Home · Insights · Fundamentals
Do you need to learn prompt engineering? Probably not
Most of what is sold as prompt engineering is just clear briefing. Here is the short version that covers what a business owner needs, with the rules we use on our own website assistant as an example.
Short answer: you need less than the internet says
"Prompt engineering" suggests a skill you must master before AI is useful. For a business owner using an assistant day to day, what you need is closer to good briefing: say what you want, who it is for, what to use, and what a good result looks like. Anthropic's own prompting guidance starts from the same place: be clear and direct, give context, and test your prompt against what you want. Nothing in it requires special tricks.
The four things that do most of the work
- Context. Who you are, who the output is for, what it is used for. A reply to a long-standing client reads differently from one to a cold enquiry.
- The task, in one sentence. "Write a reply that confirms the booking and asks for the intake form" beats "help with this email".
- The material to use. Paste the text, the price list, the policy. Do not rely on the assistant to know it.
- What good looks like. Length, tone, what to leave out, and anything it must not do. Examples help: one good past reply is worth a paragraph of description.
A rule set from our own chat widget
The assistant on our website is a good example, because it runs unattended and has to behave. Its instructions give it one job, tell it to ask at most one clarifying question, keep replies to two to four sentences, say that pricing depends on scope and never state a number, never invent a client or a result, and offer a next step only once it understands the situation. That is a short brief. It is not a clever prompt. It is a plain brief with clear limits. We describe the build in how we built the chat widget on our site.
When prompting does matter more
If you are building something that runs without you, like a reply drafter or a classifier inside an automation, the prompt becomes part of a system. Then it is worth treating like any other piece of the system: write what a good output is, keep a handful of test inputs, and rerun them whenever you change the prompt or the model. That is the same discipline as the comparison in choosing between assistants, and it is where outside help earns its place.
What to stop worrying about
- Magic phrases. There is no incantation that makes a vague request precise. Be precise.
- Prompt libraries. A pasted prompt written for someone else's business gives you their tone. Write your own context once and reuse it, as in running a business with Claude Projects.
- Getting it right first time. Treat the first answer as a draft and say what to change. That is how the tool is meant to be used.
The skill worth building is not phrasing. It is knowing which of your tasks are safe to hand over and checking the output, which we cover in what AI cannot do yet. If you want a hand setting up the first few, that is the work behind the AI systems we write and maintain for Mallorca businesses.