A proposal for a one-day training session for staff of a local authority’s cultural service. The aim: that participants leave able to produce, on a regular basis and using methods drawn from collective intelligence and artificial intelligence, the narratives their cultural policy needs.
Why cultural policies need narratives
Cultural budgets are falling almost everywhere, and elected officials across the political spectrum are turning away from culture, in their platforms as in their funding decisions. To my mind, this disengagement does not stem from hostility towards culture, but from a lack of understanding. When there is no narrative of what a territory’s cultural organisations actually do, of their social role, of the meaning of their public funding, neither elected officials nor citizens can grasp what they are for. Only the sector’s own professional community then understands their usefulness, which is not enough in a democracy, because the purpose of public money is not to support a professional community but to serve citizens.
I have often used the example of Arcadi, a public cultural agency in the Île-de-France region that supported the touring and presentation of the performing arts for twenty-five years. A political decision closed it within three months, and its website disappeared soon after, taking with it every narrative of its work and, with them, its very existence in public memory. Had its narratives been circulated somewhere other than its own official website, in newspapers, on other platforms, in the words of local residents, it would have been far harder to close it without almost anyone noticing.
In today’s digital space, it is in fact text, far more than images, that forms the traces we remember. Search engines and artificial intelligence systems index and redistribute information by relying heavily on the written word. A cultural action that has no written narrative therefore does not exist within that memory, and so does not exist within the information that now reaches citizens and decision-makers alike.
Narratives thus serve very concrete purposes: giving elected officials the material to understand and defend a budget, legitimising activities in the eyes of residents, and weaving a memory of the territory that makes a cultural policy far harder to erase. This is what I have called, borrowing Nassim Nicholas Taleb’s concept, the antifragility of cultural projects.
Narratives that change the work itself, not just its communication
A possible misunderstanding needs to be cleared up here. Producing narratives, in the perspective I am proposing, is not about communication, nor is it about adding a burden to teams that are already very busy. Narratives, once we learn to gather and produce them within the very movement of the work, provide a body of qualitative evaluation that nothing else can offer. This material then makes it possible to understand more finely what is at stake in our activities, to argue from lived experience rather than from generalities, to convince an elected official or a partner, and also to think more clearly oneself about what one is doing. The work gains in depth and in scope, and often in speed, because one no longer starts from scratch with each progress report, each grant application, each preparatory meeting.
It is this new quality of the work, made of deeper dialogue with residents, partners and elected officials, that I believe we need to cultivate today in order to defend culture. The time and resources it requires will, to my mind, prove far more useful and far more substantial than those we pour into maintaining a presence on social media whose real contribution is hard to measure.
What artificial intelligence really changes
Many staff already use ChatGPT or Claude to draft an email or summarise a document, and so believe they know how to use these tools. At that level of use, AI does render real services, often at the cost of many iterations before reaching a satisfactory result. That is already very useful, but one can go much further. The real change lies in new ways of working that few people have yet built, because they call for a long period of research and experimentation.
For three years I have been carrying out this research in depth, both in my own editorial work and in supporting cultural organisations. It was in particular while building the collaborative platform Azimut for the MPAA (Maison des Pratiques Artistiques Amateurs of the City of Paris) that I measured the scale of the problem. A resource platform stays empty as long as one lacks methods for producing those resources, because quality writing takes time that teams do not have. The methods I have developed answer this problem, and they can now be passed on.
The methods taught during the day
So as not to remain abstract, here are the concrete methods on which the training rests. All come out of this research, all are documented, and all will be addressed during the day, most of them through practice.
- Gathering material through collective intelligence. A truthful narrative is not invented in front of a screen; it is nourished by people’s lived experience. I use facilitation methods I have developed and documented (paired exchanges, written contributions through a simple digital tool, audio recordings of free-flowing conversations) that make it possible to gather, in a short time, rich and singular material, the kind that conventional questionnaires never capture.
- Gathering material day to day. The training day gathers its material within the group, but the aim is for this gathering to become an everyday practice of the service, woven into the work rather than added on top of it. It relies on light gestures, such as recording a conversation at the end of an activity (with people’s consent), taking photographs, dictating three minutes of notes on leaving a meeting, keeping and naming the documents produced. These gestures cost almost nothing in the moment, and it is the AI that then takes on the heavy work of transcription, sorting and synthesis.
