SaaS Is Dying. Long Live Whatever Comes Next

Software as a Service is not about to vanish tomorrow. But its future is changing fast, and not in a way many software vendors will enjoy.

For the last two decades, SaaS won by turning software into a subscription. Instead of buying a package once, businesses rented access to a standardised tool delivered over the web. That worked brilliantly for vendors. Build once, sell repeatedly, stack customers into recurring revenue, then upsell modules, premium tiers and services over time. For UK SMEs, it was often a reasonable trade: less infrastructure to manage, predictable billing, and access to tools that would once have been out of reach.

The problem is that SaaS was built around a hidden assumption: that businesses would continue adapting themselves to software. AI may flip that around. The next stage looks less like "buying another SaaS platform" and more like asking a machine to build, configure and maintain something much closer to the business itself. Not perfect bespoke software overnight, not yet. But increasingly, software that is generated, adapted and orchestrated around the way the business actually works, rather than forcing the business to contort itself around a generic product. Gartner now predicts that by 2028, 80% of generative AI business applications will be developed on existing data management platforms, and that those platforms could cut complexity and delivery time by 50%. That is not the death of business software. It is a warning that the shape of business software is changing.

For UK SMEs, this shift could be an opportunity. For traditional software vendors, especially the kind living off costly, rigid, one-size-fits-most systems, it looks more like a threat.

The real question is no longer whether SaaS survives

The real question is how long before a business owner can say something like:

"Build me a system for manufacturing workflow, stock control, job tracking, purchasing and reporting that fits how we actually work."

…and have the computer produce something good enough to run.

That still sounds futuristic, but it is no longer ridiculous. Reuters reported last week that the hottest job in enterprise AI right now is the "forward deployed engineer" because buying access to a powerful model is easy, while integrating it into messy real-world systems is not. That single point matters more than most vendor marketing. It tells you that the technology is moving toward bespoke application assembly, but that the hard part is no longer intelligence alone. The hard part is stitching that intelligence into real processes, real data and real constraints.

So the answer is not "yes, next month". It is closer to this: in the next two to five years, many UK SMEs are likely to gain access to AI-assisted systems that can configure or generate large chunks of their workflow software far faster than today. In the five-to-ten-year range, it becomes much more plausible that businesses will increasingly commission software by description rather than by selection from a menu of rigid SaaS products. But "and it just works" is still the dangerous part. Gartner expects over 40% of agentic AI projects to be scrapped by the end of 2027 because of escalating cost, unclear value or poor controls, even while forecasting that agentic AI will be embedded in a third of enterprise software by 2028. That is exactly what a transition period looks like: the direction is real, the hype runs ahead, and the messy middle hurts people.

SaaS is evolving into something less standardised and more generated

This is the part software vendors should worry about.

Classic SaaS depends on repeatability. The same product, the same interface, the same logic, pushed across thousands of customers with limited customisation because customisation is expensive, awkward and hard to support. AI changes that equation because it lowers the cost of tailoring logic, interfaces, automation and data handling.

Not to zero. Not safely in all cases. But enough to threaten the old economics.

Once a model can generate forms, workflows, reports, data mappings, automations, integrations and even chunks of application logic on demand, the value of a rigid SaaS product starts to erode. The product no longer has to be the thing. The system generated around the business becomes the thing. Gartner's forecast that 40% of enterprise applications will feature task-specific AI agents by the end of 2026 supports that broader direction: software is moving from static interface-driven tools toward task execution and orchestration.

That is why this is not really a story about "AI features" being added to SaaS. It is a story about SaaS slowly dissolving into AI-mediated business systems. The winners may not be the companies with the neatest dashboards. They may be the companies whose tools can be re-shaped, generated and governed around specific customer needs.

For UK SMEs, that is potentially excellent news. Many smaller firms have never been perfectly served by standard SaaS products. They either overpay for bloated systems built for larger organisations, or underfit themselves into smaller products that do not match the way they actually operate. If AI makes it easier to build near-bespoke systems at a lower cost, the buyer gains leverage for the first time in years.

This is probably bad news for a lot of software providers

That is not just theory. Markets are already nervous.

Reuters reported in February that software and data-services stocks were hit by a selloff so severe it wiped roughly $830 billion to $1 trillion from market value in a matter of days as investors worried that AI tools could upend the sector. Reuters described the episode as "software-mageddon", with big names such as ServiceNow, Salesforce and Microsoft among those hit. The reason is simple: if AI can reduce the scarcity value of traditional software workflows, then high SaaS multiples become harder to justify.

Reuters Breakingviews made the same point more bluntly in August 2025, arguing that valuations may not fully reflect the pain AI could cause SaaS companies. Later pieces suggested the selloff was exacerbating private-equity exit problems and delaying software IPO plans. In March 2026, Deutsche Bank upgraded software again, saying the worst of the downturn may be over, but even that upgrade was framed against a backdrop of a prolonged slump driven largely by AI disruption fears. In other words, the market may be calming, but it has already received the warning.

Is there a reported effect on software providers in the UK?

