Last week I caught up with a CEO we’ve been working with for a few months. He was one of the first to see how AI would change his business. We’ve been working together on an interface that allows potential customers to design an outdoor environment that his company will build. Previously, he’d seen the design stage as central to his value proposition.

The seed for the project was planted by two changes the CEO had noted. When opening new territories, he had to adapt to different ways of working. He also needed a way to work with younger designers where his relationships were not as strong. Now his thinking has turned full circle.

He told me that AI would commodify products. When a decent level of intelligence is available to almost anyone, companies will use it to produce products of a similar quality. People will then buy from the people they like.

In other words, the need for a network and the reputation for being a trustworthy partner remain crucial in the age of AI.

The CEO's observation extends beyond his own business. It also explains why the biggest AI companies are racing to sell more than models.

OpenAI and Anthropic are tackling the issue of commoditisation. The pace at which models leapfrog each other’s capabilities is increasing, as the time between new model releases shrinks. The time it takes for open source models, most often Chinese, to distil frontier models and copy their capabilities is also coming down.

This creates what the newsletter AI As Normal Technology calls a commodity trap. Undifferentiated models, similar company capital structures, low switching costs and freely adjustable prices threaten profitability.

The authors don’t deny that AI will create tremendous value. They do discuss, however, who is likely to capture that value. The insight they offer is that we do not yet know how this will play out.

To return to an old analogy, AI models are the trains that run on the railroads. There is money in moving people and goods, but only once excess capacity has been flushed from the system. Even then, the real money is made by the people who set up businesses in new places and those who service their needs in the new cities.

To address this, both OpenAI and Anthropic are attempting to bolt enterprise products onto their models. If they can embed themselves into the way firms work, then they can extract the kind of rents that Google and Microsoft already do. The challenge is that Google and Microsoft are trying to do the same and have a considerable head start.

Both already have a wide network of clients who trust their products, services and security.

The same forces eventually reach smaller businesses. If tools become commodities, the advantage lies in relationships, processes and the reputation built around them. AI will reinforce trust rather than replacing it.

There are three levels of AI usage. The first is uncontrolled use of personal and approved workplace tools by employees experimenting and doing their own thing. I expect companies to clamp down on this in time due to rising costs and as they gain a clearer idea of what they want AI to do in their business.

At MSBC, we offer education for staff and coaching for leaders to support this transition to effective AI adoption. We focus on delivering outcomes rather than tick-box training to keep regulators at bay.

The second level of adoption is using AI within existing platforms. This is triggering a turf war, with software providers keen to monetise the secret sauce contained within platforms such as HubSpot, Salesforce and SAP. Meanwhile, those companies are developing their own AI capabilities.

Whoever wins, the intellectual property remains with the software provider, which will maximise its value at the expense of clients.

The third level is proprietary AI, built in-house or for you by a company such as MSBC. Where standard packages are used, the IP belongs either to us or to another third-party provider. Where we build operating systems that are unique to your business, you own the IP and the ability to maximise your returns.

In the case of our client CEO, he is already thinking ahead to what this means for his business. His competitors will build, or may already have built, competing products that imitate his offering. They will then look to undercut him.

The point of differentiation will be more than simply how much customers like him. He already supplies best-in-class components and can argue that cheaper competitors sacrifice quality.

He has also earned clients' confidence through years of delivering projects on time and on budget. The ways of working within his business, what we now call context and teach AI systems to follow, are already proven in practice.

When the CEO says people buy from people they like he is correct. But what goes into being liked in business has only a passing resemblance to what we value in our friends. The good news is that you can concentrate on this when running your business regardless of your technology.

The networks and trust that you accumulate will be there to reward you whenever you are ready to adopt AI.