26.10.2026 — Summary. On 2 October 2026, the Federal Council adopted a report on automation and AI in the federal administration. The first step it announced is to improve the internal overview of systems. We think this is the right method, and it is precisely the one we offer SMEs: map before you automate.
Updated on 26.10.2026
What did the Federal Council announce?
In its press release of 2 October 2026 (in French), the Federal Council says it wants to make the administration more efficient through automation and AI. The press release is based on a report whose French title translates as “Automating processes and using artificial intelligence to improve the efficiency of the administration” (report in French).
The report sets out five priority areas:
- records and business processing;
- computer-assisted translation;
- review of applications and cases;
- data processing;
- management support processes (back office).
The first step is not a tool. As part of its AI strategy, the federal administration aims to improve its internal overview of its systems, whether these are AI systems or process automation. The press release gives no savings figure and no timetable. It is a report that sets a direction, not a law.
Why an inventory first?
Because you cannot choose what you cannot see. The context makes this clear. On 18 September 2026, a federal administration press release (in French) took stock of the AI strategy. More than a hundred AI projects are under way within the Confederation (the press release cites 116), and support is capped at CHF 2 million a year. Within this project, the Federal Chancellery coordinates the work, backed by an AI competence network.
We infer that the priority is not to launch more projects, but to make the most of existing resources: knowing what exists, where, and what it is for. An overview makes it possible to spot duplicates, compare results and decide where to focus the effort. Without one, each new project is simply added to the others, and nobody can say whether it was the most useful.
According to the same press release, plans for an AI handbook have been dropped, a decision we consider sensible given how fast the technology is changing, and instead of an AI marketplace, the existing AI project database will be developed. We read this as a sober choice: take stock of what exists before producing new material. The measures will be reviewed by the end of 2028 at the latest.
What we endorse: the administration talks about systems, not tools
The aim is not to create features, nor to select an LLM provider, and the same goes for us: the scope covers both AI and process automation. That is why we consider this point to be right. In any organisation, the useful question is not “which AI tool do we use?”, but “which process takes up time, and what already automates it?”. A processing rule in business software, a script, a macro or a chat assistant all fall under that question.
The five priority areas mentioned above point the same way. They are processes, not technologies: handling official business, translating, reviewing requests, processing data, running the back office. Analysis and the choice of tool come afterwards.
What can an SME take from this?
The same method, scaled to its size. An SME does not have a hundred AI projects. It often has more tools than it thinks, adopted one role at a time, with accounts and data that nobody has a complete list of.
For an SME, mapping means answering a few simple questions in writing:
- which AI and automation tools are used, and by whom;
- for which processes;
- with which data, and under which accounts;
- with what result, measured or not.
The five federal areas have their equivalents in a company: case tracking, translation, handling customer requests, data management and the back office, accounting included.
The practice we criticise is the reverse: buying a licence, then realising that most of its strengths go unused. It leads to scattered tools, and to gains or losses that nobody measures.
That is why our engagements begin with a diagnostic, before anything is built; we describe it in our approach.
We therefore share the Confederation’s approach: first know which tools are in use, with which data and under which rules. That is the starting point of our support with AI governance.
The limits of our position
Our thesis has limits, and we would rather put them in writing.
Scale, first. The federal administration and an SME do not have the same constraints or the same resources. For a small organisation, the inventory must stay proportionate to its size.
The risk of waiting, next. An inventory can become an excuse for not deciding anything. Some improvements are obvious and do not need a complete map before they are launched. Mapping must not replace action, only guide it.
Shelf life, too. An overview is a snapshot. The ways AI is used change quickly; a map that is not kept up to date loses its value within a few months.
What we do not know, as well. The press release does not say how the federal overview will be drawn up, nor how often. Our view concerns the principle, not an implementation we know nothing about.
Our own interest, finally. We offer diagnostics ourselves, so it is only natural that we argue for measuring before acting. We say so so that readers can take it into account, and because a diagnostic can also conclude that it is better to automate nothing.
The takeaway
By making the overview of systems its first step, the federal administration is applying a simple rule: know before you choose. An SME does not need to wait for a report to do the same. A map of its tools, its processes and its data is enough to decide where AI will be useful and where it will not.
By Sofia Mestre