AI Transformation Happens in Phases, AI Operating Models Will Change

If you have ever worked with an AI agent, you know that it can fill in forms. It can use various sources of information and then it knows how to fill them in. The form is sent to someone who in turn reads it into a system. Then the question inevitably arises: why the form? In the future, my AI agent will have to send or feed the requested information directly to the other party.

That is the logical next step.

For all of this to come true one day, we will now spend many years codifying our real world, writing down the meaning of our world, turning it into semi-structured data. The relationships between meanings as well.

It will take many years until everyone has done this, so that AI agents on both sides can make use of it.

It follows: AI transformation happens in phases.

The Phase of Confusion

Right now we are in a phase of confusion and panic. And of experiments.

Big tech and AI companies already offer personal AI agents. Sovereign AI agents running on local AI have also existed for a long time.

These AI agents keep getting better.

Anyone who is smart is already starting to engage with semantics, context, ontology, knowledge graphs and all the buzzwords.

But really, these are classics from library science, from data modelling and other fields, reissued and recombined.

Everything AI agents should be able to do in the future must be written down and codified as meaning.

Some companies are already doing this. The big tech and AI companies certainly are. What are you doing?

The Phase of New Companies

Large companies that are too slow to change may be left behind by new companies.

That is the next phase. New companies that form from scratch, with new roles, new processes, new structures, and codify everything right from the start, can work with far less effort and fewer people and can offer entirely different services.

The old-school company still asks its customers to fill in forms, while the small new competitor lets its customers transmit the information via their AI agents, or offers even more direct services for their AI agents.

This phase will again take a few years, until at some point the new companies have overtaken the old ones.

After this phase, the world will look completely different. That is how I imagine it.

And the Platforms?

I can also easily imagine that the platforms we use today to manage our lives and our companies will no longer exist at all, because you can do so much with local AI, so why would I still need a central place?

The data, the codified meaning, has to leave the platform anyway so that AI agents can work. The UI is already gone, keyword headless AI.

So all I need is discovery and trust. And I can distribute that across several places too.

I think we should picture these phases for a moment and consider what they mean for today.

Hierarchy No Longer Fits

A hierarchy with a chain of command like in the military makes sense when it is absolutely clear what needs to be done. When every move has to be precise. When everyone must know exactly what is required in which situation, how to react, who gives whom which instructions.

This operating model has not fitted for a long time. It did not fit before AI. It fits even less now. We are all experimenting. Nobody knows what to do. Surely there are consultants who want to give you the impression that they know. Which is nonsense. Everyone started engaging with AI at roughly the same time. The lead of a few is small.

The only justified lead some have, and I count myself among them, is experience in many different fields AND AI, and the ability to combine this knowledge to derive new insights.

In the phase of confusion we are in now, a hierarchical org chart is of little use. Nobody knows exactly what the next steps should be. So all that remains is to experiment and learn.

If you now imagine that new companies will emerge that have codified everything AI needs and can thus offer customers new things, a hierarchical concept does not fit either.

The interactions between customers' AI agents and companies' AI agents iterate so fast; should you really go up and down the chain of command? Far too slow.

The hierarchical model assumes that the superior knows more and has more information than the subordinates. That has not been the case for many years.

What Is Needed Instead

In the phase in which we codify the world, we need direct collaboration between those who hold the meaning and those who build it and those who keep it running and those who secure it and those who know the customer, etc.

None of these people knows in advance how it will turn out. So hierarchies are of too little use here.

What you do need is support, enablement, setting a direction, setting boundaries, holding the focus. And much more.

That is something different from a hierarchical org chart. Do you agree?

Outlook

What an AI operating model will look like after this phase, honestly, I have no idea.

But I know for sure that all the existing knowledge about ways of working, methods and organizational design is a good foundation. It widens the field of vision and the horizon. That is the first step.

And then it is about experimenting, learning and developing further as an organization.