I think the Team is doing it

Last week I wrote about what it sounds like when leaders haven’t found their way into AI yet and the various reasons hidden underneath; “I’m too busy.”

This is what it looks like inside the organisation.

If leaders step back, or have yet to step forward, the organisation does not stand still. It is always moving. And because nobody is watching, it moves in ways that are hard to see from the top and harder to correct later.

The most common pattern is fragmentation. Individuals and small teams start solving their own problems with whatever tools they can find. Someone in marketing is using one AI platform for content. Someone in operations is using another for rostering. Finance has built a spreadsheet workflow around a third. None of these decisions were coordinated. None of them were reviewed. And none of them are visible to the person who would normally be accountable for technology, data, or risk, because nobody asked.

A BlackFog survey found that 49 percent of employees now use unsanctioned AI tools at work. Across the organisations surveyed, 98 percent reported some level of unsanctioned AI use. And the most tolerant group? The C-suite. The same leaders who have not engaged with the tools themselves are the least concerned about how everyone else is using them.¹

This uncoordinated, widespread activity produces consequences that compound over time.

The first is a growing imbalance in capability and confidence. Everyone moves at different speeds: some enthusiastic adopters, some abstaining sceptics, some just quietly intimidated by the new technology. The gap widens within teams and between them. Two people in the same role produce work to different standards using different methods, and neither knows what the other is doing. Across the organisation, teams begin operating in AI silos, each building capability in isolation, each developing habits and workflows that are incompatible with the team next door. The organisational standard drifts through inconsistency, and nobody has a clear picture of where anyone sits.

The second is risk. Documents and proprietary data are being shared across unsanctioned platforms without policy, without visibility and without anyone governing what is leaving the building. But the exposure runs both ways. AI-generated content, analysis, and recommendations are finding their way into regulated contexts: financial reporting, HR decisions, customer communications. Nobody is asking about provenance, limitations, or liability, because the tools were never formally adopted and the governance that would normally accompany them was never established.

The third is that without leadership giving experiments purpose and permission to persist, they die quietly. A finance team handed a new platform will try it for a week, find it unfamiliar, and revert to the spreadsheet they have trusted for years. The gravitational pull of existing habits wins. The organisation thinks it tried AI. What it actually did was place a tool on a desk and walk away.

And underneath all of this, activity creates a false sense of progress. Things are happening. People are busy with AI. But none of it is building toward anything coherent, and nobody at the top can distinguish motion from direction.

The instinct when leaders recognise this is to reach for a strategy, define a policy and a governance framework, even scope a transformation programme. That instinct is understandable, but the first move is smaller than that. The first move is a conversation about what is already happening. Who is using what. Why. Where the value is showing up for them. Where do they see more value available. But also, importantly, where the risks are accumulating. What is being given away. Before you can direct the organisation, you need to see it clearly.

That conversation does not require a team of consultants or a big budget. It requires curiosity and some structured conversations. And if you are not sure where to start, or you don’t have the time, then reach out. I can help you work out what to ask and what to start thinking about when you get the answers.

¹ BlackFog (2025). “Shadow AI: The Hidden Threat to Enterprise Data.” Survey of enterprise employees across multiple industries. The 98% figure and C-suite tolerance finding are from the same study.