Every business is now a software business. That is not a prediction. It happened, and most people have not caught up to it yet.
The thing that changed is who gets to build. For my entire career, there was a wall between the people who understood a problem and the people who could fix it in software. The person closest to the broken workflow, the one who lived inside it every day and knew exactly what was wrong, was almost never the person who could change it. They had to file a ticket. They had to explain the problem to someone who had never done their job. They had to wait for a roadmap, a sprint, a quarter, a budget cycle. The fix was always six months away and always slightly wrong when it finally showed up, because something got lost in the translation between the person who had the problem and the person who wrote the code.
That wall is gone. The game has literally changed.
What does it mean that every business is now a software business?
It means the ability to build and fix software is no longer limited to engineers. AI coding agents now let the person closest to a broken workflow fix it themselves, without filing a ticket or waiting for a development team. The internal tools, spreadsheets, dashboards, and manual processes that every business runs on can now be improved by the people who actually use them. The dynamic is a direct inverse of the desktop-era trap I described in Bad Software Became the Default: for two decades the working reality of the user and the assumptions of the tool diverged because the user could not change the tool; now the user can.
The person closest to the workflow can now fix it themselves. Not someday. Not through a six-month roadmap. Now. The spreadsheet that everyone fights with. The dashboard that never shows the right number. The internal tool that was built for a version of the business that stopped existing two years ago. The manual process that every single person on the team quietly hates and works around. All of it can now be fixed by the person who actually understands what it is supposed to do, in the time it used to take to write the ticket asking someone else to fix it.
What is OpenAI Codex?
OpenAI Codex is an AI coding agent that performs software tasks such as writing code, fixing bugs, and building tools. It started as a tool for software engineers and has expanded into a broader productivity tool used by knowledge workers as well as developers. OpenAI has described it as becoming a productivity tool for everyone, not only for engineers.
I want to be precise about what I am pointing at, because this is not a vague AI-is-changing-everything claim. I am pointing at a specific shift in a specific category of tool. Coding agents have crossed a line where the person using them does not need to be a coder. OpenAI's Codex is the clearest example. It started as a tool for software engineers, and it has become something much broader. In a piece on its own site, OpenAI said Codex now has more than 5 million weekly active users, up more than six times since it launched its desktop app in February. The part that matters most is who those users are. Developers are still the largest group, but knowledge workers now make up about 20 percent of users, and that group is growing. OpenAI titled the piece about this shift Codex is becoming a productivity tool for everyone. Read that title again. Not a productivity tool for engineers. For everyone.
That is the whole thing in one line. The tool that was built to help coders write code is now being picked up by people who have never written a line of code in their life, to build the things they need to do their actual jobs.
Can non-coders build software with AI coding agents?
Yes. Tools like OpenAI's Codex have expanded beyond software engineers to general knowledge workers. According to OpenAI, Codex passed more than 5 million weekly active users, and while developers remain the largest group, knowledge workers now make up about 20 percent of users and are growing more than three times as fast. The user describes the problem in plain language and the tool handles the translation into working software.
Here is what that means in plain terms. Not everyone is a coder. That has not changed and it is not going to. But everyone can now build. Those are two completely different statements, and the gap between them is where the entire opportunity lives. You do not need to understand how the software works under the hood. You need to understand the problem you are trying to solve, and you need to be able to describe it clearly. The describing is the new skill. The person who can look at a broken process and explain exactly what it should do instead is now the person who can fix it, because the tool handles the translation from plain description into working software.
How does AI change the bottleneck between knowing a problem and fixing it?
Traditionally, the person who understood a broken workflow was rarely the person who could fix it, so ideas had to travel through tickets, backlogs, and developer queues, and most died along the way. AI coding agents collapse that distance by letting the person with the problem build the fix directly. The result is that small, everyday fixes that were never worth a developer's time become trivial to make, and that change is the same in-shape one I described in Demos Are Easy. Deployment Is the AI Company Moat.: the value is built on the ground floor of real workflows, not in the demo of what could be built.
