There's a popular narrative in tech right now that goes something like this: AI will replace everything. SaaS is dead. Traditional software is on its way out. The future belongs entirely to probabilistic reasoning and large language models.
It's a compelling story. It's also wrong or at the very least, dangerously incomplete.
The reality is more nuanced, and frankly, more interesting. In a recent interview that Ai.nl co-founder Job van den Berg had with Woodson Martin — the CEO of OutSystems who joined the company in 2025 - laid out a vision that cuts through the hype. What we're witnessing isn't a revolution where one paradigm replaces another, he argues. It's an evolution, where two fundamentally different types of systems, deterministic and probabilistic must learn to coexist. And the enterprises that understand this distinction will be the ones that capture the real value of AI.
To understand why both systems matter, Martin points to two very different scenarios from OutSystems' customer base.
Scenario 1 — Probabilistic reasoning at work
Take a travel company like TravelEssence (one of OutSystems’ Dutch clients) that designs bespoke experiences. When one of his customers finishes a day of kayaking, he needs to recommend the right restaurant for dinner. Should it be the white-tablecloth spot with a wine pairing, or the relaxed food hall where they can refuel in flip-flops?
This is a judgment call. There's no single correct answer, it depends on the customer's personality, their mood, the trip so far. Probabilistic AI can be extraordinary here. It can weigh dozens of signals and arrive at a recommendation that's as good as sometimes better than a seasoned human advisor's instinct.
Scenario 2 — When determinism is non-negotiable
Now consider a fuel logistics operation. Millions of pounds of inventory need to flow through a specific set of pumps, into precise tanks, and onto designated train cars. The sequence matters. The volumes matter. The timing matters.
If you get this wrong, the cost isn't a slightly disappointed diner. It's enormous financial loss and extraordinary waste. You don't want a system that's "probably right." You need one that is certainly right. You need determinism.
The balance of probabilistic and deterministic systems is what every enterprise really needs. You need a deterministic harness to help you leverage the probabilistic AI.
Here's the concept that too many AI-first conversations are missing: enterprises don't just need probabilistic intelligence. They need a deterministic harness the structured, rule-based, guaranteed-correct scaffolding that gives AI a safe and productive environment to operate in.
Think of it like guardrails on a mountain road. The car's ability to navigate, choosing speed, adjusting to curves, reacting to conditions is the probabilistic part. But the guardrails are what prevent a small misjudgment from becoming catastrophic. Both are essential. Neither works alone.
In enterprise terms, this harness is the workflow logic, the data validation rules, the compliance checks, the operational constraints that ensure AI's creative suggestions don't violate physics, regulations, or the laws of supply chain management. It's the layer that says: "That's a fascinating recommendation, but it's physically impossible to route that volume through pump 7 at that time."
Excel at judgment, preference, pattern recognition, and ambiguity. Best for recommendations, personalization, content, predictions, and decisions where "good enough" is valuable and context changes constantly.
Excel at precision, compliance, sequencing, and constraint management. Best for logistics, financial calculations, regulatory workflows, and any process where being wrong has severe consequences.
The idea that AI will simply replace all existing software is a misreading of how technology has always advanced. Every prior generation of technology mainframes, client-server, the web, mobile, cloud, SaaS didn't obliterate what came before. It added a layer. Each layer of the stack solved a specific type of problem, and the previous layers continued to matter.
AI is following the same pattern. It's becoming a powerful new layer in the technology stack. And like every layer before it, it will make the layers beneath it work better. SaaS isn't going away. Deterministic workflow engines aren't going away. What's happening is that AI is making all of them smarter and more capable.
The companies that try to throw out all existing infrastructure and "go full AI" will learn a painful lesson. The companies that thoughtfully weave AI into a stack that includes deterministic guarantees will build something genuinely transformative.
According to Woodson Martin, there's a temptation especially in the venture-backed world to frame every shift as a revolution. Revolutions are exciting. They make good keynote titles. But what's actually happening in the enterprise is better described as an evolution: a steady, compounding improvement where AI amplifies what already works and opens new capabilities that weren't possible before.
That might sound less dramatic, but the impact is anything but small. When a company offers personalized, AI-powered recommendations and know that his booking system, inventory, and financial operations are running on deterministic logic that never misroutes a reservation, that's a genuinely better business. Not a flashy demo. A real, measurable advantage.
Key takeaway
The future of enterprise software isn't about choosing between AI and traditional systems. It's about the blend according to Woodson Martin. Probabilistic reasoning for judgment and personalization. Deterministic systems for precision and reliability. A harness that makes them work together. The enterprises that master this balance won't just adopt AI they'll make it actually work at scale.
.svg)
