AI coding will flood the enterprise market. That is not an obituary for software. It is an overdue filter for dashboards that pretend to be decision solutions.

By Bradley Keefer, Chief Revenue Officer, Keen Decision Systems

The “SaaS is dead” headline is directionally right about one thing: AI coding is going to flood the enterprise market.

We are going to see two categories explode at the same time. First, a massive wave of trash tools. Thin wrappers, duplicated features, brittle integrations, and a lot of vaporware that looks impressive in a demo and collapses in production. Second, genuinely good tools that eliminate the tasks everyone hates anyway. The repetitive work. The glue code. The reporting busywork. The internal workflows that have never been strategic, just necessary. That is a good thing. That is progress.

Where I think we need to be more careful is the “vibe code your way into sophisticated decision agentic systems” narrative.  It makes me think about the old, “set it and forget it” commercials, but for media buying.

Marketing mix modeling, media planning, and decision-grade buying are not hard because the interface is hard. They are hard because the problem is complex across three dimensions at once: the math, the process, and the humans. The model is only one part. The assumptions, the priors, the data reality, the governance, and the way decisions actually get made in an organization are the rest of it.

Will machines make humans faster? Absolutely. Will machines simplify parts of the process? Yes. Will machines help generate code and automate the mechanical steps? No question.

But will machines reliably deliver the math, reasoning, and logic you need for business-level decision making, end to end, with stability and accountability? Not yet.

I spent a full day recently vibe coding my way to an open source MMM setup, plus an experiment and lift tool. The headline is that it is very doable. The less comfortable truth is that the models are not fully stable, and “experiments” as a standalone capability are not where the full value lives.  I don’t think anyone will be shocked to read this.

In fact, here is a take that will probably annoy some vendors: if you are paying a premium for an experiment’s workflow, you should at least ask yourself whether you are buying a commodity. You can build an experiments agent yourself for free. And if you look closely at a lot of vendor roadmaps, you can see the shift already. Experiments are becoming table stakes. They are getting de-emphasized because, in a world of AI-accelerated engineering, they are increasingly replicable.

After building the open source MMM, I would make a similar argument about modeling.  The problem is that code is unstable, the priors are missing and the decision ready output isn’t there.  If you vibe code your way to MMM, good luck defending your decisions in the long term.

The model is not the moat.

The real value is what feeds and governs the model. A priors database that is credible, curated, and continuously improved. A decision layer that is agentic enough to help guide what to do next. A planning layer that translates evidence into options, tradeoffs, and implications. Notice the language: guide, not make. The reasoning is not there yet for a machine to responsibly make these decisions on behalf of leaders. But it can absolutely help leaders move faster and with more structure.

This is the strategic point I think gets lost in the headlines.

We should chill with the “everything is over” narrative and take a breath. A lot of the loudest takes are coming from two places: poorly run businesses watching their economics unwind, and dashboard-first software companies that confused reporting with decision support. And when those stocks tank, the story becomes existential instead of strategic.

The real question is simpler and more practical.

In a world where software is cheap to build and easy to copy, what is the value of a vendor that does not help you make faster, smarter decisions?

If the product is just an interface, it will be replicated. If the value is just access to a workflow, it will be commoditized. If the promise is sophisticated decision making, it has to be backed by real math, real stability, and a system that respects how organizations actually operate.

That is where the next era of enterprise software is going. Not the end of SaaS. The end of SaaS that never earned the right to exist in the new world.