Citizen Development – Everybody is building software

“Building software is easy, when Claude does it!” “Of course people have always been able to develop some form of tooling without coding, just think of spreadsheets!”

There is yet another term making rounds in the software space, and that term is Citizen Development. In practice, citizen developers are people outside the IT organisation that are building software tools to do their job better, more efficiently, or just because they now can: apps, automation, reports, dashboards and what have you.

With AI Agents like Claude Code, Codex and the like, everyone can build applications within minutes. The latest frontier models are very good at one-shotting (creating something from a single prompt) simple and well-scoped applications. Chances are you have tried this yourself. You don’t even have to leave the chat to create small frontend React applications in interfaces that ChatGPT and Claude provide in the browser.

New builders emerge

What we see in the field is that people in various roles are building software:

  • Executives and managers answering their own questions via AI agents generating the queries ad hoc. At Anthropic, 95% of business analytics queries are handled by Claude, with about 95% accuracy.
  • Designers contribute working code, not just layouts to hand over to the developers. The gap between design file and production is shrinking fast.
  • Quality Assurance people build custom testing tools in hours that keep agent output in check. Tests that constrain what an agent may generate, scenario tests, and examples of good and bad behavior drawn from real production incidents. Practitioners at Future of Software Engineering Europe (June 2026) reported that such purpose-built tools beat generic test frameworks, partly because they keep the human review simple and hard for an agent to game.
  • Salespeople change a real app’s front end in a couple of minutes in front of a customer, or build a prototype of a new product direction for a showcase.
  • Product owners and managers show their vision with a working example instead of a long spec.
  • CTOs and CIOs test an idea, an architecture assumption or a vendor’s claim themselves before committing a team to it.

The Anthropic example is of course unfair, their organisation is optimised for this due to their domain. But it shows what is possible. Bear in mind they did not succeed overnight: There was development before they reached the levels of accuracy and adoption they did.

From Prompt to Production

Unfortunately, moving these apps from “works in my browser” to production is often a huge undertaking. Once you have to deal with realities of production level security, audit logs, monitoring, users, permissions, updating libraries, building container images and dozens of other things your run-of-the-mill citizen developer is not equipped to deal with, things really start to get hairy.

But what if there was a way to reduce that friction while keeping an organisation’s software landscape under control?

Annoyingly, we must call all previously discussed shadow IT. Anyone building software outside IT is probably doing it without following the IT’s blessed path for doing so. Shadow IT is hardly a new phenomenon, but it has become a MUCH bigger problem with the advent of coding agents.

The problem with shadow IT is that the regular IT exists for a reason: it’s the way the organisation has agreed to handle software development, operations and maintenance work. Although following the organisation’s IT operation standards might seem to add unnecessary friction to citizen developer’s relaxed and blissfully ignorant life on the software fast-lane, the standards, rules and policies have hopefully been established previously for a reason.

Offence is your best defence

The troublesome thing with these coding agents is that you cannot stop people from using them, code is text, all AI agents (even in chat) can thus produce code. This form of shadow IT is here to stay, our only option is to address it. But actually, you probably want to embrace it. When we stop and think about it, this can be tremendously useful. Building software democratised for everyone, why go back to excel macros if we can write code to solve the problem, without writing code? Duh.

Instead of blocking people from using their superpowers (coding agents), what if we would focus on the unthinkable: ALLOWING IT.

What if I told you: you can vibe code and get away with it. Enable Citizen Development initiatives. There is of course a BUT coming up, sorry about that.

The big but is that you must invest in the internal IT platform capabilities to match this demand, and never before have non-developers really used these platforms! The citizen developers commanding the coding agents are a new breed of users, and the platforms must adapt to this. Sure, you can outsource some of this to off-the-shelf platforms out there, but you still need integration work with the existing IT to get this rolling.

We have discussed the importance of platforms in the age of AI before too. When development becomes cheap(er), the bottleneck shifts.

Need Help?

There is no stopping the citizen development, but you don’t have to figure out what to do about it alone! Get in touch if this is a topic your organisation could use help with and we are more than happy to help.