Low Code is Not Dead

“Low code is dead!”

Every time this phrase pops up in the context of Power Platform/Dynamics 365 I experience déjà vu. The last time Microsoft and the community at large pushed a narrative like this, it distorted expectations, confused customers, and misled new professionals entering the field.

Remember “citizen developers”?

Go back eight or nine years. Microsoft marketed the Power Platform as the dawn of the “citizen developer.” The promise was simple: non-technical people would build high-quality applications on their own.

It didn't happen. In the best cases, skilled professionals were able to deliver better solutions, faster. In the worst cases, hobbyists confidently built hot garbage that seemed great until professionals and/or IT teams were forced to deal with it. The most successful versions of this had the citizen developer as a tiny icon in a vast, complex governance diagram.

The parallels between that era and today’s “low code is dead” narrative are hard to ignore. Back then, low code was supposed to turn citizen developers into pro developers. Today, AI is supposed to turn low coders into pro coders (or something?). Same suit, different cut and fabric.

Microsoft is telling the same story

The marketing is all too familiar: our platform will do stuff for you that you don't need to really understand. I didn’t buy it a decade ago, and I don’t buy it now. It’s not just bad messaging — it misleads everyone, especially people entering the industry.

“But AI is different!”

Exactly right, AI is indeed different. It is vastly more powerful than anything we had ten years ago. It doesn’t change the core flaw in the narrative. Whether it’s the “citizen developer” story or the “low code is dead” story, the underlying suggestion is the same:

People don't need to understand complex systems in order to reliably build complex systems.

Which is not remotely true and never has been.

“But I’ve done this! I’ve built these applications!” You unquestionably have, and you fall into one of these categories:

1. You’re a developer, regardless of whether someone stuck “citizen” in front of the word.
2. You’re a genuine citizen developer supported and guided by a team of professionals who keep your solution stable.
3. Your application was below a complexity threshold such that that luck and intuition got you there.

Personally, I've been all 3. Software sometimes works even when the thinking behind it is incomplete — but not reliably, and not at scale. Complexity punishes shortcuts.

A real project example

Years ago, one of the most successful projects I ever led produced a Minimum Viable Product “in four weeks” (back to this in a moment) to a relatively large organization — a completely no-code model-driven Power Platform app. We didn’t write a single line of code until months later, and even then it was mostly scripting.

Why did it work?

Because before those four weeks began as we waited for the contracting to complete, the team spent time:

• exploring hypothetical organizational scenarios in the absence of requirements
• prototyping data models
• mapping how the scenarios sat most comfortably within Power Platform capabilities
• stress testing our thinking while tweaking the models to satisfy the hypothetical scenarios

We hit the ground very well prepared to adapt to the actual requirements in a way that withstood future policy changes and rapidly changing requirements over time.

We succeeded not because no code magically made the hard parts easy, but because we already knew the technology and put in the work. "No code" enabled us to (extremely unsurprisingly) not write code — i.e. get the build portion done faster. That's an important portion of the work of course, but it's only a portion.

AI doesn’t do that

AI can give you a starting point and help you design, build and operate even faster and better. It cannot completely and autonomously architect a resilient, future-proof data model. Citizen developers, by definition, don’t really care about things like data models.

And a data model is just one thing among dozens and dozens of things that implementation teams must manage in order to be successful. Things like multi-layered security, requirements gathering, solution versioning, company culture, integration scenarios, defect prioritization guidelines, and I could go on and on and on about the elements of a successful project. Most of the skills and knowledge important to such projects aren’t even technical, including some near the top of the list.

None of this is a knock on citizen developers, nor low coders, nor AI users. I love those people and I want more of them. They are curious, add real value and drive business outcomes — a colleague of mine is jumping in and using Forms and Lists to improve a workflow. Awesome! Use AI to make it even better! That's not the same as architecting enterprise solutions and never was.

Low code is not dead

AI doesn’t kill low code, it supercharges it. You still need all of the same things to properly leverage it: understand your technology, have adjacent skills and overlay experience. You still need real expertise. You still need to think deeply, design thoughtfully, understand the environment in which you are solutioning and as always, learn from the past. AI just takes all that and makes skilled people faster.

"Low code is dead" is a phrase built to cleverly rejuvenate a past false promise. It is designed to get you to think that AI does your thinking for you. Ignore it. It's AI, not AGI.

Low code isn’t dead, it’s only entering adulthood.

*Written with editing input from ChatGPT*