AI Adoption Has Not Stalled. It Has Gone Invisible.

The gap between what HR needs and what the tools offer has not slowed adoption. It has pushed it underground. Around 88% of HR practitioners are already using AI in their work, largely without guardrails, shared standards, or transparency about how it is being used. That is not a failure of practitioners. It is a signal that there is demand, but the systems around them have not caught up.

Where AI use is invisible, inconsistent, or poorly understood, accountability becomes blurred and trust is eroded. That matters enormously in a profession that sits so close to fairness, safety, and legal obligation.

The opportunity is real. The 2023 Harvard Business School and BCG study found that AI lifts the floor of capability: 25% faster, 12% more tasks completed, output ranked 40% higher quality. The question is not whether to use it. The question is whether we can apply it meaningfully and safely where it genuinely helps, while keeping human resources firmly human-led.

Deploy Without Preparing, Damage Without Intending

The pattern is everywhere. An organisation buys 100 Copilot licences. Microsoft makes it easy because most organisations already have 365. Performance plans get updated with AI efficiency targets. People are told to go for it. No context, no support, no space to learn. People are being measured on something they have not been equipped to do. The result is resentment, not transformation.

How AI Accelerates Fragmentation

Organisations have always struggled with departmental silos. Remote work scattered teams further. Now AI is creating a third layer: individual silos. One person can build a set of agents that runs a whole workflow, maybe even a whole division. That sounds productive until you consider what it means for subject matter expertise, quality governance, and the connective tissue that holds an organisation together.

Most People Cannot Articulate What They Do

This is where it gets interesting. AI is starting to expose something that has been hidden for decades: most people cannot clearly describe their own work. Years of unconscious competence have made their processes invisible, even to themselves. Remember knowledge management teams and continuous improvement departments? Organisations got rid of them because the work was not considered strategic. The irony is that writing SOPs, documenting processes, creating clear workflows is exactly what AI requires to work well.

If we are going to build AI integrations and custom agents, we have to be really clear about what the rules are and how the system works. The people who can describe their work clearly are the ones who get to reshape it. Everyone else gets reshaped by it.

Talkers, Users, and Shapers

Put 100 people in a room. About 70 are talking about AI or unaware of it entirely. They are not necessarily doing anything with it. Another 25 are using it, mostly for personal tasks or summaries, but AI is still leading the interaction rather than the person. Then there are the five. These are the people who are shaping how AI gets used: building systems, creating agents, designing workflows that extend their expertise. That is where the real value sits.

AI Belongs in Organisational Development, Not the Server Room

One of the things that keeps coming up in organisations is the assumption that AI sits with IT. It does not. AI is an organisational dynamics challenge. It has to run right across the organisation as a shared responsibility. The privacy questions, the narrative, the governance, the capability building: these are people problems first, technology problems second.

There are two layers to this. The first is AI doing the work: executing tasks, generating drafts, processing information. Most organisations fixate here. But the second layer is where the real capability lives: making sense of what AI gives you. That sense of judgement, that ability to read the temperature, to interact with the system properly and then interpret what comes back. That is what keeps HR human-led. If we focus only on the implementation and not on building people's capacity to make sense of the output, we will never realise the value we think AI is going to deliver.

If you just leave it to the top to make the decisions, it will never work because they are the furthest from the actual work. You need bottom-up experimentation, top-down direction, and middle facilitation. Everyone needs to be involved.

Five Experiments for This Week