A while back, a few loud voices in tech declared that SaaS was dead. The idea was simple. AI would replace the application layer sitting on top of your data, and software as we know it would disappear.
I don't think that's right. But I do think something real is happening underneath that headline, and it's worth working through properly.
The data layer isn't going anywhere
Start with the foundation. Every business runs on data. Customer records, transactions, invoices, contracts. That's where the truth lives. Nothing about AI changes that. If anything, AI makes clean, well organised data more valuable than ever, because an agent is only as good as the information it works from.
So the real question was never whether the data layer would survive. It obviously will. The question is what happens to the layer sitting on top of it, the applications, dashboards and forms that people have spent the last twenty years clicking through.
The application layer isn't dying either. It's being rebuilt
Here's where I land. AI isn't going to kill the application layer. It's going to change who, or what, is doing the work inside it.
For years, software has meant people logging in, clicking buttons, filling in fields, and hitting save. That's the model we've all grown up with. What's changing is that a lot of that clicking and typing can now be done by an agent instead of a person.
Picture the accounts team. Today, someone opens an invoice, reads it, and types the numbers into the accounting system. Tomorrow, they scan the invoice, an agent reads it, understands it, and books it into the ledger in seconds. No typing required. If the job needs more than one kind of expertise, a second agent with different knowledge steps in, and the two work together to get it done properly.
That's not the death of software. That's software doing more of the work itself, with a person watching over the process rather than driving every step of it.
Multiple agents, working together, with people overseeing rather than doing
I think this is the real shape of what's coming.
An agent inside the application handles most of the repetitive, well defined tasks, the clicking, checking and saving that used to take up someone's day. Several agents talk to each other to get bigger jobs done, each one bringing its own area of expertise, a bit like specialists on a team. And people move from doing the task to overseeing it, checking the results, stepping in when something looks off, and making the judgement calls that still need a human. They'll be involved, just not at the scale they are today.
That last point matters. This isn't about removing people from the picture. It's about moving them up a level, from typing the invoice to reviewing the exceptions the system flags.
So will SaaS die? No. But the business model underneath it will change a lot
This is the part I feel strongly about, and it turns out the wider industry is landing in the same place.
The bit that's actually breaking isn't the software itself. It's how it's priced. For twenty odd years, SaaS has largely meant paying per user, per month, whether or not that user logs in once a day or once a month. That model made sense when the value came from a person sitting in front of a screen. It makes a lot less sense when the user doing most of the work is an agent.
What's replacing it is pricing based on outcomes. You pay for invoices processed, cases resolved, contracts reviewed, rather than paying for seats. Industry forecasts now expect the classic pricing per seat model to be largely gone by 2028, with most software vendors reworking how they charge around the value actually delivered rather than the number of logins.
That's a bigger deal than it sounds. It doesn't just change an invoice. It changes what vendors are even trying to build. A product priced per seat wants you logging in often, that's how it justifies the cost. A product priced on outcomes wants to get the job done with as little human time as possible. The less you notice it, the better it's doing its job. Those are two genuinely different design goals, and it means the software itself will start to look and behave differently, not just be billed differently.
It also raises some practical questions that businesses will need to get right.
How do you prove the outcome actually happened? If you're paying per invoice processed rather than per seat, the system needs to track exactly what it did and to what standard. That record becomes the bill itself.
Who is responsible when the agent gets something wrong? When a person used to click save, they carried the responsibility for the mistake. When an agent posts a financial entry on its own, someone needs to own what happens when it's wrong. This is why good error handling and clear escalation to a person will matter more than a nice looking interface ever did.
How do you price fairly when not every job is equal? Some invoices take five seconds to check, others need real judgement. Pricing purely by item only works if the items are roughly similar in effort. Getting this right is closer to how insurers price risk than how software has traditionally been priced.
None of this means SaaS disappears. Vertical software, tools built deeply around one industry's data and workflows, is actually growing faster than generic, one size fits all tools, precisely because that depth of data and workflow knowledge is hard to copy. The winners won't be the businesses that shout about being AI powered. They'll be the ones that quietly rebuild their pricing, their trust and their error handling around a world where the main user of the software is increasingly another piece of software, not a person.
Final take
This changes more than software. It changes how we, as businesses and as people, interact with the tools we rely on every day. We're moving from being the ones doing the work inside an application, to being the ones making sure the work being done is correct. That's a genuinely different relationship with software, and it's one that businesses of every size will need to get comfortable with, not just the big tech players.
SaaS isn't dying. It's growing up.