its object.

What the verbatim quotes are saying

The report, in its first chapter, opens on the "already concrete presence of AI, far from a uniform rollout". It lines up, soberly, the quotes gathered in the interviews. These quotes deserve to be reread for what they are, not for what the heading turns them into.

A web communications professional: "I use ChatGPT 100% of the time." Another, in the same section: "It's my deputy, really."^[1] A data observation officer, later in chapter two: "I think I'd use it less if we were three or four. As it is, when I can't ask my colleagues, I use GPT."^[1] A director on subcontracting: "We had a small budget to outsource this brief. With AI, today I'd struggle to justify outsourcing it."^[1]

And then, planted in section 5.6, a case that has already become emblematic. "In Corsica, they hired someone like that. In three days, she built an incredible internal tool. We had asked her to map tourism signage in Excel. She built an internal information system and got noticed by Apidae."^[1] A junior data analyst, hired to produce a spreadsheet, who delivers a full management infrastructure in seventy-two hours, and whose work is eventually spotted by the national network's head office.

These quotes are not illustrative. They are structural.

The report files them, however, under a heading titled "A progressive transformation, still largely assistive". They are read as signs of an integration by patches, by isolated initiatives, without organisational tipping. The summary of the first major section is explicit: "uses observed which fall more under integration through isolated initiatives of sector professionals than under organisational shifts."^[1]

This is precisely where a second reading becomes necessary. What gets filed under "isolated initiatives" is, in the grammar that the sociology of work has begun to build over the last three years, a reversal of technological power. The report sees it. It describes it. It writes it down. It cannot name it.

A category mistake

The report treats these quotes as BYOD. They are no longer BYOD. They are BYOAI. The difference is not a technical detail. It changes everything that follows.

BYOD, in the 2010s, was a transfer of devices. Employees brought their personal phone or laptop because the company-issued one was slower or uglier, and IT organised the wrapper. A usage charter, a VPN, an MDM, a security policy. The object moved. The work stayed the same. BYOAI does not move an object. It moves power. The person bringing ChatGPT to their desk is not bringing a productivity tool. They are bringing an execution capability that until then belonged to the organisation, or to outside vendors, or to colleagues no longer in the budget. I laid out this shift in BYOAI, when workers bring their own AI, showing why it is a historical reversal of the vector of automation, not a modern variant of Shadow IT.

The consequence on the reading of the report is immediate. When a communications officer says "I use ChatGPT 100% of the time", she is not describing a peripheral use of a tool. She is describing the silent reorganisation of her own function. When a data observation officer says he uses GPT because he has no one to lean on, he is not describing a workforce workaround. He is describing the appearance of a substitute colleague his employer does not know exists, has not paid, and cannot manage. When the director says she can no longer justify the outsourcing budget, she is not describing a cost saving. She is describing the disappearance of an entire line of the value chain, subcontracting included.

And when the data analyst in Corsica delivers an information system in three days instead of a spreadsheet, she is not overshooting her brief. She is practising what sociology has called job crafting for the last twenty years, this time technologically armed. Her job description calls for mapping signage. She equips herself, alone, and redefines her job from below, mid-execution.

Read with this grid, the sequence of quotes stops being illustrative. It becomes the main event of the report. And the logical next step of the analysis changes nature. A one-page usage charter, of the kind the report recommends in closing, addresses the ergonomics of a phenomenon that is cognitive and political. It will arbitrate nothing, because there is nothing to arbitrate ergonomically: what has already happened does not wait for a charter to be signed.

The format locks the reading

The next question is why a report this serious could not write what its own quotes were shouting at it. The answer is not the authors' competence. It is the structure of the commission.

A study is commissioned by a federation to a consulting firm under a public procurement framework, here the Studies and Foresight on Employment and Skills programme run by AFDAS.^[1] The deliverable must produce actionable recommendations for the commissioning body, that is, collective devices recognisable by the organisations that placed the order. A national hub. A charter. A reference person. A training plan. A framework. The format of the delivery constrains the format of the possible reading of the field.

This is exactly what Marie-Anne Dujarier analyses in Le Management désincarné ("Disembodied management"), where she shows how the planneurs she identifies, around forty percent of French middle management today, produce devices that never operate on their own ground.^[2] The ADN Tourisme report is itself a product of that layer. And that layer, by construction, cannot read a phenomenon that plays out by short-circuiting it. BYOAI short-circuits the planneur layer. A study commissioned to a planneur on BYOAI mechanically hits this limit. This is not a moral flaw. It is a structural consequence of the device for producing knowledge.

To be clear: this is not a criticism of individuals. The researchers who carried out this study did, within their frame, a clean piece of work. The lock is upstream, in the very nature of the relationship between a federation that funds, a body that pilots, and a firm that delivers.

The report documents a revolution. It was commissioned to document a tool.

This tension also explains why the report concludes, in plain words: "Generative AI does not replace the jobs of DMOs, it shifts their work."^[1] The sentence is correct. But it stays at the scale of the task and the function. It does not descend to the sociological grammar of the shift. And this shift is not organisational. It is individual, at the scale of the person, function by function, at different speeds, in a single direction: the agent equips themselves, the organisation follows. This is the dynamic I have documented in The revolution that never happened, and which the verbatim quotes of the ADN report now confirm, sector by sector, in French institutional tourism.

Picking up the work where it had to stop

The ADN Tourisme report did what the format of its commission allowed it to do. That is already a great deal. The logical next step is not another delivery in the same format, by another firm, under another public contract, with alternative recommendations produced at the same cadence. That would simply repeat what I have just described. The logical next step is elsewhere.

It consists in moving the work outside the perimeter of institutional commissioning. In pushing it toward what it always should have been, given the order of the object encountered: a sociological research effort over the long term, crossing the sociology of work, the economics of the firm, and the political economy of digital technologies. A piece of work that does not have to deliver actionable guidelines in six months, and that can therefore descend to the grammar of the phenomenon. A piece of work that can name what the commission, by construction, cannot name. And that will then produce recommendations that hold, in time and at scale, for institutional tourism and for adjacent sectors.

What DMOs are already doing, under the radar, is the real material to study.

The stakes, in fact, reach well beyond tourism. What the twenty-two ADN interviews document inside DMOs plays out, at varying paces, in local government, in education, in law, in healthcare, in almost the entire qualified service sector. Institutional tourism in France simply offers, through the density of its network and the diversity of its functions, a particularly readable observation field. The phenomenon itself is cross-sectoral. And its treatment cannot remain at the level of a sector-specific consultancy on an eighteen-month contract.

So I am not proposing here to replace one recommendation with another. Drawing from the ADN report, in the same institutional language, a set of alternative shortcuts would do exactly what I just criticised. The right question is not "which charter rather than this one, which hub rather than that one". The right question is "which frame of thought for what is happening, and which actors have the legitimacy to build it". By construction, these are not the sector federations and the firms they hire. They are research laboratories, universities, public observatories, long-form action-research programmes. That is where, rather than in another wave of charters no one will follow, the network's leaders would gain most by investing their commissioning power today.

The work of naming what is happening will not be done inside consulting firms. It will be done elsewhere, at a different pace.

That is what I see, every week, in the rooms where I train. The agents who have crossed over come back with outputs their leadership discovers in meetings two months later, often without understanding where they came from. The leadership teams I present this reading frame to react as if to a fresh diagnosis, while their own people have been living it for eighteen months. The gap between the two is not a question of information. It is a question of the capacity to read what is already in front of them. And that capacity is not commissioned by public contract.

Jean-Jérôme DANTON