The Rise of Asymmetric Software: The Beginning of the End for Traditional SaaS

We've spent the last 30 years building software around categories. CRMs to manage customers, ERPs to manage operations, project management tools to manage people who are already tired of being managed.

We've built an entire industry around helping humans operate software. Now we're building machines that can increasingly operate it themselves.

That's a slightly uncomfortable situation for an industry that makes billions selling people tools.

I think we're approaching the beginning of the collapse of traditional software as we know it. Not because software will disappear, but because the categories we've spent decades building around may no longer make much sense.

And what comes next could be far more interesting than another AI-powered CRM.

Wait, what's collapsing?

Think about how companies operate today. Customer information lives in a CRM, finances in accounting software, projects in another application, and internal conversations somewhere else. A single business decision might require someone to jump between five different tools.

We've accepted this as normal because software has always been built around specific functions. Building specialized applications was practical when engineering was expensive and intelligence had to be programmed through predefined rules.

But AI changes a fundamental assumption.

What if the software understands what you're trying to accomplish and coordinates the work across those systems? You may still need databases, accounting records, and reliable infrastructure, but you might not need five separate interfaces or fifty employees updating them manually.

This is where things get interesting for SaaS companies. Many charge based on how many people use their applications. If fewer people need to operate the software, that business model comes under pressure.

The first things to disappear probably won't be Salesforce or SAP. It'll be the smaller tools whose entire purpose is moving information between other tools, generating reports, or making people click through workflows that machines can increasingly handle.

The collapse begins when software stops being something we constantly operate and becomes something that increasingly operates on our behalf.

So, what is asymmetric software?

I use asymmetric software to describe products that deliver capabilities far beyond what their size, cost, or development resources would traditionally suggest.

But there's another part to it. These products don't necessarily fit into any existing software category.

Imagine someone running a manufacturing business. They need to manage suppliers, forecast demand, track inventory, evaluate pricing, and coordinate deliveries. Today, that probably means an ERP and several additional systems.

Now imagine a small team building an intelligent system that understands the entire operation and coordinates these activities through existing data, APIs, and a few carefully designed interfaces.

Is it an ERP? A procurement platform? A financial planning tool? An operations assistant?

Maybe all of them. Maybe none.

That's the interesting part. Instead of starting with a software category and deciding which features belong inside it, you start with a problem and assemble whatever capabilities are necessary to solve it.

And because intelligence is becoming a reusable building block, the resources required to experiment with these combinations are falling.

A product that previously needed a large engineering team might become possible with a few skilled people, existing infrastructure, and AI. Production reliability still takes serious engineering, of course. Nobody wants an enthusiastic AI agent accidentally ordering 40 tonnes of steel.

But the economics of what can be attempted are changing.

The weird people might have an advantage

This is probably the part I'm most curious about.

There have always been people who don't naturally think within established boundaries. Polymaths, unconventional thinkers, people who connect ideas from completely unrelated fields.

Historically, turning those ideas into software required engineers, capital, and a considerable amount of specialized knowledge. So even unconventional ideas often ended up being simplified into recognizable products.

You had an unusual idea? Great. Now explain which existing software market you're entering.

AI is gradually removing part of that constraint.

Imagine a farmer who understands weather patterns, commodity pricing, logistics, and soil conditions better than most software engineers ever will. Or someone in healthcare who understands the strange relationship between patient behavior, administration, and clinical workflows.

These people might imagine products that wouldn't occur to someone whose starting point is building the next CRM.

And now, for the first time, more of them can experiment with those ideas without first assembling an entire technology company.

I suspect some of the most interesting software products of the next decade will come from people who never intended to become software founders.

Not because engineering stops mattering, but because understanding a problem deeply becomes more valuable when building an initial solution becomes easier.

Are we imagining the future wrong?

When smartphones arrived, we initially thought of them as better phones. Eventually, they created entirely new businesses and behaviors that had very little to do with making phone calls.

I think we're making a similar mistake with AI.

We keep imagining AI versions of existing software. AI CRM, AI ERP, AI project management, AI productivity tools. Basically, the same applications with a chatbot attached.

Some of these will be useful. But we're still imagining the future using today's categories.

What if the next useful product combines personal finance, scheduling, decision-making, and behavioral understanding into something that isn't really any of those applications?

What if a business system understands an objective and dynamically assembles the tools required to achieve it, instead of asking employees to navigate through predefined workflows?

Once intelligence becomes a fundamental building block, there's no particular reason every new product must resemble something we've already built.

This is also why the next major software category might be difficult to predict. We're looking for it using the vocabulary of the previous generation.

But software isn't becoming free

There's a difference between generating an application and building something people can trust.

AI can help write code, but production software still needs security, infrastructure, testing, integration, and accountability. Established platforms also have decades of business logic and operational trust behind them.

So I don't think traditional software companies will simply disappear. Many will become the infrastructure that intelligent systems depend on. Others will adapt their products around outcomes rather than interfaces.

But companies whose only advantage is a collection of familiar features may have a harder time defending that position.

This changes how we should think about building products, too. At Beitroot, I'm increasingly interested in a question that comes before architecture, technology, or even product design: Does this problem actually need the kind of software we're imagining?

Because as building becomes easier, deciding what deserves to be built becomes more important.

Maybe that's the interesting part

For decades, software development has been constrained by what we could afford to build. Engineering resources, capital, and technical complexity determined which ideas made it into the world.

AI is beginning to loosen those constraints.

There will probably be an enormous amount of useless software as a result. We already have enough applications reminding us to be productive while making us less productive.

But we'll also see people exploring ideas that would previously have been too strange, too expensive, or too difficult to explain.

And some of those ideas might become entirely new kinds of products.

I don't think the future of software is simply smaller teams building the same applications faster.

I think it's people building things we haven't learned how to categorize yet.

Maybe the next big software company won't be the one that builds a better CRM. It'll be the one that makes us wonder why we needed a CRM in the first place.