PowerXFrodx blog

AI vendor dependency: the advantage fades, the dependency stays

Written by Igor Pauletič | Nov 19, 2026, 11:31:00 PM

FrodX is an SAP partner. Mikro+Polo is our client. And in this story, I’m on the client’s side.

Marko Verdev, managing director of Mikro+Polo, recently described on LinkedIn how, after five years in SAP’s cloud, he expected a price adjustment of just over 17% and instead received an offer for a new five-year term at more than three times the annual price. Replacing the system would take two to three years. He has about six months.

On paper, he can leave. In practice, he can’t go anywhere.

The price increase affects the ERP, not the part of SAP we manage at Mikro+Polo. Another partner handles that, and we have no influence over it. I don’t know all the contractual details and can’t judge whether SAP’s offer is justified. But I do know what happens when switching suppliers is so hard that the supplier stops worrying about you leaving. That is no longer a negotiation between equals.

But this story doesn’t end with SAP. Right now we are rushing into a new dependency, one that may run even deeper. This time on artificial intelligence. And we have very good reasons for it.

Advantage first, then the bill

Dependency was not invented by the cloud. When companies started using electricity, they became dependent on it. Early adopters gained a competitive edge. Today, electricity gives no one an advantage, but the dependency remains.

It was much the same with the cloud. I remember when we bought our own servers, installed software, took care of backups and paid people to keep it all running. The cloud freed us from much of that.

Those who moved early could develop faster, with less investment, for a few years. Using the cloud is no longer a competitive advantage in itself. Yet the bill arrives every month.

Something else changed too. We used to depend on our own IT people. Today we depend on vendors who often have far more negotiating power than we do.

Here it helps to separate dependency on a technology from dependency on a single vendor. You can buy electricity from different suppliers. With a business system you have spent twenty years adapting to your company, switching is much harder.

The most expensive supplier is not necessarily the one with the highest price. It may be the one that is hardest to replace.

With AI, we no longer rent servers. We rent capabilities.

We handed infrastructure to the cloud. Now we are starting to hand the work itself to AI.

Why? Because it will do much of it cheaper, faster or better.

Whoever is first to genuinely change how they work with AI can build a significant advantage. But that advantage won’t lie in access to a model their competitors can use too. It will lie in the processes they build with it.

Once others work the same way, much of that advantage will disappear. The dependency on the technology will remain.

I’m doing exactly what I’m warning about. Our voice agent, Kinetara, talks to callers using OpenAI models. If OpenAI raises its prices significantly tomorrow, either we or our clients will pay the difference. We can switch models, but that means making adjustments and checking again that the system works as it should.

Every day we build a product on technology whose future pricing we can’t predict.

And we do it knowingly.

Cheaper tokens, bigger bills

A future AI price increase may not even be the biggest risk.

According to the OECD, quality-adjusted prices of language models fell by nearly 80% between January 2024 and April 2026.

Tokens, the units AI is billed in, may get cheaper every year. Our bills may still grow.

Today AI writes ten emails. Tomorrow it processes a thousand requests. The day after, it does work that used to need a whole team.

Tokens get cheaper, but tasks don’t necessarily, because AI uses more and more tokens per task. The total bill can still grow. And with every new task we hand over to AI, being without it becomes a bigger problem.

Still, money is not what worries me most.

When we forget how to do the work

With any luck, I’ll soon be driving a Tesla. I’m not even particularly fond of the car. What excites me is the idea that its driver-assistance system will take over as much of the driving as possible. For now, I still have to supervise it. But after five years of driving less and less myself, will I be as good a driver as I am today?

I’m not sure.

Researchers have already seen a warning sign in medicine. After a period of working with AI, experienced endoscopists detected adenomas in 22.4% of examinations done without it, down from 28.4%. That’s a fifth fewer. The study doesn’t prove that AI caused the drop, but it raises an uncomfortable question (The Lancet Gastroenterology & Hepatology, 2025).

If AI does the work for us every day, will we still know how to do it ourselves in five years?

Think of a salesperson who hasn’t prepared a proposal without AI in years. Or a company that automated customer support and then found that nobody understood the whole process any more.

With the cloud, we gave up our own servers. With AI, we may give up the knowledge of how to do the work.

And you don’t get that back by clicking “export data”.

Build your exit before you need it

I don’t think this is a reason to slow down AI adoption. Quite the opposite.

Whoever uses it to improve their business today can enjoy, for a while, an advantage their competitors don’t yet have. Whoever waits three years may get the same technology, but not the chance to use it to pull ahead.

But the transition has to be handled so that the competitive advantage stays inside the company and we don’t become dependent on a single vendor.

Marko Verdev advises negotiating the exit terms when you sign. With standard AI services, we often won’t even have that option. We’ll click “accept”.

So we have to build the exit ourselves. Data, processes and instructions must remain ours, not locked into one model. The business must be able to survive a change of vendor. And we must keep enough knowledge to understand and supervise the work AI will be doing.

That’s why I’ll still drive myself now and then.

And when a bill three times higher arrives, you may no longer have enough time to leave.

igor.pauletic@frodx.com

More on planning a digital transformation that keeps data, processes and know-how inside the company.