When intelligence becomes abundant

The thing that changed is the price of intelligence

What has changed enormously since then is the cost of intelligence. Reading information, understanding a situation, comparing alternatives, applying preferences, generating a recommendation or navigating a software interface are all becoming dramatically cheaper. The models are improving, of course, but the deeper change is that intelligence itself is becoming something that can be applied almost everywhere, at very low marginal cost.

That matters because a large part of the customer-service industry was built around the opposite assumption. Understanding a problem was expensive. Somebody had to listen to the customer, recover the context, interpret the company's policies, search through several systems and work out what should happen next. Because human judgement was scarce, entire operating models were designed around allocating it efficiently: queues, routing, skills, tiers, offshore locations, occupancy, average handling time.

The first response to AI has naturally been to make that model more efficient. Give the human agent a better copilot. Automate simple contacts. Improve self-service. Reduce handling time. Those are real improvements, but they still assume that the interaction itself remains the basic unit of the system.

I am increasingly unsure that this is where the larger change sits.

If intelligence becomes sufficiently cheap and abundant, then understanding the problem and generating a sensible answer may no longer be the hard part. Take the hotel example again. There is no particular reason why the intelligence required to decide that I should move from Hotel A to Hotel B needs to sit inside a customer-service department. My own agent could eventually know my itinerary, my preferences, how much I paid, whether I care about breakfast, how far I am willing to move and how much additional cost I would accept. The hotel may have its own systems. The booking platform may have another agent. Any of them may be perfectly capable of working out a good answer.

What remains difficult is making that answer real

Does the replacement room actually exist? Can the original reservation be cancelled? Who has the authority to make the change? And what happens if something goes wrong?

This is where the old CX Travel Network idea starts to look more interesting to me than it did in 2024. The network does not need to be the smartest participant in the system. Intelligence may eventually be everywhere. Its value comes from connecting intelligence to reality: seeing that something has changed, understanding what outcome is required, coordinating the different participants, obtaining the necessary authority, executing the actions and dealing with the exceptions when the normal path breaks.

Today, a human interaction often acts as the glue holding all of this together. The human understands the situation, moves between systems, applies judgement and gets the different pieces to line up. If intelligence becomes abundant, much of that reasoning can happen automatically, but the need for coordination, authority and execution does not disappear. In some ways, it becomes more important because there are now many more intelligent actors capable of proposing what should happen.

The same structure appears almost everywhere once you start looking for it. For years, CX has been defined mainly as the interface between a customer and a company. I am not sure that definition survives a world in which both customers and companies have their own intelligence. The interface may become much less important than the infrastructure underneath it.

That leads me to a broader interpretation of what an Operating Layer of Global Business could actually mean: It would not simply be a larger customer-service platform, and it would not be an AI layer placed on top of existing workflows. It would be much bigger and far greater: it's the layer through which human and machine intentions become real business outcomes. A layer that can understand enough context to know what is trying to happen, authenticate the parties involved, determine what they are allowed to do, coordinate across systems and organizations, apply policy, execute the relevant actions, handle the exceptions and verify that the expected outcome actually occurred. To me, this is much more valuable than the current setup.

I believe that once intelligence is no longer the scarce resource around which the whole architecture is organized and when it becomes one of the inputs, the scarce assets become access to authoritative state, permissions, trust, orchestration, execution and accountability.

Looking back, I think that is what I was beginning to sketch in September 2024 without really having the language for it. I thought I was drawing a more intelligent customer journey, but I think I may have been drawing an early version of a much larger system: one designed for a world where intelligence is abundant and where businesses still need a trusted layer capable of turning decisions into actions and actions into real-world outcomes. I'm very curious to see if CX will finally head toward that idea and evolve in this direction over the next few years.