Better Prompting: Getting More from AI Tools

The previous post in this path covered prompt engineering as clear thinking — the idea that working effectively with AI starts with knowing what you want. This post builds on that foundation with specific techniques that consistently produce better results from AI language models.

These techniques are not tricks. They are communication skills. The same principles that make you a better communicator with humans — being specific, providing context, giving examples, stating expectations — make you a better communicator with AI. The difference is that AI takes your instructions literally. It cannot read between the lines, infer your unstated preferences, or ask clarifying questions unless you prompt it to.

Be Specific About What You Want

"Write me an email" is a prompt. It will produce an email. It will probably not produce the email you wanted, because the AI has no idea who the recipient is, what the purpose is, what tone is appropriate, or how long it should be.

"Write a professional email to my team lead requesting a deadline extension for the Q2 report. The current deadline is Friday. I need until next Wednesday because the data from the finance team arrived two days late. Keep it brief and direct — my team lead prefers concise communication." This prompt contains the context, constraints, and preferences that produce a useful result on the first try.

The specificity principle: every piece of information you leave out is a decision the AI will make for you. Sometimes those decisions are fine. Often they are not what you wanted, and you spend more time editing the output than you saved by being vague. Front-load the specificity and you get usable output faster.

Provide Examples

When you want output in a particular format or style, showing an example is more effective than describing it. This is sometimes called "few-shot prompting" — giving the AI a few examples of what you want before asking it to produce more.

If you want product descriptions in a specific style, provide two or three examples of descriptions you like. If you want code that follows your team's conventions, show a snippet that demonstrates those conventions. If you want summaries at a particular level of detail, show a summary you consider good.

The AI learns the pattern from your examples and applies it to new input. This works better than describing the pattern in words, because patterns are easier to demonstrate than to articulate. You might struggle to describe what makes a product description "punchy and benefit-focused," but you can easily point to one that is.

Assign a Role

Telling the AI who it should be when responding provides implicit context about expertise, vocabulary, and approach. "You are a senior software engineer reviewing code" produces different feedback than "You are a technical writer editing documentation" — even when the underlying content is the same.

The role does not need to be elaborate. "Act as an experienced project manager" is enough to shift the response toward project management vocabulary, concerns, and frameworks. "Act as a patient teacher explaining to a beginner" adjusts the complexity and tone toward accessibility.

This works because the role frames the response. An "experienced data analyst" will notice different things in a dataset than a "marketing manager" will. Assigning the right role for your task gets you a response shaped by the relevant expertise.

Break Complex Tasks into Steps

A single prompt asking the AI to "analyze this data, identify trends, create visualizations, and write a report" combines four distinct tasks. The AI will attempt all four simultaneously, and the quality of each will suffer because it is spreading its attention across the entire request.

Instead, break complex tasks into sequential steps. First: "Analyze this data and identify the three most significant trends." Review the output. Then: "For each trend, explain the likely cause and business implication." Review again. Then: "Create a summary suitable for a non-technical audience." Each step builds on the previous one, and you can course-correct between steps.

This approach — sometimes called "chain-of-thought" when the AI does it internally — produces better results because each step gets full attention and you maintain control over the direction. If step one identifies the wrong trends, you correct it before step two builds on a flawed foundation.

Tell the AI What Not to Do

AI models have default behaviors that may not match your preferences. They tend to be verbose, they add caveats and qualifiers, they use formal language, and they repeat your question before answering it. If these defaults annoy you, say so.

"Do not include an introduction or conclusion. Do not repeat my question. Do not add caveats. Answer directly." This kind of negative instruction — specifying what you do not want — is surprisingly effective at shaping the output toward your actual preferences.

Similarly, if you want the AI to stay within certain boundaries: "Only use information from the document I provided. Do not add outside knowledge." Or: "If you are unsure about something, say so rather than guessing." These constraints prevent the most common frustrations with AI output.

Use the Output to Refine the Input

Your first prompt rarely produces perfect output. That is expected. The real skill is using the imperfect output to refine your next prompt.

If the response is too long, add a length constraint: "Keep it under 200 words." If the tone is wrong, specify: "Use a more casual tone, as if explaining to a colleague." If it includes information you did not want, exclude it: "Do not mention pricing." If it misunderstood your intent, rephrase: "I meant competitive analysis, not feature comparison."

This iterative approach — prompt, review, refine, re-prompt — is how experienced AI users work. They do not expect the first output to be final. They treat the AI as a collaborator that needs direction, and they provide that direction through specific, concrete feedback.

The efficiency gain comes from learning which specifications matter for your recurring tasks. After a few iterations, you develop prompt templates that produce good results on the first try for the types of tasks you do frequently.

Context Is Everything

The AI only knows what you tell it in the current conversation. It does not know your company, your audience, your preferences, or your constraints unless you provide them.

For recurring tasks, build a context block that you include with your prompts. "I work at a B2B SaaS company. Our audience is mid-market HR leaders. Our tone is professional but approachable. We avoid jargon. We never make claims we cannot support with data." Prepending this context to your prompts produces output that is consistently aligned with your needs.

For one-off tasks, take thirty seconds to think about what context the AI needs to do a good job. Who is the audience? What is the purpose? What constraints apply? What does "good" look like? Answering these questions for yourself — and including the answers in your prompt — is the single biggest improvement most people can make to their AI results.

The Takeaway

Better prompting is not about memorizing magic phrases. It is about clear communication: being specific about what you want, providing examples of what good looks like, assigning an appropriate role, breaking complex tasks into steps, constraining what you do not want, and providing the context that shapes a useful response.

The investment is small — an extra minute of thought before you press enter. The return is output that is usable on the first or second try instead of the fifth. Over time, the patterns become second nature, and working with AI tools becomes a genuinely efficient part of your workflow.

---

Next in the "Prompting and Working with AI" learning path: We'll cover prompt templates for common tasks — ready-to-use frameworks for writing, analysis, coding, and decision-making that you can adapt to your specific needs.