AI Prompting Guide

Stop Guessing.
Start Prompting.

Prompting is no longer a technical trick — it's a crucial, learnable skill. Master the five core principles and transform AI from a cool tool into a paradigm-shifting efficiency machine.

Learn the Framework See Examples
Scroll

A few years removed from AI's initial boom, prompting is the skill that separates average users from power users. Across real projects — from building a personal portfolio to designing an AI receptionist service — the five principles of Persona, Product Description, Tone, Constraints, and Call-to-Action consistently deliver precise, high-quality results. Here's how to use them.

The Framework

Five Principles of Effective Prompting

01
🎭

Persona

Assign the AI a role before the task. A WW2 historian will frame answers differently than a general assistant. Persona shapes vocabulary, tone, and judgment — tapping directly into the expertise you need. You can also "pre-warm" the AI by asking if it knows a topic, pointing to sources, or uploading documents.

02
📋

Product Description

AI must know exactly what you want to produce. Without a clear description, it fills the gaps with generic assumptions. Describe your project, its purpose, and your expected output in detail. The fewer assumptions the AI makes, the closer the result is to your vision.

03
🎯

Tone

Without guidance, AI defaults to a flat, generic middle-ground. Specify who the audience is and the desired tone. Writing an email to a guest speaker? Say so. Want it to sound professional and warm? Say that too. Tone specificity is the difference between passable and polished.

04
🚧

Constraints

Often, telling AI what not to do is more powerful than telling it what to do. Set limits: no jargon, no unverified claims, specific word counts, only certain sources. Constraints eliminate unwanted directions and sharpen the AI's focus on exactly what you need.

05

Call-to-Action

Give the AI one clear task at a time. Just like multitasking reduces human performance, it reduces AI quality. For multi-deliverable projects — report, use case, and model — prompt each piece separately with its own focused CTA. When tasks conflict, start a new chat.

The Prompt Formula
🎭 Persona + 📋 Description + 🎯 Tone + 🚧 Constraints + ⚡ CTA = Precise Output
Before vs. After

See the Difference in Action

❌ Weak Prompt
Write me an email.
✅ Strong Prompt
You are a professional communications specialist. I need to write an email to a guest speaker I'm inviting for my student organization, BIAS. The tone should be professional, warm, and grateful. Do not exceed 150 words. Do not use filler phrases or clichés. The CTA is to confirm their availability for a panel on April 15th at 5 PM.
Why this works
The persona establishes expertise. The description clarifies context and audience. The tone removes ambiguity. The constraint sets a clear length limit. The CTA gives one specific goal — confirm a date.
❌ Weak Prompt
Explain machine learning to me.
✅ Strong Prompt
Act as a data science professor explaining a concept to a sophomore business student who has no coding background. Avoid technical jargon. Use one real-world analogy from business or finance. Do not assume the student knows statistics. Limit your response to 200 words. Explain: what machine learning is and why it matters for business analysts.
Why this works
Setting the audience ("sophomore business student, no coding background") forces the AI to calibrate complexity. The analogy constraint ensures relatability. The topic is narrowed to a business context — not a blanket definition.
❌ Weak Prompt
Help me with my project. I need a report, a model, and a use case.
✅ Strong Prompt
You are a senior business analyst. I am working on a capstone project analyzing supply chain disruptions. Right now, focus only on the report. Audience: professor evaluating MIS students. Tone: academic, concise. Do not include the use case or model yet. Format: Executive Summary (100 words), Problem Statement, 3 Key Findings. Write the report outline now.
Why this works
Splitting a multi-deliverable project into one CTA per prompt prevents confusion and keeps quality high. Explicitly telling the AI to ignore the other deliverables further reduces cognitive load and drift.
Anatomy of a Prompt

Breaking Down a Real Prompt

Annotated Prompt Example — AI Receptionist Brief
🎭 Persona
You are an experienced product manager and UX writer who has built AI-driven customer service tools for startups.
📋 Description
I am building SimpleSec, a bilingual (English/Spanish) AI receptionist service for small businesses in El Paso, TX. It handles inbound calls, routes questions, and books appointments.
🎯 Tone
The audience is non-technical small business owners. Tone: conversational, confident, and approachable — not corporate or overly technical.
🚧 Constraints
Do not use AI buzzwords like "leverage" or "synergy." Limit response to 3 paragraphs. Only describe features I've specified — do not invent new ones. Pull wording from the context I've provided, not general knowledge.
⚡ CTA
Write a 3-paragraph product description for SimpleSec suitable for a pitch deck slide aimed at potential business clients.
Sample Output →
SimpleSec is your always-on, bilingual front desk — handling calls, routing inquiries, and booking appointments 24/7 so you can focus on running your business. Designed for small businesses in the El Paso–Juárez corridor, it speaks your customers' language — literally — switching seamlessly between English and Spanish.

No more missed calls or double-booked appointments. SimpleSec integrates with your existing calendar and CRM tools, routing each caller to the right service or scheduling window with zero friction. Think of it as your most reliable team member, available even when you're not.
Going Further

Advanced Best Practices

📚

Feed Examples

Give the AI multiple samples of what you want. When generating content for BIAS, provide existing BIAS material. More examples = higher accuracy.

🔍

Ask It to Explain

For nuanced outputs, ask the AI to explain its reasoning. This surfaces assumptions and catches errors before they reach your final deliverable.

🧑‍💼

Human in the Loop

For automated tools — especially those that take real-world actions like sending emails — always include a review step before execution. Risk mitigation matters.

🔄

Iterate, Don't Restart

Treat your first output as a draft. Follow-up prompts that refine a single aspect ("now make it more concise") outperform writing a new prompt from scratch.

🗂️

Separate Complex Tasks

When tasks are unrelated or risk confusing the AI, switch to a new chat. Clean context windows produce cleaner outputs.

🎓

Prompting Is a Skill

Like writing or coding, prompting improves with deliberate practice. The structured five-principle framework makes that practice repeatable and systematic.