- Building corpora and managing context. This is the methodological heart of the matter, and it is what almost no one yet practises. AI produces singular results only when fed with singular, well-prepared documents. This calls for precise know-how: applying speech recognition to recordings, naming and sorting files, building successive step-by-step syntheses so as not to saturate the AI’s working memory, and framing requests that do not lock the machine into our own way of thinking.
- Writing techniques with AI. Getting from an AI a text that carries a voice, and does not read like machine prose, requires several passes of revision, with explicit instructions about the flaws to correct. I pass on these instructions, which I built by spotting and naming, one by one, the tics of AI writing, and we co-construct them as skills specific to the service, that is, reusable writing instructions that carry its editorial voice.
- Reuse and pooling. Material that is well gathered and well sorted serves several times over. The same corpus feeds a narrative for the website, an argument for an elected official, a section of an activity report, a grant application. I pass on concrete techniques of reuse, which turn each documentation effort into a lasting resource, and techniques of pooling between colleagues and between organisations, so that corpora and methods circulate.
- Reusable instructions. Once a method has been found for a given type of narrative (an account of a cultural activity, a portrait of a resident, a territorial narrative, an educational resource), it can be written up as a permanent instruction, what is called a “skill” in technical vocabulary. The service thus gradually builds an editorial chain of its own, and producing narratives becomes a regular practice that never starts again from scratch.
- Confidentiality, security and sovereignty. Working with AI in a public service calls for clear strategies about what is and is not entrusted to it, about the choice of tools, about where data is stored and about its durability. I pass on these guidelines, which are an integral part of the method and which allow for responsible use, consistent with the values of public service.
Passing it on and cooperating beyond the service
These methods are not only for internal use. Once acquired, they can be passed on, first to elected officials, who need this understanding, but also to the other cultural actors of the territory, and the training gives staff the first tools for doing so.
They also open up a field I believe to be decisive: cross-sector cooperation. Working across culture, social services, education, health or urban planning is notoriously difficult, because each sector has its own documents, its own logics, its own timescales and its own vocabulary. Artificial intelligence, used with method, makes it possible to cross-reference these documents and logics, to produce syntheses that bring different frames of thought into dialogue, and so to cooperate where coordination had so far failed for lack of shared material.
The shape of the day
The day alternates between input, experimentation and real production. It is designed so that participants leave not with notes, but with a narrative they have produced, a method they have tested and a tool they can reuse.
Morning
- Welcome and a collective intelligence exercise on the group’s current uses of AI, in order to start from each person’s actual practice.
- The political stakes of narrative within cultural policy, and the place of narratives as material for qualitative evaluation and reflection, drawing on concrete examples from various territories.
- A first gathering of material: paired exchanges and audio recordings about the service’s recent cultural activities, which will form the working corpus for the afternoon.
Afternoon
- Preparing the corpus: speech recognition, naming, sorting, step-by-step syntheses.
- Producing a first real narrative with AI, working on context management, multi-pass revision and the reuse of the corpus for other purposes (arguments, reports).
- Collectively drafting a reusable instruction specific to the service, together with confidentiality and sovereignty guidelines for everyday use.
- A closing session on day-to-day gathering, on the dissemination of narratives (website, local press, social media, partners) and on the internal organisation to be put in place.
By the end of the day
Participants will be able to gather the narrative material of their territory, both day to day and in dedicated collective sessions, to build corpora that AI can work with, to produce narratives that carry the singularity of their action, to reuse that material across their various working documents, and to sustain this production through reusable instructions, with clear confidentiality and sovereignty guidelines. They will also have the first tools for passing these practices on to their partners and for putting them at the service of cross-sector cooperation.
To my mind, it is this capacity for regular production, far more than any communication tool, that can durably strengthen the place of a cultural policy in the eyes of elected officials and residents, by changing the very quality of the work that carries it.
Practical arrangements. A single day from 9.30 a.m. to 5.30 p.m., for a group of 8 to 12 staff. Each participant brings a laptop; the AI access needed for the day is provided. A modular room, with movable tables, is preferable. A remote follow-up session, a few weeks later, can be added to support the first independent productions.
This programme is a working basis, one that will be adapted in detailed and attentive fashion to the realities of the service, to its current projects and to the narratives it most needs.