Think about what that does to the old bottleneck. The old model was that good ideas for fixing a workflow had to survive a long, lossy journey. From the head of the person who had the idea, into a ticket, into a backlog, into a prioritization meeting, into a developer's queue, into code, and finally back out into the world as a tool that was usually a little off from what was actually needed. Most good ideas died somewhere on that journey. They got deprioritized. They got misunderstood. They got stuck behind something more urgent. The friction was so high that people stopped even suggesting fixes, because they knew the fix would never come. They worked around the broken thing forever.
Now the journey is short. The person with the idea and the person who builds the fix are the same person. There is no translation loss because there is no translation. There is no queue because you are the queue. The fix that used to be six months away is now an afternoon, sometimes an hour. And because the cost of fixing something dropped so far, the things that were never worth fixing before, the small daily annoyances that quietly tax everyone, suddenly become worth fixing. That is the part people are missing. It is not only that big projects get faster. It is that a whole universe of small fixes that were never economical to build are now trivial, and those small fixes are where most of the friction in a real business actually lives.
I keep coming back to the same realization. The constraint on improving a business was never a shortage of good ideas about what to fix. The people doing the work always knew what was broken. The constraint was the distance between knowing and building. That distance is collapsing, and it is collapsing fastest for exactly the people who were furthest from the code before, the people on the front line who understand the work but were never going to learn to program.
Let me make this real with the world I know best. Think about a real estate team that runs its whole business out of a spreadsheet held together with hope. The transaction tracker that breaks every time someone sorts a column wrong. The commission calculator that one person built years ago and nobody else dares to touch. The follow-up list that lives in three different places and never agrees with itself. Every team has this. Everybody knows exactly what is wrong with it. And for twenty years the answer was the same, you either pay a developer you cannot really afford, or you live with it. So everyone lived with it. The person who actually ran the transactions, who knew precisely where the tracker fell apart and why, was the one person guaranteed not to be able to fix it. That was the trap. The knowledge and the ability to act on it lived in two different people, and getting them together cost more time and money than the problem seemed to be worth.
That is the trap that is breaking right now. The person who runs the transactions and knows where the tracker fails is now the person who can rebuild it, by describing what it should do instead. No developer. No budget request. No waiting. The expertise that used to be stranded on the wrong side of the wall is now the most valuable thing in the room, because the person with the expertise can finally act on it directly.
Will AI coding agents replace software engineers?
No. Deep, complex engineering still requires years of expertise and is not going away. What has changed is the floor. The baseline ability to turn a well-understood problem into a working tool is now available to people who never had it before, which expands who can build rather than removing the need for skilled engineers on hard problems.
I am not saying this makes everyone a software engineer. It does not. There is still deep, hard, real engineering that requires years of expertise, and that is not going away. What I am saying is narrower and more useful. The floor moved. The baseline ability to take a problem you understand and turn it into a working tool is now available to people who never had it before. The wall between understanding a problem and fixing it has come down for an enormous category of everyday business problems, and that category is the one that touches the most people.
So here is the question every business should be asking right now. What are the broken things everyone has been working around for years because fixing them was always someone else's job and never a priority? Make that list. The spreadsheet held together with tape. The report someone rebuilds by hand every week. The handoff between two teams that loses something every time. The dashboard nobody trusts. All the things that got filed under that is simply how it works here. Most of those are now fixable by the people who live with them, this week, without waiting for anyone's permission or roadmap.
The businesses that win the next few years will not be the ones with the biggest engineering teams. They will be the ones where the people closest to the work feel free to fix what is broken the moment they see it, because the tool to do it is finally in their hands. The wall is down. The only question left is who notices first and starts building.
It is a new future. The person who understands the problem is now the person who can solve it. That has never been true before. It is true today.
Judd Hoffman is CEO and Co-Founder of Ethica AI, building AI-powered tools for real estate transaction workflows.