Automation, agents, or an app? What you actually need.

These three words are used interchangeably in sales conversations and mean genuinely different things in engineering. Buying the wrong one is the most common expensive mistake in this market.

Automation: rules you already know

If this, then that. No judgement required. Moving data between systems, triggering notifications, generating a document from a template. Cheap, reliable, and frequently the right answer — a lot of what gets sold as AI is automation with a language model bolted on for decoration.

If your process has no ambiguity, do not pay for AI. Pay for automation.

Agents: judgement inside rules

Reads something unstructured, decides what it means, takes an action within boundaries you set, escalates when uncertain. This is where AI earns its premium — claims, prior authorisations, contract review, transcript evaluation. It costs more because it needs permissions, audit trails and human override designed in.

Apps: a product your users touch

An interface built around AI that your customers or staff use directly — a web application, a mobile app, an assistant. The AI is a component; the product is the thing. Priced like software, because that is what it is.

How to tell which one you need

Answer these in order:

• Is there ambiguity in the input? No → automation. Yes → keep going.
• Does a human need to see the output before anything happens? Yes → an assistant. No → an agent.
• Do your users need somewhere to go and do this themselves? Yes → an app.

Many businesses need a combination, built in that order. Almost nobody needs all three at once, and any vendor proposing all three in the first engagement is selling a budget rather than solving a problem.