Why AI Projects Fail: The Real Culprit Isn't the Technology

The numbers are unambiguous, and they contradict conventional wisdom: roughly 70 percent of AI projects fail — not because algorithms malfunction or systems run unreliably, but because the people expected to work with them were never prepared for the change. Any organisation that treats change management as a dispensable overhead cost when investing in AI will eventually pay the full price: failed projects, frustrated teams, and burned budgets.

The Real Problem: People, Not Machines

As early as 2015, McKinsey research found that 70 percent of all change initiatives fail due to employee resistance and insufficient management support — long before generative AI became a boardroom topic. The pattern hasn't changed with the AI wave; if anything, it has intensified. A Gartner survey from July 2024, covering 473 HR leaders, found that 73 percent of employees are change-fatigued, and 74 percent of leaders feel unable to lead change at all.

The pattern holds when you isolate technology projects specifically: a 2016 KPMG study found that transformations with a pure technology focus are twice as likely to fail as those with a clear strategic focus. Treating AI as an IT project rather than a leadership mandate sets the wrong course from day one — throwing AI at broken structures simply breaks them faster.

Why Swiss SMEs Are Particularly Exposed

A joint study by HWZ and Swisscom from September/October 2024, covering 123 Swiss SMEs, identifies three central barriers to AI adoption: skills shortages, regulatory uncertainty, and lack of internal expertise. Notably, none of these three barriers can be solved with better software — they require structured change management: training, clear processes, and leadership that provides certainty instead of amplifying anxiety.

This observation aligns with data from Deloitte Switzerland: only 24 percent of companies offer mandatory AI training, and daily AI usage in Switzerland ranks near the bottom globally. Without training, no organisation can expect competence to emerge on its own — and as of 2 August 2026, Article 50 of the EU AI Act makes AI competence a legal requirement, not an optional extra, for organisations with EU exposure.

The Cost of Skipping Change Management

An analysis published by mybusinessfuture.com puts the difference in stark terms: without structured change management, AI adoption rates hover around 15 percent — with a deliberate change approach, they climb to as high as 75 percent. The same source confirms the figure cited earlier: roughly 70 percent of AI projects fail, and the root cause is almost always how organisations manage people, not the technology itself.

Prosci, a research firm specialising in change management, confirmed in a 2023 study that 67 percent of change practitioners now use AI themselves — yet the biggest challenge remains constant: team acceptance, not the technology. Anyone who believes the next tool upgrade will solve the problem is confusing the symptom with the cause.

What Actually Works: From Top-Down Mandate to Ambassador Model

A frequently cited example comes from Siemens: instead of mandating AI adoption top-down, the company built an ambassador model in which employees from operational departments acted as internal AI multipliers. The result was an acceptance rate of roughly 80 percent — significantly higher than classic top-down rollouts achieve. The difference isn't in the tool; it's in who explains and models the change. This analysis of AI in project management explores why this success factor is so decisive.

The pattern holds at scale: a 2025 BCG survey found that 72 percent of employees already use AI in their daily work — yet only 26 percent of companies actively shape that change. The gap between individual usage and organised leadership is precisely where projects get stuck between pilot and production, never scaling into sustained value.

Change Management as Business-Critical Insurance, Not a Cost Centre

The error many leadership teams make is simple: change management gets treated as an add-on, the first line item cut when budgets tighten. The opposite is true. Change management is the insurance policy that ensures the underlying technology investment actually generates a return. Organisations that respond to AI with headcount reduction instead of workforce enablement compound the problem rather than solving it.

1. Early workforce involvement: affected employees become participants before the system goes live.
2. Visible leadership: executives take visible ownership instead of delegating responsibility to the IT department.
3. Internal ambassadors instead of top-down announcements: colleagues explain to colleagues what is changing and why.
4. Mandatory training instead of optional offerings: capability building becomes part of working hours, not an after-hours extra.
5. Measurable milestones: acceptance and usage are tracked with the same rigour as technical performance metrics.

The Swiss Approach: Precision Over Haste

Swiss companies traditionally hold an advantage they too rarely apply to AI transformation: thoroughness. The same rigour that secures quality and trust across Swiss business culture translates directly to AI adoption — provided it isn't undermined by time pressure or uncertainty at the top. This assessment of AI strategy for Swiss SMEs describes precisely how leadership uncertainty can stall progress at the executive level.

In the end, one uncomfortable but clear conclusion remains: technology is rarely the limiting factor. The question that determines success or failure isn't 'which AI system should we buy?' — it's 'how do we prepare our people to work alongside that system?'. Organisations that take this question seriously avoid the second, far more expensive investment: cleaning up after a failed project.