Yes, although the UK evidence is more visible in pressure and anxiety than in a neat "AI killed this company" headline.

The cleanest direct UK example right now is Capita, which is not a pure SaaS business but is a major UK technology-enabled outsourcing and services provider. Reuters reported today that Capita's shares fell 15% after it warned of a margin decline, with weakness in contact-centre operations and rising costs. Crucially, the article notes changing market dynamics including increased use of AI in customer support. That is not proof that AI alone is wrecking British software and service firms, but it is clear evidence that AI-driven changes are already shaping revenue quality, pricing pressure and investor reactions in UK-listed tech-enabled businesses.

There is also broader market evidence that UK and European software-related names are being repriced because of AI fears. Reuters reported that UK wealth-management shares fell in February as AI disruption fears spread across Europe, and noted that software companies' shares had already suffered a sharp selloff on similar concerns. LSEG, another UK-listed business with a strong data and software dimension, has been under activist pressure partly around margins and partly around whether its Microsoft partnership is delivering enough visible benefit. That is not a straightforward SaaS case, but it fits the same pattern: markets are asking much harder questions about software economics in an AI era.

The near future is messy, not magical

The likely near-term future for UK SMEs is not "describe a whole ERP and it appears flawlessly by Friday". It is more uneven.

First will come AI-assisted configuration. More software will include agents that can create workflows, automate repetitive tasks, generate reports, build forms, design interfaces and connect data sources with much less manual work. Microsoft's push into Copilot Cowork is a useful example. Reuters reported that the tool can independently create applications, manage data and build spreadsheets, and that Microsoft is folding this capability into M365 Copilot with extra usage sold on top. That matters because it shows the large platforms moving from assistant-style software toward delegated task completion and application assembly.

Then comes AI-assisted orchestration. Instead of just helping inside an app, the system starts coordinating work across multiple systems. That is where on-demand business systems become more plausible. A business does not necessarily buy one giant application; it describes the process and an agentic layer assembles the required pieces on top of databases, automations and existing enterprise tools.

Only after that does the more radical future start to emerge: software commissioned by intent rather than selected from catalogues.

The medium future is worse for vendors than for buyers

If AI keeps moving in this direction, classic SaaS vendors face a nasty squeeze from both sides.

At the bottom end, low-code, no-code and AI-generated workflow tools make it easier to produce acceptable business software without paying premium subscription rates to rigid vendors. At the top end, the major platforms — Microsoft, Google, OpenAI, Anthropic, maybe a few others — become the orchestration layer on which more software is generated and managed. That leaves a lot of mid-tier SaaS providers in an ugly middle. Too expensive to justify as static software. Not powerful enough to own the AI layer. Too easy to replicate around the edges.

This is why "SaaS evolving into AI-generated business systems" is not just a feature story. It is a value-capture story. The question becomes: who captures the value when software is easier to generate than to sell? If AI makes logic, interface and workflow generation cheaper, the scarcity shifts away from the old SaaS wrapper. That is why markets have been so jumpy.

UK SMEs may finally get the upper hand — if they are patient

There is a real opportunity here for buyers.

For years, smaller businesses have been told to accept compromises because bespoke software was too expensive and enterprise software was too bloated. SaaS often meant choosing the least bad option. AI-generated and AI-configured systems may finally start to shift that balance. Not because every business will become a software company, but because the cost of shaping software around a business is likely to fall.

That could be particularly useful for UK SMEs, which often sit in the awkward middle: too complex for very small business tools, too small for heavyweight enterprise suites. If AI can compress the cost and time required to produce tailored workflow systems, smaller buyers gain leverage they have not had for a long time.

But patience matters. Reuters reported in December that business leaders still believe AI is the future, but wish it worked properly now, while Forrester expected companies to delay about a quarter of planned AI spending by a year. That fits the real picture: the direction is right, the speed is overhyped, and the implementation layer is still hard.

So how long before you can ask the computer to design a manufacturing workflow, stock control and job-tracking system and have it just work? Near future: it will increasingly help design and assemble one, but human oversight, integration work and testing will still matter. Medium future: the first draft is generated largely by AI and refined by specialists rather than built from scratch. Longer term: many businesses may stop buying software in the old sense and start commissioning systems by intent.

The warning for software vendors

Software providers should be very nervous about mistaking "AI features" for a sufficient response.

Adding a chatbot to a clunky product is not a strategy. Nor is sprinkling "agentic" over a pricing page and hoping no one notices the software underneath still behaves like 2019. Gartner has already warned that much of the current agentic-AI market is hype-driven and that only a fraction of vendors offer the real thing. If your value rests mainly on workflow convenience rather than deep domain data, trusted execution or serious integration capability, your moat may be thinner than you think.

The threat is not just that AI replaces coders or support staff. The threat is that AI reduces the scarcity of the software product itself.

And once that happens, the customer starts asking a more dangerous question:

Why am I paying a premium subscription for your standard system when I can increasingly have something built around mine?

If you want to talk about what this shift means for your business systems, software buying strategy or the future of bespoke workflow tools for UK SMEs, use the contact page.