In the 1990s, adding “.com” to a company’s name could send valuations soaring, even if the business had little real internet capability. During the blockchain boom, almost any company mentioning “crypto” or “Web3” suddenly appeared innovative. Beverage firms became “blockchain companies”. Kodak launched a cryptocurrency initiative.
Today, “AI” risks becoming the latest version of the same corporate costume.
A large portion of what is marketed as AI today is still fundamentally hard-coded logic, statistical models or traditional algorithms wrapped in fashionable terminology. If a system merely follows predefined rules — “if X happens, do Y” — it is not intelligence. It is automation. Calling a glorified decision tree “AI” does not make it intelligent, any more than putting a racing sticker on a Proton makes it a Ferrari.
The term AI has become so commercially valuable that many firms now use it less as a technical description and more as a branding strategy — a way to appear more advanced, more scalable, more futuristic and more valuable than they really are. In some cases, “AI-powered” simply means a chatbot connected to a database.
See also: The great data delusion — where to invest for AI winners
An algorithm is simply a set of instructions — it follows a predefined pathway. Examples include traffic lights adjusting based on traffic flow; airline ticket pricing changing with demand; GPS rerouting based on congestion, e-commerce recommendation engines; banks auto-rejecting loans based on credit scores or credit card fraud detection; trading systems executing buy/sell rules automatically; dynamic surge pricing in ride-hailing apps; retail inventory auto-reordering systems; and rule-based cybersecurity alerts.
These systems may be useful, complex and efficient, but they do not reason, infer, create or truly adapt beyond what was explicitly designed.
Traditional software is deterministic. If the inputs are known, the outputs are predictable. Modern generative AI is different because it can generate new content — text, images, code, analysis and reasoning pathways — from learned patterns rather than explicit instructions. It is probabilistic, not fully deterministic.
See also: The AI grid rewrites the value chain
Agentic AI goes further still. It can plan tasks, interact with tools, make decisions dynamically, iterate towards objectives, and operate with partial autonomy.
In other words:
- A traditional algorithm is like a calculator following formulas;
- Generative AI is like a writer synthesising an idea; and
- Agentic AI is like a junior employee attempting to complete a task independently.
The real AI revolution is not software becoming faster. It is software beginning to imitate elements of reasoning, judgement, adaptability and autonomous actions. And that is precisely why the term should not be diluted into meaningless marketing jargon.
