A CFO I spoke with recently has been vibe coding (aka natural language programming), on his own, a solution that handles a lot of what his company pays category leaders like Workday and SAP Concur for today. He is not a developer. He has no engineering background. He is planning to work with his IT team to roll it out module by module, and he cannot wait to start turning off some of the SaaS subscriptions his finance team has been paying for.
That conversation has been sitting with me for a few weeks. It is the clearest signal I have seen that the disruption story around SaaS is no longer theoretical.
Not every SaaS product is at risk. Some categories have real moats: deep integrations, regulatory footprint, network effects, decades of hardened workflow logic. But the idea that a finance executive can build something functional enough to replace parts of a multi-million dollar stack, on his own, in evenings and weekends, would have sounded absurd eighteen months ago. Today it is happening, and the CFO I spoke with is not the only one.
The pedestal we built around code
Somewhere along the way we turned coding into a destination. It became an identity, a career track, a credential, a gatekeeping mechanism. That was never what it actually was.
Coding was always just a way to get to an outcome. It was how you built a product, fixed a broken process, lowered a cost, improved an experience, or unlocked something humans could not do before. Self-driving cars. Fraud detection at scale. Recommendations that learn your taste better than your friends do. The goal was never the code. The goal was the thing the code made possible.
What is happening with the current generation of AI coding tools is that the pedestal is being lowered. Not because coding is less valuable, but because the ability to translate an idea into working software is becoming more accessible to anyone with a clear outcome in mind.
The tools doing the lifting
Claude Code & Cowork, Google's Antigravity, ChatGPT's Codex, and others are democratizing capabilities that used to take years of training to tackle. They are not perfect. They produce slop. They hallucinate. Sometimes the code does not work. Anyone who has used them seriously knows this.
But the curve is steep. What these tools could not do six months ago, they do routinely today. What they cannot do today, many will do by the end of the year. Betting against that trajectory has been a losing trade for three years running. The real disruption isn't just building cheaper; it's AI's potential to maintain and debug that code just as cheaply.
More importantly, the people using them do not need to be engineers. They need to understand their business, articulate what they want, iterate with the model, and have enough judgment to spot when something is off. Those are skills a lot of operators already have. The CFO I mentioned is one example. I expect to see many more in the next twelve months.
Why this matters for SaaS
For a long time, a certain class of SaaS vendor had an effective monopoly on a certain class of workflow. The workflow was valuable enough that building it from scratch was not worth the effort. The vendor charged accordingly. Customers were locked in and paid.
That calculation is shifting. When a capable operator can assemble a working prototype in a weekend, the build-versus-buy conversation changes. It does not mean the SaaS vendor loses. It means the vendor has to be offering something the customer cannot quickly recreate. Depth of functionality, reliability, ecosystem, compliance posture, customer success, the network of integrations built over a decade. Those are still real moats. Comfort is not.
The SaaS companies that will keep winning are the ones that treat this moment as a forcing function. They will invest in the things that are genuinely hard to replicate, and make sure their customers feel that value every quarter. The ones coasting on inertia are in a more exposed position than they realize.
The enterprise reality check
A weekend prototype is not an enterprise system. Production software at scale needs things that do not show up in a demo: access controls, audit trails, data lineage, disaster recovery, security review, integration testing, change management, and support processes.
This is where the real work begins. The interesting organizations I am watching are not the ones saying "anyone can build anything now." They are building lightweight enablement paths where business users can prototype, IT can vet and harden, and the best ideas graduate into properly supported internal systems. That is a very different operating model from the one most enterprises run today, and building it is going to take intention.
The sprawl risk is also real. Without the right guardrails, you end up with hundreds of half-finished internal tools, each with its own security posture, each depending on one person who remembers how it works. That is not a step forward, it is a different kind of mess.
What I would be doing right now
A few practical moves, depending on where you sit.
If you are an operator, pick the one or two workflows that are most painful and least well served by your current tools. Spend ten focused hours in one of the AI coding environments. Not to replace anything in production, but to calibrate. The gap between what you think these tools can do and what they can actually do is probably larger than you expect.
If you are a SaaS vendor, audit your product honestly. Which parts of what you sell could a motivated customer rebuild in a weekend? What are you doing to stay worth the price when that becomes obvious to more of your buyers? The answer needs to be better than "we have always been here."
If you are in IT, start designing the rails now. Business users are going to build things. The question is whether you shape that energy into something safe and scalable, or whether you discover it in a breach report eighteen months from now.
The wave is already here
The people who see this clearly are in for an uncomfortable but exciting few years. Software is becoming something more people can create, and that changes almost every assumption the last decade was built on. The winners will be the operators who embrace it, the vendors who raise their game, and the IT teams that get ahead of it.
Coding was never the point. Creating was. And creating is about to get a lot more democratic.