We went to Action Day to talk about AI.
We came back talking about people.

At Fondazione Agire's Action Day Evolution 2026, held on 10 September at Padiglione Conza in Lugano, we went to hear real cases of artificial intelligence inside companies. We heard them. But the theme that kept coming back, on stage and at the tables, was not a technology: it was people, their skills and the responsibility for making decisions.

A day for working, not for listening

The event, part of the Swiss {ai} Weeks programme, chose a format quite unlike a classic conference: a handful of talks, mixed tables of manufacturers, service companies and suppliers, and a LEGO® Serious Play® workshop run by Essere Agile and Rasmussen Consulting. For us it was a real find: a method we did not know, and it works. You build a model out of bricks, explain it, connect it to your neighbour's. Ideas become visible, and within minutes the table is discussing things that would have stayed unspoken in an ordinary conversation. Hard to remain a spectator.

Fondazione Agire set the principle that ran through the whole day in its opening remarks: technology is a means, not an end. What has changed compared with earlier waves is speed: these days you can find yourself switching the AI tool you use every day three times in a single year. Hence two concrete behaviours suggested to companies: set aside time for innovation regularly, until it becomes a habit like going for a run, and experiment while accepting that some attempts will not work.

A live poll among participants gave a snapshot of the room: almost everyone already uses AI, but the most common pattern is still individual, occasional use; the obstacles mentioned most often were privacy, change management and skills. The organisers themselves warned that the sample does not represent Swiss SMEs. It does match what we see in companies across Ticino, though: the technology is available; the hard work is adopting and governing it.

From the shop floor to the office: the knowledge lives in people

The most concrete case came from a long-established manufacturer, with ageing machinery and a great deal of manual work, that in recent years has brought robotics and computer vision into production and in-house software into the office, replacing spreadsheets. A journey made of small, fast prototypes rather than one big project.

What struck us most, however, was the part about people. To teach a system how a process works, someone first has to describe it. And often, the owner explained, that is the moment a company discovers it does not know exactly how it works: rules, checks and exceptions live in the operators' experience and have never been written down. Hence the choice of a gradual path that gives staff time to understand the change.

There was a practical warning too. If the company does not provide tools and training, employees will subscribe to whatever service they like and feed it company data. Better to give them something governed than to leave them to improvise. It is the same issue we found in the figures on AI adoption in Swiss SMEs: personal accounts and vague rules are the quietest risk.

Even a digital agent needs an owner

The session on people and skills began with a useful definition, taken from the ISO 56000 standard: innovation is something new that is adopted and creates value. A human resources manager from a local company then brought the reasoning inside the organisation. A memo or a directive does not change behaviour: people need new habits and support, and they need the skills to check what AI produces. The example of the meeting where someone challenges a colleague because "ChatGPT says so" drew a laugh from the room, but it describes a real problem.

The most interesting passage concerned digital agents, meaning systems that do not just answer but carry out tasks. The proposal is to treat them with the same clarity you would give a member of staff: they appear in the org chart under the person accountable for them, with a purpose, a scope, a spending ceiling and a way to stop them. You need to know which process stalls if the agent stops working and who can step in. Above all, decisions affecting customers, money and reputation stay human: the system proposes, a person decides.

It may sound like red tape. In practice it is what separates an interesting experiment from a tool a company can rely on, and it applies to automations far smaller than an agent.

The Third Way: innovating around what already works

The Rasmussen Consulting facilitator closed the plenary with the story of LEGO and the Third Way of Innovation model. In the late 1990s the company, alarmed by video games, launched products far removed from the brick and came close to closing down. The recovery did not come from another break with the past but from a series of complementary innovations built around the core product: stories, films, series and digital components that reinforce the building experience rather than replace it.

Hence the model: identify your "crown jewel", the product or service that embodies the promise you make to customers, leave it untouched, and work on the friction around it. The examples came from very different sectors: a paint manufacturer that gives decorators a tool for quotes and colour choice, a camera maker that looks after mounts and video sharing, a brewer that helps people open venues consistent with the product experience. The test is that every element reinforces the same promise: complementary ideas, not accessories.

For a Ticino SME with a solid product, that is a far more approachable question than "let's reinvent ourselves": what do customers already value, and what makes it a chore today to buy it, use it or order it again?

What we are taking home

For us the day confirmed the way we work. In short:

• Start from a process and a result to improve, not from the tool.
• Fit the automation into the tools the team already uses, to cut the friction of adoption.
• Involve the people who do the work in mapping rules and exceptions: that is where the knowledge lives, which is why every project of ours starts from an organised company knowledge base.
• Define human oversight, ownership and the way to stop things before going live, not after the first problem.
• Proceed one bounded use case at a time, measure, and only scale after validation.

That is the approach behind our automation services for SMEs: small automations resting on clear processes, with people who remain accountable for the result.

Our thanks to Fondazione Agire, the Swiss {ai} Weeks, Essere Agile and Rasmussen Consulting, the speakers and everyone who worked with us at the tables.

Frequently asked questions