This is the part of AI transformation nobody puts in the strategy deck. The tooling is the easy bit. The hard bit is that the people you need most are often the ones most convinced this is being done to them rather than with them. And in a Swedish workplace that conviction is fatal, because our whole model runs on trust and on change being negotiated, not imposed. Here is how we walk Nordic leaders through it, in three commitments.
Commitment one: sign the contract before you ask for anything
Not a legal contract, though in Sweden you already have one of those. Co-determination is not a nice-to-have here. It is MBL and the samverkan model, and it means change lands best when people feel consulted rather than restructured. Use that to your advantage. Say out loud, in public, what you will and will not do to your people.
The fear is simple and rational: this is coming for my job. There is still no consistent evidence at a national scale that AI is cutting employment, and Swedish labour protection makes the raw layoff story less acute than the American headlines suggest. But the anxiety is real anyway. People read the same global news about founders cutting thousands to focus on AI, and they translate it locally into omorganisation, into who gets moved, retrained or quietly sidelined. No aggregate statistic touches that worry.
So address the elephant directly. The leaders who do this with integrity say something concrete: we are keeping the company lean, so our plan is to slow the pace at which we hire, not to cut the people who are here. That is honest about headcount without threatening anyone currently in the room, and it is exactly the kind of commitment a consensus culture rewards.
Then reframe the whole thing. Jensen Huang has been blunt that he expects enormous productivity gains from his engineers and is letting go of none of them. He has said that leaders who cut staff simply lack the imagination to use their teams in the age of AI. That framing lands, because it moves the conversation from taking something away to expanding what you can go after. Your job as a leader is the wider vision. AI is the productivity engine underneath it. Say that, and you have their attention and their trust.
Commitment two: pick one narrow thing that moves the bottom line
"We are becoming an AI company, everyone start doing AI" fails twice over. People do not believe it, and they have no idea what it means for their Tuesday.
Here the Swedish setup helps you. These are small, lean organisations with unusually high digital maturity, reasonably clean data, and people who are comfortable in English and in new tools. You do not need a moonshot. You need one narrow area that genuinely moves your numbers, whether that is expanding revenue or cutting tooling costs, so the motivation is real. Tie it to something AI has already proven it does well. Never make your first project an unsolved research problem.
Customer service is a common first pick: routine cases move to AI agents, complex cases stay human, and AI works the CRM and policy records in the background so every human conversation is a good one. Engineering projects are the other common start, chosen because teams want to learn AI-native ways of working. Either is fine.
The trap is how you define success. If success is "people used the tool", you have already lost. Uber said publicly this year that it regretted how much it was spending encouraging AI usage, then started talking about token budgets. Imagine being told use AI, use AI, finally doing it, and then being told to stop because of cost. Now leadership looks like it has no idea what it is doing. Define success as the way we work changed and the customer felt it. That is tangible. Usage is not.
One more thing decides everything: you need a passionate champion in middle management, where the work actually happens. Directors and VPs can talk all they like in the steering group. If the team-level leader is not genuinely excited, nothing ships. And leaders have to be visibly on the journey themselves, learning the tools and talking openly about what works and what does not. AI transformation is a personal change as much as an organisational one, and in a flat Swedish hierarchy people will watch whether you actually walk it.
Commitment three: find the harsh ground truth before you scale
AI agents keep taking on more complex work, which means your harnesses have to evolve to match: how you call tools, how data flows, how information is processed across the whole business.
Before you expand anywhere, get honest about whether the pilot actually helped. Most teams do one of two wrong things. They persist out of stubbornness, "it's AI, we just have to push through", or they take one bad rollout as proof the whole thing is a dead end and walk away. Both skip the actual question. Was there real managerial commitment? Were people taught how AI works, where it fails, and what good looks like in their role? Did the tool do a job that genuinely helped, or was it another Copilot licence nobody opens?
Get to that truth first. Then either carry the lessons somewhere new, or scale deliberately, one team at a time, with the technical implications mapped for your actual stack. The technical details always turn into people details. The moment your memory files are markdown, someone asks "so do I still write in the wiki?" You are not rolling out a tool. You are evolving a capability where humans and AI agents work together, and that sentence is worth saying out loud.
When you widen it, lead with the customer impact you saw, then the business case, then the long-term vision for how information flows, without drowning people in architecture. And with GDPR and Swedish data expectations in the room, name your safeguards plainly: this is how we evaluate agent output, this is what agents cannot touch, this is where the data lives and stays. That is not bureaucracy here. It is the price of trust, and trust is the thing your rollout runs on.
The human edge is not shrinking
The value is real and we can see where it is: coding, faster iteration across writing and analysis, deep work like a twenty-tab spreadsheet modelling a position. But if you want something professional, a human still takes out the LLM-isms and checks it against the actual strategy.
There are plenty of voices telling you the agents just keep getting better forever until there is nothing left for humans to do. We do not see that world. We see roles changing fast and an enormous amount of human work, because figuring out how to work well with these models is the hardest thing the corporation has faced in five hundred years. Engineers are becoming system designers who write evals and push agents against a standard until the code holds. That is not less work. It is different, harder, more human work, and it plays straight to a Swedish strength: small teams that take real ownership.
That is the whole answer to "what do you do when your people hate AI". You commit to protecting them, you start somewhere small and real, and you tell the truth about where the human edge lives. People will move mountains once they can see where they fit in the picture.
That is also where emberloom.ai comes in. We help you pick the narrow first project that moves the bottom line, build it with the safeguards your people and your regulators expect, and take operational responsibility with SLA. The client buys the outcome, not the anxiety.