# We replaced AI subscriptions with four senior developers. It was the best AI decision we made.

Author: Norbert Biedrzycki
Published: 2026-04-23
Language: en
URL: https://norbertbiedrzycki.pl/en/we-replaced-ai-subscriptions-with-four-senior-developers/
Summary: AI was supposed to reduce costs, increase productivity, and allow smaller teams to do more. So we did what most tech companies are doing right now.

For a while, everything seemed obvious.

AI was supposed to reduce costs, increase productivity, and allow smaller teams to do more. So we did what most tech companies are doing right now. We bought more tools, added more subscriptions, and started "scaling AI."

At first, it worked. The team moved faster, more things were being produced, and experimentation was at an all-time high. It felt like we were finally getting closer to that promised efficiency.

But after some time, something stopped adding up.

Costs were growing faster than results. Token usage became one of the biggest line items in our budget. Subscriptions kept piling up almost unnoticed. And every team was using AI differently. It became harder and harder to answer a simple question: what is the actual business impact of all this?

At some point, we had to pause and ask a much more uncomfortable question: are we actually buying productivity, or just activity?

Because AI is incredibly good at generating motion. Ideas, drafts, iterations, variations. But motion is not the same as progress.

So we did something that would have sounded absurd a year ago.

We started shutting down subscriptions. One by one, without much sentiment. In the end, we cut several AI tools, and with the same budget, we hired four senior developers.

Not because "AI doesn't work." That's too simplistic. The real issue was elsewhere.

AI is not a tool you just plug in and expect results from. It's an operational layer that amplifies whatever is already there. If your processes are messy, AI will make them faster and more expensive. If your workflows are unclear, you'll just get more output that's harder to evaluate. That's exactly what happened to us.

More prompts. More experiments. More output. But without structure and control, it all started to look like a very expensive system for generating work but not value.

Four senior developers changed more than all those tools combined. Not because they work faster, but because they remove what shouldn't exist in the first place. They simplify processes, make decisions, and build things that actually make sense. And only then does AI start delivering real value.

That was the turning point.

AI stopped being "the solution." It became a tool, used intentionally, where it actually makes sense. Looking at it now, the conclusion is much simpler, even if less comfortable.

The problem isn't the models or their cost. The problem is trying to replace thinking with them. And that doesn't scale.

The most expensive part of AI isn't the tokens. It's the lack of strategy.
