Many of us are using AI tools like Claude, ChatGPT, and Copilot every day, but the results are sometimes leaving us frustrated. The outputs don’t always live up to what we had in mind.
Before assuming the tool is falling short, it’s worth looking at how we’re asking for help, because the way we ask matters more than we might think.
A prompt is simply the instruction we give an AI tool: a question, task, or request typed into the chat window. It can be a single sentence or a detailed, multi-paragraph brief with other documents attached. It’s the main way we communicate what we want, and it shapes the response we get back.
Prompts are not the only factor shaping AI-generated results. Most tools also allow users to establish ongoing preferences for tone, format, and areas of focus through their personalization settings. Because these preferences carry across conversations, they can significantly impact the responses the tool provides and are definitely worth spending some time on, especially as you get started using a new AI tool.
But today, I want to focus on three principles that can improve any single prompt, regardless of what settings are already in place.
Principles for Better Prompts
- Give the tool context, not just a task. Remember, your AI tool isn’t a mind reader. It only works off what we tell it. Instead of asking, “Write a memo about our new policy,” try: “Write a two-paragraph memo for department directors explaining the new visitor policy in plain language, with a warm but direct tone.” The better context we provide about audience, purpose, and tone, the closer the first draft will be to what we actually need. However, going too far the other direction is also not helpful.
- Making prompts longer doesn’t necessarily lead to better results. The more you place into one conversation, the more the LLM has to weigh and it can inadvertently start prioritizing recent materials over things buried in the middle of a huge upload. In fact, a 2025 Chroma study found that performance grows increasingly unreliable as the input gets longer. Our prompting needs to be focused and not just longer. Anthropic’s own guidance recommends starting a fresh conversation for a new topic rather than letting one thread grow indefinitely. This leads to my last recommendation that builds on this point.
- Be specific about what “good” looks like. If you have a format or structure in mind, share it with the tool. Asking for “a summary” invites guesswork; asking for “a summary formatted as three bullet points, each under 20 words” gives the tool a clear target. When possible, include an example of similar work you liked. The AI tool will work to match its shape and style.
As with any skill, we can get better at prompting with a little practice.
The next time a response doesn’t land quite right, take another look at your prompting language. Use a few moments to create a clearer, more specific prompt to increase your odds for a usable result on your very next output.
Until next time, stay curious, lead thoughtfully, and keep humans at the center of the work.
By Chris Stump, Managing Associate | AQORD Consulting
Share This: