Why AI Projects Run Into Trouble Long Before Implementation Starts

Implementation rarely fails because of technology. It usually struggles because organisations prepare systems before preparing people for meaningful AI adoption. My aim is not to suggest that technology is less important but to encourage a broader discussion about where successful AI implementation really begins.

The answer may depend less on the technology organisations buy and more on the people they recruit, develop and support.

The role of artificial intelligence on the boardroom agenda has increased significantly. Companies invest in generative AI, intelligent automation and predictive analytics in order to increase efficiency and drive innovation and competitive advantage. Yet despite considerable investments, many organisations face difficulties moving from successful pilots to wider implementation. AI demonstrates great results at the pilot stage, works well in the trial and is enthusiastically received throughout the organisation. But then progress stalls.

In such cases, it is usual to refer to common explanations for the problem. Technology was not yet mature enough. The quality of data was insufficient. Too late, governance was introduced. Change management was difficult. While all these aspects do influence implementation, they may not necessarily explain where the problem started.

Maybe AI projects run into trouble long before they start implementing the first piece of technology.

AI is More Than Technology Investment

Most companies continue treating AI as a technology initiative. The first discussion focuses on software platforms, infrastructure, cybersecurity, implementation partners, budgets, and expected return on investment. When all these issues are sorted out, the focus moves to people. People attend training sessions, get guidance and are encouraged to experiment with the new technology. It looks quite logical. But it also implies that the organisation has the appropriate workforce. Technology can usually be acquired within weeks or months. Creation of organisational capability takes longer.

People bring different experiences, skills, and expectations to AI. Some use AI on a daily basis. For others, it remains a theory. Even in the same department, people use AI in completely different ways. While some people check every answer of AI, others trust every result.

These differences will not appear during implementation. Sometimes, they reflect decisions that the organisation made months or even years ago.

Decisions Made Ahead of Time That Shape the Success of AI. 

Each organisation continually shapes its future workforce through recruitment and onboarding. 

Job descriptions define the skills and experience required. The recruitment process defines who joins the organisation. Onboarding affects the way people understand what is expected of them.

But most organisations continue recruiting people for roles that existed before AI became a reality.

Technical skills and professional experience are still important but no longer sufficient. Now, more and more organisations need people with characteristics such as learning agility, curiosity, collaboration, and professional judgement. People who are ready to learn and adapt constantly to changes brought by technological innovations, rather than rely only on their existing knowledge.

These skills are usually not developed within the framework of a short implementation project. They start with recruitment and are developed throughout the onboarding process.

A Klarna Story

Klarna became one of the most prominent examples of fast AI adoption with the introduction of generative AI into its customer service department. Significant productivity gains were reported, and it was shown that AI can deal with a large share of customer requests.

But the company's CEO, Sebastian Siemiatkowski, later made public statements that showed the reality was more complex. He emphasised the importance of human expertise, empathy, and customer relations, and the company decided to increase its investment in people who work with customers. The lesson was not in the failure of AI technology. Technology brought tangible benefits. The lesson was about the importance of organisational capability in generating long-term value.

And this difference is crucial. AI implementation is not just the implementation of the new technology. It is the creation of an organisation where people know how to work with this technology.

Another Approach to AI Adoption

Consider two organisations that purchased the same AI platform. One organisation implements AI technology. People are given access to the system, take the training course, and are encouraged to use the technology independently.

Another organisation starts much earlier. Prior to starting the implementation process, it evaluates its job descriptions and ensures they reflect AI-enabled work. The recruitment process identifies not only the technical competencies but also people's ability to learn and adapt. In the course of onboarding, everyone learns not only how to use AI technologies but also how they can help to achieve organisational goals and support its decision-making processes.

Two organisations own the same technology. One of them intentionally prepares its workforce to use it.

The difference between them is noticeable over time. One organisation develops its working practices, builds confidence and ensures reliable implementation. Another one faces inconsistent implementation, differing behaviours across teams, and slow progress despite continued investment.

Three Questions for Executive Teams

Ask yourself three questions before you approve the next AI initiative:

Ø Are your roles designed to reflect our organisation, where AI becomes a part of everyday work, or are they still based on old assumptions?

Ø  Is our recruitment process identifying people who can adapt, collaborate and exercise professional judgement, in addition to having technical competencies?

Ø  Does our onboarding process explain how AI can be used in our organisation or just provide access to new technology? It shifts the discussion away from software toward organisational capability.

Final Note

Artificial intelligence does not generate business value by itself. An organisation creates value when technology and people evolve together. Maybe the first question to ask is not whether the organisation is ready to implement AI technology, but whether it has already begun building the right workforce.

In next week’s post, I will address why Technology Readiness and Workforce Readiness are two distinct capabilities and why misunderstanding them might be the hidden problem behind many failed AI projects.