Practical guide
Keep the instructions. Change the right information.
A variable is information that changes from one use to the next: the topic, the audience, the tone or the format. The prompt keeps its structure; only this information needs to be filled in.
Illustrative example
One example is enough to understand
Write a LinkedIn post announcing our new service. Target audience: small business leaders. Use a clear and professional tone and end with a question that invites a response.
- Aa Information you enter
Your request is ready.
Local demo: Do Prompt calls no AI and does not collect what you type. “Open in…” sends the prompt to the service you choose.
What stays, what changes
Instructions describe the expected work; variables adapt that work to the topic at hand. In the example above, the request stays the same: only three pieces of information change. A third category, durable context, hardly ever changes.
- Instructions: goal, rules, expected format.
- Variables: topic, audience, tone, language.
- Durable context: offer description, brand voice, glossary.
Start with the information that matters
Choose the information that changes often and genuinely shapes the request. Information that never changes is an instruction; a question that does not change the request is one question too many.
- Marketing: product, target, channel.
- Support: issue, language, tone.
- Consulting: assignment, industry, goal.
Keep durable context separate
A client description, a brand voice or a glossary do not change with every use. They can live in the workspace’s reusable content: prompts reference them, and you only fill in what varies.
- Durable information lives in one place.
- Prompts that reference it pick up its current value each time they are prepared.
- Text already copied into an external AI tool does not change.
Make the prompt easy to use
A prompt with too many variables becomes hard to use. Separate required information from options, offer a default value when possible and group stable information in the context.
- Field names people understand.
- Few required questions.
- Guided choices when they help.
Your turn: pick a prompt you use often
Spot the information you change every time: these are your first variables. Turn them into fields, keep the rest as instructions, then have a colleague try the prompt. Variables make the request clearer, but the answer still depends on the AI tool you choose: review the result before using it.
Explore further: variable types and scopes
A variable can take several forms: free text, a choice from a prepared list, or a value reused without typing it again. Its scope tells you where the value comes from and who fills it in.
- Task field: filled in by the person when preparing the request, such as the topic or the recipient.
- Choice list: a few prepared options that guide the answer, such as tone or channel.
- Reusable content: durable workspace information, picked up by the prompts that reference it.
- Company, user or system information, such as today’s date: filled in when the request is prepared.
Frequently asked questions
Do I need technical skills to create variables?
No. A variable is simply a question asked to the person using the prompt: which topic, for which audience, in which tone. Technical details remain optional.
Should we use the same variables everywhere?
It helps to align frequent variables, such as audience or tone, while keeping specific variables for certain roles or deliverables.
Do variables guarantee a better AI answer?
No. They reduce omissions and ambiguity in the request, but the answer depends on the AI tool and the task. Always review the result.
Your turn: pick a prompt you use often.
Spot the information you change every time: those are your first variables.
Free during beta
