AI starts with people

AI is no longer optional for businesses of any size. The question is not whether to adopt it, but how. And I believe that the organisations that don’t understand that are likely to pay a significant price for it.

 

Most organisations genuinely want AI to work, and they go about it in the same way: the tools are bought, the rollout is announced, the training session is booked into the calendar, and then everyone waits to see the transformation arrive. And then, largely, nothing changes.

 

A BCG (Boston Consulting Group) study in late 2025 found that just 5 per cent of companies achieve substantial value from AI, while 60 per cent report no material value at all. McKinsey estimates that generative AI could add up to $4.4 trillion annually to the global economy, which means the gap between what AI could theoretically deliver and what organisations are actually getting represents one of the most expensive missed opportunities in business right now. And that pattern stands whether you are a ten-person team or a ten-thousand-person organisation.

 

The instinctive response is to point at the technology and assume that maybe the tools were wrong, the vendor oversold the capability, the implementation was badly sequenced or rushed, and sometimes that is true.

 

However, more often the failure point lies with the humans.

 

The human element is less a problem to be solved; it is more a factor that was never properly accounted for in the first place, one that no amount of better tooling will fix on its own.

 

The wrong strategy

According to BCG, in successful AI-driven transformations, 70 per cent of the value comes from people-related actions rather than technology-related actions, meaning the technology itself is actually the smaller part of the equation.

 

That ratio will surprise most leadership teams because it is almost exactly the opposite of how AI implementation budgets are typically allocated in practice. The tools get the lion’s share of investment and attention and headspace, and the human layer, the behaviour change, the cultural readiness, the individual psychology of adoption, gets a one-day training session, a FAQ document, and an all-staff email that most people scroll past on a Monday morning.

 

This is not a communication problem but a behaviour change problem, and the two require completely different responses, timelines, and kinds of leadership.

 

As a behavioural intelligence strategist, I work with founders and senior leaders to identify the hidden patterns shaping how they make decisions, respond to change, and lead their teams through it. Behavioural Intelligence is the practice of understanding how individuals are wired, the unconscious patterns, emotional responses, and psychological drivers that determine how people actually behave under pressure, rather than how they intend to. This is the foundation that determines whether any strategy holds.

 

The pattern I see most consistently in AI adoption is that organisations that are struggling are doing so because they deployed technology without first doing the human work that makes it stick, and now they are trying to retrofit the cultural conditions that should have been in place from the start.

 

IBM’s guidance on AI-focused change management makes exactly the same point: that building trust, transparency, and genuine change agility across a workforce is not a nice-to-have layer on top of implementation, but the prerequisite for implementation to actually work. A workforce’s relationship with uncertainty, with perceived threat, with the anxiety of professional disruption, does not dissolve because you bought a licence or ran a well-meaning lunch-and-learn.

 

 

The people gap

There is a significant gap between what leadership teams assume about their workforce’s AI readiness and what is actually happening at desk level, and most organisations are operating on the wrong set of assumptions.

 

BCG’s research found that 76 per cent of executives believed their employees were excited about AI, yet only 31 per cent of employees actually felt that way. A separate BCG survey found that only 36 per cent of employees say they have been trained on the skills needed for AI transformation, which means nearly two-thirds are underprepared for what they are being asked to adopt and do differently every day.

 

Some people are experimenting but not saying so, because the culture around it feels ambiguous and they are not sure whether it counts as cheating or cutting corners. Some are using AI tools regularly but shallowly, dipping a toe in without getting anywhere near the full capability of what they have access to. Some are actively avoiding the whole thing, for reasons that range from genuine scepticism to real, unaddressed anxiety about what this change means for their professional identity and the skills they have spent years developing. And a meaningful number are somewhere in between, not opposed or resistant, just genuinely unsure where to start and waiting for someone to make it clearer.

 

They are actually entirely predictable responses to a technology that arrived fast, that nobody was formally taught, that carries cultural weight far heavier than most workplace tools, and that comes loaded with existential undertones that a training session does not address. When someone fears that AI might diminish or replace the parts of their job they are most proud of, asking them to embrace it enthusiastically is not a messaging challenge. It is a behavioural-change challenge, and it needs to be treated as one.

 

The gap between understanding what AI can do and actually doing it, what could be referred to as the confidence gap, is where adoption most commonly stalls. And it is worth being clear about what that gap actually looks like from the inside, because it is rarely simple reluctance. Most people sitting in it are not resistant to AI. They are capable of more than they are acting on, and their potential competence exceeds their current confidence. They are waiting to feel ready before they try, and without the right conditions in place, that moment never quite arrives.

