The Layer AI Is Moving Into Now

Last year I spent several months inside a large professional services firm, helping them think through where AI fits in their work. Not as a technical question. As an organisational one.

What struck me was not what the tools could do. It was where the conversations kept landing: one level higher than expected.

From Task to Coordination

The first wave of AI at work was horizontal. Tools spread across teams. People drafted faster, summarised meetings, rephrased text. Useful, visible, and largely peripheral to how organisations actually make decisions.

That is changing. The tools are moving up.

Not dramatically, not all at once. But the pattern is consistent. AI is beginning to touch coordination: which cases to prioritise, what information to surface before a meeting, where process bottlenecks are forming. Work that used to live in the heads of middle managers or senior analysts is being absorbed into systems that run quietly in the background.

I have watched this happen in practice. Knowledge-intensive organisations, firms where expertise is the core product, are finding that AI does not just support individual workers. It begins to shape the rhythm of the organisation itself.

What Managers Actually Do

There is a useful way to think about this. Management, at its core, has always been an information routing problem. Information flows in, gets filtered, prioritised, and passed upward or sideways. Decisions come back down. The manager sits at the intersection.

AI is now capable of doing parts of that routing. Not the judgement calls. Not the conversations that require trust, context, and accountability. But the aggregation, the pattern recognition, the flagging of anomalies: those are moving into systems.

This is what the shift from hype to workflows actually means in practice. The visible, chatbot-style layer was the beginning. The deeper layer is less visible and more consequential.

The Delegation Question

When AI reaches this level, a different question becomes urgent: what do you keep, and what do you hand over?

I think about this constantly in my own work. Preparation, synthesis, structural analysis: I delegate those. Orientation, judgement, the decision about what actually matters in a situation: those stay with me. The line is not arbitrary. It follows accountability. If I have to explain a decision, own a relationship, or stand behind a recommendation, I have to have made it myself.

Vision, judgement, and creativity remain human not because AI cannot simulate them, but because they require someone to be accountable for them.

Most executives I speak with have not drawn this line explicitly. They use AI, sometimes extensively, but the question of where their own judgement begins and the tool's output ends is often unexamined. That is manageable when AI is a writing aid. It becomes a governance question when AI is shaping priorities.

The Quiet Governance Problem

The firms that are getting this right are not necessarily using more sophisticated technology. They are asking better questions about which decisions should remain human, and building that into how they use the tools.

This connects to something I have seen repeatedly: organisations where AI disappoints are often not disappointed in the model. They are disappointed in what the model reveals about their own knowledge infrastructure, the undocumented decisions, the expertise that lives in one person's head, the processes that were never written down. As I have written elsewhere, AI that is built in rather than bolted on requires something underneath it to be structured enough to be useful.

When AI enters the coordination layer, that structural requirement moves up a level too. The question is no longer just whether your documents are organised. It is whether your decision-making is legible enough for a system to support it, and explicit enough for a person to remain responsible for it.

Not a Tool Question

What I find interesting about this moment is that it is forcing a conversation that was overdue in most organisations: what does management actually consist of, and which parts of it require human presence?

That is not a question about AI capabilities. It is a question about the informal layer of authority and judgement that most organisations have never had to make explicit, because humans filled it automatically.

AI entering the coordination layer does not answer that question. It makes it impossible to avoid.