Documentation and artificial intelligence

Generative AI can profoundly enrich our professional practices. To exploit their potential for constructive synthesis and analysis, we need to rethink our relationship with the documentation of our activities.

Anthropological changes

With the use of generative artificial intelligences that is now democratized in personal and professional life, whether for research, translation, daily and life advice, learning, writing, creation, task automation, etc., I believe that it is time at this stage to consider for organizations, and particularly cultural organizations, collective best practices in the use of these technologies. I know that some do not yet feel concerned, as they do not yet use AI, but it must be noted that young people live with it, AI is now part of their living environment, their practices and references, so it’s a beginning. In the middle of 2025, almost three years after the arrival of ChatGPT, it is certain that we are at the beginning of a profound anthropological change with no return, despite its ecological risks.

I think it is always useful to anticipate, not to stupidly “stay in the race” without knowing why we are running, but rather to understand how these technologies allow us to produce other things and differently, for the benefit of the human community.

All professions seem to be impacted in their future by artificial intelligence, we fear being replaced by machines that would format life more than it already is. But in my opinion, the real question, faced with a technological evolution that brings new human uses, is this evolution of people’s uses and needs. It’s not that professions will be replaced, it’s that human activities are enriched, modified, not directly by technologies, but by the human uses that are produced by technologies, which meet sociology and living environments.

What is magnificent about humanity is its capacity to evolve, to enrich itself, to discover, always. Thus professions, concrete human activities, reinvent themselves. And to produce other things, we must work differently, technologies or not. Throughout time, new professions have been invented to make other things, because there was a need for these other things. And conversely, in a complementary way, the advent of new technologies has also made it possible to create other things. It is a dialectic between techniques and uses. We are today, with artificial intelligences, at a moment of choice, of precise positioning on what we want to do differently and how we want to do it.

The synthesis and deduction capacity of AIs

Generative artificial intelligences initially give extremely standard, disembodied, statistical results, made of averages and not of biases. This is very uninteresting in a professional context, it remains anecdotal, even if it can boost productivity here and there. We see completely generic emails written by AI, insipid texts, distressing images of conformism... On the other hand, in organizations, one of the very great assets of artificial intelligences is their capacity to synthesize documents (written, images, audio, video, web). From these syntheses, artificial intelligences are able, thanks to their reasoning capacity, to propose action plans, really relevant, constructive and singular suggestions on the subject.

If artificial intelligences are fed with documents specific to an organization, they will be able, through their inhuman capacity for synthesis, to open doors for enriching activities, open new avenues of work, detect trends, make proposals, etc. This is extremely rich.

But for this power of enrichment to exist, we human beings must adopt a new way of doing things, to be built: it is to document much more all the activities of our organization. This involves photographing, making narratives, recording audio, collecting documents and properly classifying all these corpora so that they can be processed by artificial intelligences, which will be able to deduce work paths that we would not have been able to find by ourselves. This is extremely rich and can allow for the renewal of many things, by building on the particular history of the organization and its activities, and by deepening the singularities of the professions in which we are engaged.

New skills to develop

I insist again, this requires new vast documentation work, for the moment very little invested in professional practices. This involves scanning documents, naming files, classifying them properly, making audio recordings, applying quality voice recognition to them, archiving important email exchanges between people, keeping traces and versions of texts, etc. This is a new responsibility and it is the work of a documentalist, ultimately, which in my opinion is a new skill to be cultivated by humans. It is about preparing corpora, structuring them, so that AI can make the most of them.

It’s simple to say, but it’s not at all simple to implement, because yes, AIs have these immense synthesis capacities, but they also have their processing capacity limits, depending on the choices of AI agents we make, and depending on the capacities we allocate to them. The way generative AIs work is to take all the documents together in their RAM, and if the documents are numerous this will require a very large memory (a capacity that not all AI services have), and by the joint presence of all these elements in memory, the AI can relate all this, make syntheses and deductions, which take everything into account, which exceeds human capacities.

We will often find ourselves at the limits of AI capacities, which means that we can see them produce quite deplorable results on this type of proposals. We must also learn to establish a path for AIs, successive syntheses by stages. For example, rather than directly giving the audio of a meeting in the corpus, first have voice recognition done for the meeting, but it’s a very large document with a lot of “noise”, many useless things. So, from this voice recognition of a meeting, we will ask an AI to make a synthesis of it, but a synthesis as complete as possible, because a synthesis that is too reduced would not provide a sufficiently singular corpus. Then, it will be these syntheses with which we can feed the AI to ask for advice on strategic paths, on new activities to develop, on which audience to contact, on what type of new articles to write, which book to publish, etc.

It is possible thanks to very detailed documentation, structured and validated by humans, that AI will allow us not to be more productive, but to be more accurate, to be deeper, to be more constructive in relation to the objectives of our professions. This modifies professions quite profoundly, because we are not used to detailing the documentation of our activities so much. But the contributions, in all domains, can be unprecedented, in the service of human beings.