By Dr. Xenia Wade | The Human Side of AI at Work

74% of companies have yet to show tangible value from their AI investments.

BCG survey of 1,000 senior executives across 59 countries, 2024. 

And when BCG looked at why, many of the implementation challenges weren’t about the technology. They were about people and processes.

Most organisations already know this. And yet the most common response to slow mandatory AI adoption is still the same one: add another mandate.

The problem isn’t the mandate itself. Mandates can work. The problem is what usually gets skipped alongside them: the change management, the psychological safety, the meaningful involvement of the people who are supposed to actually change how they work. When that piece is missing, you don’t get adoption. You get what I call Checkbox Adoption.

What Is Checkbox Adoption?

Checkbox Adoption is what happens when mandatory AI rollouts produce the metrics of adoption without the substance of it. Usage is logged, training is completed and dashboards look healthy. But employees have found the path of least resistance through the requirement, and nothing about how they actually work has changed.

It’s not cynicism or laziness on the part of employees. It’s a predictable human response to change that’s been announced rather than managed. When people have no meaningful involvement in a transition that directly affects their work, minimal compliance is a rational response.

Getting the tool selection right is the easier part. What comes after the announcement is where most rollouts fall apart.

Is your organisation building genuine AI capability, or Checkbox Adoption? The diagnostic measures what your dashboards can’t. -> Take the Free Diagnostic

Why Mandatory AI Adoption Backfires Without Change Management

To understand why mandates without change management produce Checkbox Adoption, it helps to understand what the research tells us about how people decide to use new technology.

The Technology Acceptance Model, known as TAM, is the most replicated framework in this space. TAM identifies perceived usefulness and ease of use as the primary drivers of adoption. But the research shows those relationships change in mandatory settings. Huang and Hsu’s study of workplace users of mandatory IT found that when employees are forced to use a new system, workload pressure, perceived control, and anxiety become far more significant predictors than usefulness or ease of use alone. You cannot assume that findings from voluntary adoption research carry over cleanly to mandatory contexts.

Cieslak and Valor’s 2024 integrative review of 63 studies on employee resistance to digital transformation, published in Cogent Business and Management, found that resistance is rarely about the technology itself. It’s driven by perceived threats to resources, job security, workflow autonomy, and professional competence. A mandate that arrives without change management doesn’t resolve those threats. It amplifies them.

The BCG finding is worth sitting with: 70% of implementation challenges in AI rollouts are people- and process-related. Not technical. And yet most organisations allocate the majority of their AI investment to the technology. The people side is treated as a communication plan, or a training module. Not as the actual work.

There’s also a significant cost to high-performing employees. These are often the people with the strongest professional identity and the clearest view of whether AI outputs are actually trustworthy. When a mandatory rollout happens without psychological safety underneath it, they don’t engage more deeply. They disengage. Not from the tool. From the whole change programme. This dynamic connects to what I explore in my article on AI Shame: the private embarrassment of not understanding a technology that others appear to have mastered. Mandatory rollouts amplify AI Shame by removing the psychological safety to admit confusion without career risk.

What the BCG Data Actually Tells Us

74% of companies haven’t generated tangible value from AI. That number is striking, but it’s not because AI doesn’t work. BCG’s report is explicit: the organisations that are seeing real returns treat people and processes as the primary investment. They follow what BCG calls a 70-20-10 approach: 70% of resources on people and processes, 20% on technology, 10% on AI algorithms.

Most mandatory AI rollouts invert that ratio entirely.

When Li, Zhu, and Hua examined what separates organisations that capture AI value from those that don’t, writing in Harvard Business Review in 2025, the conclusion was the same. Most firms struggle not because the technology fails, but because their people, processes, and organisational readiness haven’t been treated as the actual work.

The research doesn’t say mandates always fail. It says mandates produce compliance without commitment when they’re not accompanied by genuine change management. That becomes a serious problem when organisations skip adequate training, ignore employee concerns, or deploy tools without demonstrating clear value in the employee’s day-to-day work.

The Checkbox Adoption Cycle

The pattern is predictable.

• Mandate issued. Leadership announces a company-wide AI requirement. Tools are deployed. Training is scheduled.
• Compliance achieved. Usage metrics hit their targets. Completion certificates are filed. The rollout is declared a success.
• Quiet resistance grows. Employees find workarounds. Outputs are rubber-stamped without review, or quietly ignored. The tool becomes shelfware dressed up as productivity.
• Capability gaps emerge. Six to twelve months later, the business asks why AI isn’t generating the promised return.
• The cycle repeats. A new mandate, a new tool, a new training module. And the same result.

The break in this cycle isn’t a better tool. It’s treating the human side of the rollout as the actual work, not the afterthought.

I use the concept of Emotional Carrying Capacity to describe an organisation’s genuine ability to absorb the psychological weight of change. Mandatory AI adoption doesn’t happen in a vacuum. It lands on people already navigating uncertainty, workload pressure, and real questions about how AI affects their professional future. Rolling out a mandate without first assessing that capacity is one of the most common, and most avoidable, reasons AI adoption stalls.