 

What compounds this is how people tend to respond when they feel uncertain about something professionally significant. Under pressure, most of us default to one of three unproductive patterns - either defending the status quo, finding reasons why the new thing won’t work, or performing enthusiasm without genuine engagement. All three feel safer than honest experimentation, and all three keep people firmly in the gap. What actually closes it is the willingness to hold your current knowledge lightly, treat your first attempts as experiments rather than verdicts, and allow yourself to be a beginner in a domain where you are otherwise experienced. That approach does not arrive through instruction. It arrives through psychological safety and discomfort that feels productive rather than threatening.

 

This is where behavioural intelligence does its most important work. Understanding the patterns that govern how someone responds to uncertainty, whether they tend toward over-caution or overconfidence, whether they learn by watching or by doing, what their relationship with failure looks like, is what makes it possible to design conditions that actually move people. Closing the confidence gap is about creating an environment where trying feels safe enough to be worth it.

 

 

The risk

There is a risk in AI adoption that receives far less attention than hallucinations, data security, or job displacement, and in my view it is one of the most consequential ones for business leaders specifically.

 

The risk is that leaders and their teams lose the habit of thinking for themselves and do not notice it until the habit is already gone.

 

AI produces confident-sounding output quickly, and under genuine time pressure and cognitive load, the temptation to accept that output without properly interrogating it is entirely real and entirely understandable. Over time, the critical muscle, the one that says "Wait, does this actually make sense? Have I really thought this through?" can weaken from disuse, and the decisions that get made start to feel like they belong to the process rather than to the person. This is already evident in how some senior professionals describe their reasoning and decision-making when they reflect on it honestly.

 

The organisations that will genuinely thrive with AI are those whose people remain clear, confident, independent thinkers who use AI as a thinking partner rather than a replacement for thought, and that requires deliberate, consistent leadership around modelling critical engagement with AI outputs and making it explicitly clear that judgement remains a human responsibility, not something to be outsourced to the machine.

 

 

What to do instead

Three shifts matter most for businesses that want to get this right.

 

1. Start with psychological readiness, not tool readiness. Understand where your people genuinely are, not where you assume they are or where you need them to be for the rollout to stay on schedule. BCG identifies multiple overlapping barriers to AI adoption: skills gaps, real uncertainty about whether AI use is culturally acceptable within the organisation, and genuine grief about changes to professional identity that nobody is acknowledging. Each of these needs to be named and addressed directly, rather than assumed away or buried under enthusiasm for the technology. The confidence gap closes when people feel safe enough to try, fail, and try again in the context of their actual work.

 

2. Create real safety for experimentation. AI requires sustained trial and error, and if your culture, even without meaning to, treats mistakes as failures rather than data, your people will not experiment openly. They will use the tools in ways that feel safe and unnoticed, or worse, share confidential information or I.P. with an AI language model that isn’t as secure, and you will lose the organisational learning that comes from people genuinely trying things, getting them wrong, and sharing what they found. In the context of genuine skill-building, discomfort is a sign that something is going right. Creating the conditions for that kind of productive discomfort where trying and failing is framed as progress rather than incompetence is one of the most important things a leader can do.

 

3. Model it from the top, seriously and visibly. BCG’s research is clear that every leader should be able to describe one specific way they applied AI in the previous week. Change follows behaviour rather than instruction, and the most powerful signal a leadership team can send is to use AI openly, talk honestly about where it helps and where it falls short, and demonstrate through their own behaviour that critical judgement stays firmly in human hands.

 

The opportunity here is real and substantial. McKinsey’s $4.4 trillion estimate represents what becomes possible when AI is genuinely embedded in how people work at scale across an organisation, with the human conditions for adoption in place. The organisations that will seize that opportunity are not the ones that moved fastest to buy tools or launched the most ambitious rollout plans.

 

The competitive advantage in AI will belong to the leaders who invested in the human layer first, who understood their people well enough to bring them along, built psychological safety for genuine experimentation, and never let the machine become a substitute for their own judgement.

 

That is where it starts…and the rest follows from there.

 

---

 

Cat Paterson is a behavioural intelligence strategist and creator of behavioural intelligence methodologies that help senior leaders, in organisations of all sizes, identify the hidden patterns shaping decisions, performance and change.

 

Main image courtesy of iStockPhoto.com and gremlin