When an AI initiative stalls, the post-mortem usually blames the tool, the data, or the vendor. In my experience the cause sits elsewhere. AI initiatives in mid-sized companies fail for the same reasons as any other change programme. Nobody at the top owns them, they start in the wrong place, and the leadership team never builds real conviction.
What this edition argues
- AI delegated to the IT function becomes a tool rollout. The decisions that determine its value are business decisions, and they belong to the leadership team and the board.
- The right starting point is not where the technology impresses. It is where the economics improve fast enough to matter, and over time where AI strengthens something the company already does better than others.
- Adoption is a leadership task. Time saved is not value created until someone decides where the freed capacity goes.
The thesis
AI initiatives in Swiss SMEs rarely fail on technology. In my view, the tools are mature enough and affordable. The initiatives fail through a sequence that every experienced leader will recognise from earlier change programmes.
The sequence starts with delegation. The leadership team treats AI as an IT topic and hands it to the technology function. The technology function, reasonably, selects use cases that show what the technology can do. A pilot impresses in a demonstration and changes little in daily work. Employees are not convinced, because nobody explained what the change demands of them or what it makes possible. The time the tool saves dissolves into the working day. A year later, the company concludes that AI does not yet work for a business of its size. That conclusion is wrong. The technology was never the constraint. Leadership was.
Ownership: AI belongs in the boardroom
The first failure is a framing error. AI is not an IT topic with strategic implications. It is a strategic topic with technology underneath. When a board receives an AI roadmap from the head of technology and approves it politely, it gives away a decision that belongs to it.
Three questions belong at governance level, and none of them is technical. Where does AI create a real advantage for this specific company? What are the risks around data security, quality, and dependency on vendors, and who governs them? And how does the company build its own capability over time, instead of buying every step from outside providers?
The governance question is more urgent than it looks. Every AI tool a company uses processes its data: customer data, financial data, strategic documents. In my conversations with SME boards, the question of where that data goes is rarely asked in a systematic way.
Test for this in your organisation: Who presented the last AI decision to your board, and which of these three questions did the board actually discuss?
Starting point: choose the problem before the tool
The second failure follows from the first. A technology function asked to “do something with AI” will start where the technology impresses. A leadership team that owns the topic starts somewhere else: where human time is high and human judgement is low. Proposals, tender responses, meeting summaries, and internal reports follow recognisable patterns. Commercial tools can already reduce that workload, without custom development and without specialist hires.
The opposite error is just as common. A company decides it needs an AI strategy and commissions a comprehensive assessment. Months later it owns an impressive roadmap and no operational result. One team, one task, one tool, and a thirty-day experiment teach more than any strategy document.
The starting point has a second layer. Efficiency gains that every competitor can also achieve are worth having. But they are not an advantage. The advantage comes when AI strengthens something the company already does better than others: an exceptional customer relationship, proprietary data, a process that is hard to copy. Generic use cases become commodities quickly. Specific ones, tied to capabilities the company already owns, are where lasting value sits.
Test for this in your organisation: Can you name the one task where your first AI experiment will run, and the one capability that AI should strengthen over time?
Conviction: adoption is a leadership task
The third failure is the most familiar. Most change programmes over-invest in process and under-invest in conviction. AI initiatives repeat the pattern. Licences are bought, training is scheduled, a rollout plan exists. What is missing is a leadership team that can explain why the change matters, what it demands of each person, and what becomes possible on the other side.
I learned how much this matters before AI was on any agenda. In an international retail group, I led a large digitalisation programme. For the first time in the company’s fifty-year history, data from the European store network was collected and made available to the executive board and the business functions. My team went further. We derived the commercial implications and turned them into clear recommendations. Then the executive responsible for both category management and marketing looked at me and asked: “What does this data have to do with me?” For a moment I was speechless. The data was sound. The analysis was sound. And the most important commercial function in the group had decided not to use either.
It was not an AI project. But it was the foundation every AI initiative depends on, and it stalled for the reason AI initiatives stall today: one member of the top team did not see why the change concerned them. Conviction is only half of the problem. The other half is capability. A leadership team that cannot read what the data says will not act on it, however clear the recommendation.
If people experience AI as a threat rather than as help, adoption stalls, however good the tool is. In my case the doubt was spoken openly. More often it stays private, and the organisation still notices within weeks.
Even successful adoption hides a trap. Time saved is not value created. If nobody decides where the freed capacity goes, it spreads silently across the working day, and the business case stays on paper. Redirecting that capacity – towards customers, towards selling, towards work that was always postponed – is a leadership decision. No tool will make it.
Test for this in your organisation: When your first AI experiment saves time, who has already decided where that time will go?
Three questions for your next leadership or board conversation
- Who in our leadership team owns AI as a business decision, rather than as a technology project?
- Where does AI strengthen something we already do better than our competitors, and where does it only make us more efficient?
- When AI frees capacity, who decides where it goes, and how will we see the result?
AI does not fail Swiss SMEs on technology. It fails them on the three things that decide every change programme: ownership, starting point, and conviction. The companies that treat AI as a change programme will see the return. The rest will buy tools and wonder where the value went.
