The technology gap: between revolutionary AI and broken HTML forms
January 10, 2026
The technology gap: between revolutionary AI and broken HTML forms
We live in insane times.
On one hand, we have new AI models from Google, Anthropic, OpenAI and others that enable incredible things:
- Prototyping applications in days instead of months
- Solving math problems whose first line of statement the average person can't understand
- Testing new drugs and discovering new materials
- Speaking with anyone on the planet, in any language
And on the other hand?
- My bank that can't make an HTML form work
- My ISP that can't process a direct debit
- The tax authority website that seems written in Braille / Mandarin
It's hard to find a more frustrating technology gap.
Why such a disconnect?
It's not a technology problem. The tools exist, the skills exist too. The problem is structural:
The technical debt of large organizations is colossal. Systems built 15, 20, sometimes 30 years ago, stacked on top of each other, maintained by teams that have neither the budget nor the mandate to rebuild everything.
Decision cycles in large companies and administrations are incompatible with the pace of innovation. By the time a project is validated, budgeted, launched and delivered, the technology has already changed three times.
The lack of competitive pressure for institutions. Your bank can have the worst website in the world — you're not going to switch banks for that (or hardly). The tax authority has no competitor. When there's no market pressure, there's no urgency to improve.
Customer support: a massive market
And this is where it gets interesting. The gap between what technology enables and what organizations actually deliver creates an enormous opportunity.
Customer support — in the broad sense: support, maintenance, modernization, redesign — is going to be a massive market in the coming years. Because AI isn't just going to allow creating new things faster. It's also going to allow fixing what already exists.
AI assistants are getting very good at working on legacy code. So there's new hope of seeing these large organizations catch up — provided they want to.
In the meantime
In the meantime, we'll keep living in this absurd world where you can ask an AI to generate a complete application in an afternoon, but it takes three weeks to get a certificate from your health insurance.
It's the Haussmann building of tech: it's old, poorly insulated, but it's become heritage.
Originally published on LinkedIn.
