Who from the union was there when the digital system arrived

Nearly nine in ten people in Sweden say it is important that employees and their representatives are involved when new technology is designed and introduced. The share is 88 percent against 77 percent across the EU as a whole, and the highest in the union are Finland and the Netherlands at 90 percent, according to Special Eurobarometer 554 on artificial intelligence and the future of work.

The opinion, then, is not contested. What actually happens is harder to put a number on. Handels, the Swedish commercial employees' union, has asked its own safety representatives how often they are involved before a digital system is introduced or changed at the workplace. A safety representative in Sweden is a statutory role: elected by the employees, with a defined right to be consulted on the working environment.

Safety representatives are rarely in the room when the system arrives

In retail, 36 percent of safety representatives answer no, never. A further 24 percent answer mostly not. In warehousing and e-commerce the corresponding shares are 19 and 32 percent. The question was put plainly, asking whether the representative is involved before digital systems or tools are introduced or changed, with four options running from never to always.

The same survey asked whether risk assessments are carried out before digital systems are introduced or changed. In retail, 28 percent of safety representatives answer never and 24 percent mostly not. In warehousing and e-commerce the shares are 15 and 30 percent. Two questions, one pattern: the share of representatives answering never or mostly not falls between 45 and 60 percent, depending on the question and the sector.

The population is Handels' own safety representatives in retail and in warehousing and e-commerce, one sector rather than the whole labour market. That is also what makes the figures useful: they come from the people who would have been in the room.

Members are not against the technology

The usual explanation for the union arriving late is that the union slows things down. That explanation runs into a problem in the same survey. Asked how technological change affects their own work, 80 percent of retail employees answer very or fairly positively. In warehousing and e-commerce the share is 74 percent. Among the safety representatives themselves the shares are 75 and 66 percent.

Whoever negotiates about a system meets members who, in large majority, already want it.

That makes the absence more expensive than it looks. A union representative who arrives after go-live has two things left to do: demand correction and manage the consequences. One who is there during the design has a third option, and it is the only one of the three that costs the employer less than the other two.

The difference follows participation, not technology

The survey splits members by how they answer the statement that they are involved in how the business develops new technology and new ways of working. Among those involved to a high degree, 89 percent in retail and 96 percent in warehousing and e-commerce are positive about technological change. Among those involved to a low degree the shares are 76 and 65 percent.

Alongside participation sits competence. The share of members who say they have received training to handle new technology is 25 percent in retail and 30 percent in warehousing and e-commerce. Safety representatives were asked an adjacent question, whether there is enough time to learn new ways of working: in retail 13 percent answer never and 43 percent mostly not, in warehousing and e-commerce 12 and 41 percent.

These are two different shortfalls and they have different remedies. Participation is about when someone is called in. Competence is about what that person can do once the call comes. A safety representative who is at the meeting but cannot read what the system does to staffing has in practice left the question unanswered, and next time the representative is not called at all.

Exposure follows the size of the employer

How much of this touches a particular union is decided by where its members work. According to Statistics Sweden, 71.9 percent of companies with at least 250 employees used AI during 2025, against 49.6 percent of companies with 50 to 249 employees and 30.8 percent of those with 10 to 49. The survey covers companies with at least ten employees.

A union whose members mainly work in large companies therefore has a different exposure from one whose members work in small firms, and that difference can be calculated from its own membership records before any decision is taken about the scale of training. Which systems are involved is a separate question, and one that can be answered from the employer's own records.

The two measurements do not say the same thing about training

Special Eurobarometer 554 contains a figure that pulls the other way. Among those in work, 68 percent agree that their employer gives them the tools or training needed to work effectively with the latest digital technologies. The Handels measurement gives 25 and 30 percent. The gap is wide, and the explanation lies in the questions.

The figures do not measure the same thing. One is a judgement about whether what exists is enough, put to everyone in work across the EU. The other is a report of what has actually been provided, put to members in two sectors in Sweden. Someone who is managing can answer yes to the first question and no to the second without contradicting themselves. Which of the two figures matters to a union is therefore decided by which of the two questions the union puts to its own members.

The evidence has weaknesses. The Handels member survey drew 1,117 responses in retail and 600 in warehousing and e-commerce, from mailings to 10,000 and 5,491 members, so 11 percent in both sectors. The safety representative survey drew 21 percent. And the relationship between participation and attitude is a cross-section: someone already positive about new technology may well be the one who seeks out the work of introducing it, which would reverse the arrow.

Even among the least involved, a clear majority is positive, and that figure needs no causal relationship.

The objection about the direction of the arrow therefore leaves the simplest conclusion untouched. What the cross-section cannot settle is how much participation is worth. What it can settle is that the union does not have an opinion problem, and that is enough to move money from campaigning to competence.

The criterion sits in two numbers you already have

Both numbers are in the union's own records. The first is the number of safety representatives and elected representatives. The second is how many of them have taken a course on digital systems and AI in the past twelve months. The ratio decides which of two problems is yours, and they have different solutions.

If the ratio is high and attendance is still low, the problem sits in the timing. What is missing then is not training but a local routine triggered when a system changes and not only when a new one is bought, because a system already in place can gain new functions without a new purchasing decision.

If the ratio is low, the problem is the reverse. Attendance is then worth nothing until someone in the room can read what the system does to staffing, measurement and follow-up, and that training is bounded: it concerns the systems your members actually meet, not AI in general.

A third outcome is possible, which is that neither ratio can be calculated, because the union does not record which of its elected representatives has taken which course. That answer is also a finding, and it is the one to deal with first.