For the past two decades, the software industry has been dominated by a single idea: why build when you can subscribe? SaaS became the default answer to almost every business problem. Need a CRM? Salesforce. Project management? Monday.com. Marketing automation? HubSpot. The logic was sound — building custom software was expensive, slow, and risky, while SaaS platforms offered instant access to battle-tested tools at a fraction of the cost.

But something fundamental is shifting. AI is dramatically lowering the cost and complexity of building custom software, and it's forcing us to ask a question that would have sounded absurd five years ago: are we approaching the end of big SaaS as we know it?

The trade-off that created SaaS

To understand where we're headed, it helps to remember why SaaS won in the first place. Custom software development has historically been brutal. You needed specialised developers, months (or years) of build time, ongoing maintenance teams, and a budget that made most CFOs wince. SaaS solved this by amortising those costs across thousands of customers. You got 80% of what you needed for a fraction of the price. The remaining 20%? You adapted your processes to fit the software.

That trade-off — conforming your business to someone else's workflow in exchange for affordability — became so normalised that we stopped questioning it. Entire consulting industries emerged just to help companies implement and customise these platforms. The irony was never lost on anyone paying attention: businesses were spending enormous sums to make generic software feel slightly less generic.

What AI changes

AI-powered development tools are attacking the core economic assumption that made SaaS dominant. When the cost of building custom software drops by 80–90%, the calculus changes entirely.

Here's what's happening in practice. Tasks that once required a senior developer working for weeks — building a dashboard, creating an internal workflow tool, setting up a customer portal — can now be accomplished in hours or days using AI-assisted development. We're not talking about no-code drag-and-drop toys. We're talking about production-grade applications, built to exact specifications, that do precisely what a business needs and nothing more.

This matters because the "nothing more" part is actually a feature, not a limitation. Every SaaS platform comes loaded with complexity you don't need. You're paying for features built for other companies' use cases. Your team navigates interfaces designed to serve everyone, which means they're optimised for no one. Your data lives in someone else's schema, structured around their assumptions about how your business works.

A bespoke application, by contrast, can mirror your actual operations. It can use your terminology, follow your workflows, and surface exactly the information your people need — no training required, no workarounds, no "yeah, just ignore that tab."

The new economics of custom software

Let's put some rough numbers on this. A mid-sized company might spend £50,000–£150,000 per year on a suite of SaaS subscriptions — CRM, project management, HR tools, analytics platforms, and so on. That's recurring, forever, and it typically increases year over year as vendors raise prices and you add seats.

Building equivalent custom tools using AI-assisted development? The upfront cost is dropping fast, and the ongoing maintenance is increasingly manageable. More importantly, you own it. No per-seat pricing. No annual increases. No vendor lock-in. No panic when your favourite tool gets acquired and the new owners decide to "sunset" the features you depend on.

The maintenance argument — long the strongest card in SaaS's hand — is also weakening. AI doesn't just help build software; it helps maintain it. Bug fixes, feature additions, security patches — all of these become faster and cheaper when AI is involved in the development lifecycle.

Where SaaS still wins (for now)

Let's be fair about where this argument has limits. There are categories of SaaS that aren't going anywhere soon.

Platform-level infrastructure — cloud computing, databases, authentication services — these aren't really about workflow; they're about capability. You're not going to build your own AWS. Similarly, tools that derive their value from network effects, like communication platforms or marketplaces, benefit from having millions of users on the same system.

Highly regulated domains — accounting, compliance, payroll — carry enough legal complexity that most companies will still prefer a vendor who takes on that liability and keeps the software current with changing regulations.

And for very small businesses without any technical capacity, SaaS remains the path of least resistance. A sole trader doesn't need a bespoke CRM; they need something that works out of the box.

The middle is where things get interesting

The real disruption is happening in the vast middle ground — the workflow tools, the internal dashboards, the customer-facing portals, the reporting systems, the operational platforms. This is where most SaaS spending lives, and it's exactly the territory where AI-built bespoke applications make the most sense.

Think about it from a competitive standpoint. If your competitor is running the same Salesforce instance with the same templates and the same workflows as you, where's the advantage? But if they've built a customer management system that's perfectly tuned to their sales process, with AI-powered insights specific to their market, integrated exactly with their other systems — that's a genuine edge.

We're entering an era where software can be a differentiator again, not just a cost centre. And that's a profound shift.

What this means for SaaS companies

Smart SaaS companies are already seeing the writing on the wall, and the clever ones are adapting. Some are repositioning as platforms and APIs rather than end-user applications — providing the building blocks that bespoke applications are built on. Others are embedding AI deeply into their products to stay ahead of what a custom build could offer. A few are pivoting toward the regulated, high-liability niches where custom builds are still risky.

But for those selling generic workflow tools at premium prices? The next few years are going to be uncomfortable.

What this means for businesses

If you're a business leader, this shift is worth paying attention to now, even if you're not ready to act on it immediately. Start by auditing your SaaS spend with fresh eyes. For each tool, ask: is this providing genuine, irreplaceable value, or are we mostly paying for convenience that's getting cheaper by the month?

Consider running a pilot. Take one internal tool — something that frustrates your team, something you've been meaning to customise for years — and explore what it would take to build a bespoke replacement using AI-assisted development. The answer might surprise you.

The bigger picture

What we're witnessing isn't just a technology shift; it's a philosophical one. For twenty years, the software industry told businesses to conform to standardised tools. AI is now making it economically viable to flip that script — to build software that conforms to your business instead.

That doesn't mean every SaaS product disappears overnight. It means the bar for what justifies a subscription just got dramatically higher. The SaaS products that survive will be the ones that genuinely can't be replicated — because of their data, their network, their regulatory expertise, or their sheer technical depth.

Everything else? It's now a build-versus-buy conversation that, for the first time in two decades, might actually favour building.

The age of bespoke is coming back. Only this time, it's fast, it's affordable, and it's powered by AI.