Nobody inherits a fractured tech stack it turns out they’ve mostly commissioned them.

You may need to read that twice, I’ll wait.

You've seen the slide, I sure as hell have because I’ve built plenty of them. Boxes, arrows, a spaghetti of integration points rendered in angry red. It's the standard opener for anyone selling transformation to a large established tier one telco.

That's them. Years of acquisitions, regulatory retrofits, a CRM built for enterprise accounts nobody sells to anymore.

But when it goes up int front of the independent MVNO operators their response is different.

“Not us. We started in 2019. We're clean, we're cloud, we run lean and we launch fast, and the reason we launch fast is that we don't carry any of that tech debt.”

Fair enough. It's true today. Give it four years. Ok… (glances over shoulder at Claude) maybe two.

The clean slate has an expiry date

The diagram often drawn is really a time lapse. Every box on it was added by someone competent, solving a real problem, against time pressure, using the best option available that quarter. Twenty years of saying yes to things worth saying yes to will produce that picture reliably.

What should worry this market is the compression. Telcos took two decades to build that slide. Plenty of MVNOs are on track to build the same thing in four or five, because the tempo is higher, the margins are thinner and there is real pressure to make AI a fundamental part of the business. Am I right?

You needed to onboard a second wholesale partner, so you bolted their API onto the provisioning flow. Your billing platform couldn't handle eSIM the way you needed it to, so you wrote a service alongside it. Marketing wanted a customer view the ops system wouldn't give them, so you stood up a warehouse and started syncing. The second brand needed its own portal, its own logic, its own edge cases. Then came the third.

Every one of those calls was correct. Good decisions, made under time pressure, with nobody in the room representing the decision after next. Complexity arrives as a by-product of competence, which is why competence on its own won't keep it out.

The audit that will lie to you

The usual advice is to map your systems and count your dependencies. Run that exercise four years in and it will come back reassuring, which is exactly why it's the wrong instrument. Three experiences need to be considered…

Your last launch, against the one before it. How many people were in the room? How many systems were touched? How many days from decision to live? Ask your ops lead at 3pm on a Thursday and use whatever number comes back. Then put the two launches side by side and see which way the line points.

Your exception handling. When a billing anomaly hits, does it follow an automated path, or does it follow a person? When that person is on leave, is there a second person, or is there a spreadsheet? A human in the loop at year four is a dependency that nobody has written down yet.

Your data foundations. Everyone in this market is fielding AI questions right now: churn prediction, personalisation at scale, automated provisioning. The one to answer honestly is whether a customer exists once in your business or four times, and whether those four copies agree with each other. Nothing you build on top will fix that. It’s foundational now.

None of this is a gotcha. These are the three places where a young operator finds out whether it is actually simple or merely early.

The decision worth making once

The operators pulling ahead in 2026 got there by treating the stack as a single decision rather than a running series of them. One converged foundation. One home for the customer record. Launching a bundle becomes a configuration. Partner number six becomes an entry in a table. That structure is what survives your next twelve deadlines, because it stops asking your team to choose between speed and cleanliness every time a window opens.

Discipline won't do that work. Discipline loses to deadlines, every quarter, in every operator, including yours.

Simplicity at year four is a condition, and conditions expire. The operators who make it permanent choose to, while the picture is still small enough to hold in one hand.