When AI Makes Building Things Too Easy

I recently built the most complicated Power Automate workflow I have ever attempted. By the time it was finished, it had roughly thirty-seven steps, with conditions branching in several directions and a collection of Apply to each loops nested inside one another. A couple of years ago, I would have looked at a design like that and decided it was beyond my skills.

I am not an elite Power Automate developer. I know enough to build useful things and to recognize what ought to be possible, but there has always been a wide gap between knowing something should be possible and knowing exactly how to make the platform do it. AI closed most of that gap. When I didn't know how to accomplish a step, I described what I was trying to do. When something broke, I worked through the problem with AI, adjusted, and tried again. Eventually the whole workflow ran from beginning to end and did exactly what it was supposed to do.

It was a satisfying win. But after watching it run a few times, a bigger thought took over.

The Real Shift Is Not About Developers

If AI could help me build something that complicated, what is about to happen everywhere else?

Professional developers have always been able to build workflows like this one. The change is that ordinary employees, the people who understand their own jobs and processes better than anyone, are gaining the same ability. As AI gets better at turning a described idea into working technology, an enormous amount of technical capability is moving closer to the people who understand the business problems in the first place.

That may be one of the most valuable opportunities AI offers. It also deserves more scrutiny than it usually gets.

AI dramatically reduces the cost and difficulty of turning an idea into a working solution, and it does that whether the idea is a good one or not. It makes implementation easier. It does not make our judgment any better.

We Have Seen This Before

Microsoft 365 already gave organizations a preview of this problem. It hands ordinary businesspeople an impressive set of building materials: Teams, SharePoint sites, Lists, libraries, workflows, apps, pages and dashboards. When those tools are applied with some thought about purpose, structure and ownership, they can make work considerably better.

When they are applied without that thought, organizations end up with a different result. Give everyone the ability to create things without much consideration of what belongs where, who owns it, who should have access, how long it should exist or what problem it is supposed to solve, and you get a modern collection of tools sitting on top of the same confusion you were trying to escape. The chaos is not eliminated. It is simply upgraded.

AI is going to make that outcome easier and faster to reach.

Losing the Accidental Speed Bump

None of this is an argument against employees building things or experimenting with AI. The fact that capabilities once reserved for specialists are reaching the people who do the work is exactly what makes AI exciting.

The concern is what happens when our ability to create solutions improves faster than our discipline for deciding which problems deserve solutions.

There used to be natural resistance between having an idea and turning it into technology. Significant solutions usually required someone with specialized skills. That person cost money, the project took time, and at some point someone had to explain why the idea was worth the trouble. That process never prevented bad decisions, but it created an accidental speed bump.

AI is flattening that speed bump. Mostly that is good news, because a useful idea no longer has to die simply because a small company lacks a developer. The catch is that lowering the cost of experimentation also lowers the cost of acting on ideas no one examined carefully.

For years, the central question in technology projects was, "Can we build this?" Increasingly, the answer will be yes. When that happens, the more valuable question becomes, "Should we build this?" And even that question may come a step too late.

Friction Before Features

Before deciding whether to build something, it helps to understand what made someone want it in the first place. That is the idea behind a principle we return to often: Friction Before Features.

Before discussing a feature, an automation, an AI agent, an app or a new Teams structure, describe what is making the work harder than it should be. Perhaps project managers spend hours each week chasing updates because no one can see a project's status in one place. Perhaps employees type the same customer information into three different systems. Perhaps no one is sure which version of a procedure is current.

Then ask whether that friction matters enough to address. Some annoyances cost real time, add risk, slow decisions or frustrate customers. Others are just annoyances. If removing a minor irritation means creating an elaborate system that someone must own, maintain and explain for the next five years, the cure may be worse than the problem.

Next, understand what is actually causing the friction. Symptoms are remarkably good at disguising themselves as problems. If people cannot find the right procedure, the answer might be better search or an AI agent. Or there might be six copies of the procedure scattered across the organization with no one responsible for the authoritative version. Those are very different problems. AI might help with the first. The second needs someone to take ownership, decide where the real version lives and clean up the duplicates. Skip that diagnosis, and you risk building a sophisticated way to navigate a mess that should have been cleaned up instead.

Only after understanding the friction, why it matters and what causes it does it make sense to talk about solutions. The answer may well be AI, Power Automate, SharePoint, Teams, a CRM system or a new app. It may just as easily be a clearer process, an agreement about ownership, better training, an approval that can be eliminated or a better doorway into information that already lives in the right place. Occasionally, the honest answer is that the friction is not worth the money or energy to fix right now. That is a legitimate answer too.

Capability Is Growing. Judgment Has to Keep Pace.

Picture someone handy who suddenly gains access to the best fabrication shop in the county, complete with a welder, a plasma cutter, a drill press and an expert standing by to explain every tool. That is an enormous increase in capability. It does not mean that every time the tractor acts up, the right response is to weld new parts onto it.

The first step is still to walk around the tractor and figure out what is wrong, why it matters and what is causing it. Sometimes a new bracket really is needed. Sometimes the fix is tightening a bolt. The shop loses none of its value because someone paused to diagnose the problem. In fact, the better the tools become, the more valuable that judgment is.

That is where AI is taking us. Ideas that once died somewhere between "wouldn't it be nice if" and "we don't have the budget for that" are about to become practical, and that is worth celebrating. But if we can build five things for what it once cost to build one, it does not follow that we needed five things.

The complicated workflow I built still works, and I am still a little proud of it. The most valuable thing I took from the experience, though, was a reminder that being able to build something and having a good reason to build it are two separate accomplishments. AI is making the first one easier every month. The second one is still up to us.

The goal was never more technology. The goal is less friction and better work for the people doing it.