The launch of GPT-5 has resurfaced the old chestnut that “AI is going to kill SaaS”. When I see this kind of generalisation, I get a little frustrated.
There’s no denying that GPT and other LLMs are rapidly being embedded into workflows, operations, and customer interactions. But let’s be clear: GPT-5 did not just obliterate the SaaS and enterprise app industry.
Yes, GPT-5 brings new features like better reasoning, code generation, and agentic capabilities. But there are also claims that it hallucinates more than GPT-4, which raises the question of how reliable it would be as a sole replacement for SaaS systems.
Ultimately, this isn’t an either-or choice; it’s a better together story. The most powerful applications of LLMs come from deep integration with SaaS platforms.
Why Predictability Still Matters
Much of what we do in business requires predictability and consistency. LLMs by their nature do not provide that; things like workflows, rules, and predictive logic do.
For many runtimes use cases, LLMs are less predictable – and far more expensive to run – than more traditional SaaS architectures. A GPT-5-powered prompt is no substitute for decades of enterprise software engineering. While GPT-5 can be great for ad hoc needs, it remains probabilistic: the same input won’t always deliver the same output. That’s a problem for enterprises that require compliance, auditability, and repeatable results.
By contrast, SaaS systems bring domain-specific workflows with predictive logic, ongoing maintenance, and rules that businesses rely on. For example, an order processing system in SaaS will always calculate tax in the same way, applying explicit business logic. They provide auditability and traceability that afford businesses historical oversight over workflows.
Relying on AI alone risks creating isolated proofs of concept that fail to integrate or scale, leaving gaps between strategy, processes, and customer experiences.
Trust is Non-Negotiable
For businesses, technology isn’t just about function; it also about trust.
SaaS platforms have earned that trust by embedding compliance controls, maintaining certifications, and offering service-level agreements (SLAs). If a SaaS CRM system goes down, accountability is clear. GPT-5, on the other hand, doesn’t provide the same contractual guardrails.
Trust extends to how sensitive information is handled. SaaS platforms have spent decades hardening their systems to protect everything from medical records to bank statements. These systems have been stress-tested in real world settings and have proven security frameworks. By contrast, GPT-5 stores information less securely and is vulnerable to risks such as prompt injection, model poisoning, or data exfiltration.
That said, as AI adoption accelerates, regulatory scrutiny is ramping up too. Initiatives like the 2024 EU AI Act aim to provide a framework for responsible and accountable AI that will enable businesses to adopt the technology with greater confidence.
The Sweet Spot: When SaaS Meets AI
Where LLMs do shine is as design engines. Gone are the days of trying to fit your software into rigid, pre-built workflows and logic. By using an LLM to shape the experience, with the proven backends of SaaS architectures, you can achieve the ‘vibe coding’ experience, while maintaining enterprise-grade confidence.
Further to vibe coding is what I like to call ‘vibe transformation’. Instead of just writing snippets of vibe code that quickly hit their limits in complex enterprise apps, vibe transformation is about starting with a business outcome and using AI to redesign processes for speed, adaptability, and better experiences. Meanwhile, SaaS ensures compliance and execution.
However, usefulness doesn’t mean SaaS systems should be replaced entirely. I’ve helped clients deploy RAG-based (Retrieval-Augmented Generation) answer agents, and while they can be powerful, they’re not a replacement. Shutting down existing search mechanisms can backfire, especially when customers still find them useful. We’ve seen this play out in real cases, where companies faced backlash after customers found AI tools slower and less reliable than traditional index-based search. These scenarios highlight that forcing AI as a substitute for SaaS can sometimes degrade the customer experience rather than enhance it.
The real value comes when LLMs enhance SaaS. At design time, AI can reinvent processes; at runtime, it can identify and execute the right workflows seamlessly. By embedding AI into existing platforms, businesses can speed up routine tasks, improve operational efficiency, and create intelligent feedback loops that enable systems to continually learn and develop.
A Hybrid Future
The reality is that businesses don’t need to choose between AI and SaaS.
Businesses that embrace a hybrid approach, that combines the integration, compliance and security frameworks of established SaaS platforms with the adaptability and automation provided by AI, will ultimately win.
This kind of integration will only succeed if businesses invest in people, training users to work effectively with AI-enhanced systems and extract maximum value from them.
AI has enormous potential, but it won’t deliver in isolation. Its real value comes from being woven into enterprise systems, and from knowing when to rely on deterministic SaaS logic versus when to harness the creativity of large language models.
