How AI really lands in a business - with the people meant to work with it every day. Six points on the question that rarely comes up in the sales meeting, but decides every project.
An AI solution that your employees do not get behind will fail. No matter how good it is technically.
This is not soft talk - it is experience. Anyone who builds software and integrates AI quickly learns that the hardest part is not the model, not the interface, not the data structure. It is the people who are supposed to deal with it - or not. This page is for everyone who wants to take the topic seriously.
When you announce AI in a staff meeting, everyone listens. No one asks. But everyone thinks the same thing: "Will I be replaced?"
You have to answer this question before it is asked - and honestly. If AI does not cost any jobs at your company, say so and say how you know. If AI means less routine work and frees up time for other things, say that too - and say what that other thing concretely is. If AI really does take over tasks someone does today, say so openly and talk about how that person will be deployed in future. Silence is read as confirmation.
The truth is usually less spectacular than expected. AI rarely replaces whole jobs. It changes individual tasks.
A concrete example: someone who used to spend two hours a day sorting and answering customer enquiries now spends twenty minutes checking and correcting AI suggestions - and can use the other hour and a half for real conversations, for advice, for the places where humans are still clearly better. This mix will settle in across many professions. It is not a threat - but it is a shift. And shifts need guidance.
In every business you will encounter roughly three reactions.
The enthusiasts. Usually young, often tech-savvy, already playing with ChatGPT privately. They want to start right away. Your risk: they take the AI more seriously than their own judgement and overlook errors. Your lever: give these people responsibility, but with clear limits - four-eyes principle, regular spot checks.
The sceptics. Often experienced, often older employees, often the ones with the most valuable knowledge. They distrust AI - sometimes rightly. Your risk: they block, because they feel overlooked. Your lever: ask them first what they do not believe about the AI. Then show how their experience makes the model better.
The quiet ones. The largest group. Say little, go along with it. But not won over inside. Your risk: they use the AI minimally, find workarounds, hide their scepticism. Your lever: create spaces where questions can be asked without losing face - small groups, anonymous surveys, a clear learning curve without an exam.
Classic AI training fails. Three days of theory, lovely slides, everyone is impressed - and after two weeks no one uses the tool. The reason: knowledge without application evaporates.
What works looks different. A short introduction - one hour, not three days. Then immediately a real task from everyday work, solved together. Then a week of working on your own. Then a second meeting, in which everyone names one place where the AI was good and one where it failed. Learning happens between the meetings, not in the meetings. We build our guidance exactly that way.
One of the most common sources of conflict: management introduces an AI without asking the team. Suddenly an algorithm decides which enquiries are important, which emails take priority, which candidates are screened. The people who used to decide for themselves feel disempowered.
The clean solution: whoever does a task today should have a say in how the AI supports it tomorrow. Not because that is democratic, but because it works. Whoever helped build it corrects it afterwards. Whoever was overlooked refuses. With every one of our projects, a workshop round with the actual team is part of the deal before we really build in the AI.
The key term is the Technology Acceptance Model (Fred Davis, 1989). Two factors decide whether a tool is accepted: perceived usefulness and perceived ease of use. Both are subjective. An objectively good AI that employees do not believe is useful to them will not be used.
Later studies brought out the third factor: co-creation. Whoever was part of the development - even if only through honest questions in a workshop - accepts the result. The predictive power of the co-creation factor is astonishingly high.
With every AI project we start, before the first build we run a round with the people who will operate the software later. Not with management - with the people on the front line.
We ask what they do today, what annoys them, what they want to keep, what they do not want to give up. We listen. We take notes. And we design the AI so that it serves those answers - not the fastest efficiency calculation. That makes our projects a little more expensive, but considerably more successful. An AI that gets discarded costs more in the end than any extra workshop hour.
An AI is only as useful as it is used. And what gets used is what people understand, support and can influence.
Anyone who forgets the people when introducing AI buys technology that stands in the corner. Anyone who takes the people seriously - before, during and after the introduction - gets an AI that truly becomes part of the business. The order matters.
Where to even start with AI is under Where to start with AI?. What AI can realistically do today is under What AI makes possible and What AI cannot do. How smaller businesses change overall is under What does the future of companies look like?.