AI Will Expose Weak SaaS. Not Replace It.

This article was originally published on LinkedIn and is reshared here for broader access.
‍
When $300 billion in market value was wiped from enterprise software stocks in a week following a wave of AI product releases, people reached for apocalyptic language.

“The SaaSpocalypse Has Begun.”

“The SaaSacre.”

“Is SaaS Dead?”

I’ve seen this movie before.

Every technological revolution looks chaotic in its first act. Capital floods in, and valuations swing. Commentators jump straight to the end state. And before you know it, headlines are predicting robots replacing everyone’s job, AI agents becoming the new front door to enterprise software, and entire industries collapsing overnight.

But AI is not a SaaS extinction event. It’s an architectural stress test, and not every system will pass.

This Isn’t the End. It’s the Build.

I look at moments like this through a historical lens. In Technological Revolutions and Financial Capital, Carlota Perez describes how major technological shifts move through two primary stages: an installation period marked by speculation and infrastructure build-out, followed by a longer deployment phase, when productivity gains actually materialize.

Installation is messy.

Twenty years ago, mobile networks were unreliable. Infrastructure had to be built before smartphones could reshape behavior. Uber didn’t emerge until coverage was ubiquitous.

My kids assume 100% mobile coverage was always there. But that took years of installation before the deployment phase reshaped daily life.

AI feels similar.

Everyone is debating what AI agents will eventually do. Far fewer are asking whether the underlying systems are built to support them.

Uber wasn’t possible without reliable mobile networks. AI agents won’t function without modern, interoperable infrastructure.

There’s a popular narrative that AI agents will become the new front door to enterprise software. That assumes the back end is ready. In many companies, it isn’t. Before agents become the interface, infrastructure has to become interoperable.

Architecture is the Real Divide

The question you should be asking is whether your architecture was built for a batch world or whether it can operate in a real-time, agent-driven one.

Much of enterprise software was designed for a different era: overnight processing, file transfers, manual reconciliation across systems, static dashboards. That made sense when real-time integration was difficult and infrastructure was more constrained.

The companies that win this cycle won’t be the ones shipping the most AI features. They’ll be the ones whose architecture allows AI to operate cleanly.

This is why I don’t see AI as a SaaS extinction event. I see it as a stress test.

If you’re a true system of record that holds structured, regulated, permissioned data, then AI doesn’t replace you. It runs through you.

But if your product is a layer of human workaround sitting on top of brittle integrations, AI exposes that immediately.

In fintech and benefits, we see where AI delivers real value today: repeatable, rules-based processes.

Take health benefits claims, for example. Historically, someone submits a reimbursement request. It gets reviewed. It gets flagged. It goes back and forth. It creates friction and frustration.

In our experience, once we introduced real-time validation, rejected claims dropped by 90%. Participants received immediate clarity instead of waiting days. Administrators reduced manual review cycles. That’s the difference between automation as a feature and automation as infrastructure.

These aren’t glamorous workflows. But when someone sets aside time to deal with their benefits, they want it resolved in one sitting. AI is particularly good at that kind of structured, rules-based work.

What’s interesting is what business customers are actually asking for at this moment.

Despite the noise, most enterprise buyers aren’t demanding full autonomy. They’re asking for optionality.

The companies selling “fully autonomous enterprises” are often talking to investors. The companies actually buying AI are asking for guardrails.

That doesn’t look like panic. It looks like an installation.

This Is a Steve Jobs Moment.

This is a founder-mode moment. Not because AI demands chaos, but because it demands re-architecture.

In stable cycles, operational optimization wins. In inflection points like this one, you need to think more like Steve Jobs: reimagining the product, challenging assumptions, and rebuilding core architecture before the market forces you to.

The leaders who navigate this shift well won’t be the loudest voices debating whether SaaS is dead. They’ll be the ones quietly rebuilding for an agent-driven world.

They’ll ask harder questions:

Are we architected for interoperability? Can AI agents securely operate across our systems? Are we event-driven instead of batch-based? Are we designing for delegation rather than just dashboards? Are we modernizing before the market forces us to?

Markets often overestimate short-term disruption and underestimate long-term transformation. We are still in the installation phase, rebuilding infrastructure and modernizing core systems. AI will not eliminate enterprise platforms, but it will change how we interact with them.

The companies that adapt early will be the ones building for what comes next.

‍