Should we pause AI? No — but we should stop letting it drive.
Some of the world's leading AI companies now say society may need the option to slow down or pause the development of the most advanced AI systems. It is a serious discussion.
But if you build real products for real customers, that is not your question. Your question is simpler: how close should you let AI get?
"At arm's length" is a principle borrowed from politics, culture and governance — a way to allow influence without control, and support without dependency. Applied to AI, it means one thing: use its power, but never let it dictate.
AI is not the problem. The problem starts when AI becomes the part your product cannot run without.
Used as a tool, AI is powerful. Used as the driver, it is a risk.
The first risk: dependency
Many companies now build products where AI sits inside the core of the system. The product depends on the model to work.
So ask yourself: what happens if the model has an outage? If the provider changes the terms? If access is cut off? If the model simply gives a bad answer?
If your product stops when AI stops, that is not control.
That is dependency.
The second risk: lock-in
There is a second risk most companies do not price in early enough: cost and lock-in.
AI agents do not make one simple call. They reason, call tools, check themselves and run again. One task becomes many model calls. So even when token prices fall, your total AI bill keeps rising.
And if your core product runs on AI — what happens when the cost doubles? Can you switch providers? Can you self-host? Can you turn it off? Or is there nothing underneath to fall back on?
The real negotiating power is the ability to walk away.
Want to know how dependent you really are?
Run the test. Turn off every AI API in your product and watch what happens.
Most products stop. There is nothing underneath to keep them running.
Flowen keeps running. Here is why.
How we build Flowen
At Flowen, we use AI every day — to code, to analyse, to move faster. But we keep it at arm's length from the core.
Humans review and approve every line before it reaches production. AI helps us build Flowen. AI does not run Flowen.
In production, AI is only a thin layer on top — limited support features like the chatbot, never the system itself.
So when we switch off every AI connection, Flowen runs business as usual. Quotes, orders, projects, purchasing, delivery notes, WMS, invoicing — all of it keeps working.
That is the difference.
Control is architecture
Control is not something you add later. You build it in from the start.
For us, that means:
We are not against AI. We use it openly and often. We are simply not willing to lose the ability to say no to it.
That is what arm's length means: cooperation, not surrender.
