A vendor told one of our clients their new "AI-powered claims processing system" would transform operations. We looked under the hood. It was a rules engine with a chatbot bolted on. The rules engine was fine. Solid, reliable, doing what rules engines do. But calling it AI set expectations it could never meet, and when it didn't "learn" or "adapt," the client wrote off AI entirely.
That's the damage hype-washing does. Not to the vendor (they got paid). To the enterprise that now thinks AI doesn't work.
What Is and Isn't AI
Let's be direct:
Not AI: Rules engines, decision trees, if-then logic, workflow automation, RPA bots following scripts, SQL queries with conditional logic, Excel formulas, hardcoded scoring systems.
AI: Systems that learn from data, recognise patterns, generate content, understand natural language, make predictions based on statistical models, or adapt their behaviour based on new information.
The distinction isn't academic. It determines what the system can and can't do, how it should be maintained, what governance it needs, and what value it can deliver over time.
A rules engine is deterministic: give it the same input, you get the same output. That's a feature, not a limitation. It's auditable, predictable, and explainable. It's also completely incapable of handling ambiguity, processing unstructured text, or improving with experience.
An AI system is probabilistic: it handles ambiguity and adapts, but it also needs monitoring, governance, and fundamentally different expectations around accuracy and error handling.
Why the Mislabelling Hurts
It sets wrong expectations. When a client buys "AI-powered" software expecting it to learn and improve, and what they get is a static rules engine, the inevitable disappointment poisons the well for actual AI initiatives.
It misallocates governance effort. Calling a rules engine "AI" triggers unnecessary governance overhead (risk assessments, bias audits, monitoring frameworks) that a deterministic system doesn't need. Meanwhile, actual AI systems may get less scrutiny because the organisation is already experiencing "governance fatigue."
It obscures real opportunities. If a business process is being handled by a rules engine labelled as AI, nobody asks whether actual AI could do it better. The label implies the AI opportunity has already been captured.
It erodes trust in AI. Every "AI-powered" product that's really just automation contributes to the growing scepticism that AI is all hype. That scepticism makes it harder to get funding, executive support, and user adoption for genuine AI capabilities.
40%
of 'AI startups' in Europe did not use AI in any meaningful way in their products
Source: MMC Ventures, The State of AI Report, 2024
The Honest Taxonomy
Here's what to call things:
| What It Does | What to Call It |
|---|---|
| Follows predefined rules | Automation |
| Moves data between systems | Integration |
| Follows a scripted workflow | Process automation or RPA |
| Learns patterns from data | Machine learning |
| Understands and generates language | AI (specifically, LLMs) |
| Combines retrieval with generation | AI (specifically, RAG) |
None of these are better or worse than the others. A well-built rules engine solving the right problem is more valuable than a poorly deployed AI system solving the wrong one. The point isn't that everything should be AI. It's that everything should be called what it actually is.
What to Do About It
Next time a vendor says "AI-powered," ask three questions:
- Does it learn from new data, or does it follow predefined rules? If it's rules, it's automation. That's fine. Just price it and govern it accordingly.
- What happens when it encounters something it hasn't seen before? AI handles novel inputs (with varying accuracy). Rules engines either have a rule or they don't.
- Does it improve over time without manual updates? If every improvement requires someone editing rules or logic, it's not AI. It's software.
The enterprises that get the most value from AI are the ones that know exactly what AI is, and exactly what it isn't. They deploy rules engines where determinism matters and AI where adaptation matters. They don't confuse the two, and they don't let vendors confuse them either.
Stop calling everything AI. Start calling it what it is. Your AI strategy will be better for it.