Ask Better Questions, Get Better Answers

Prompting Fundamentals for ChatGPT

Why Two People Using the Same Tool Get Completely Different Results

When the agency dug into what separated the two groups, the answer wasn't technical skill or AI experience. It came down to one thing: how people were asking. The frustrated group was typing short, vague requests, 'write a blog post about social media' or 'make this email better.' The successful group was giving ChatGPT context, a clear role, a specific format, and a defined audience before they asked for anything. They were, in effect, briefing the AI the same way they'd brief a talented but brand-new junior colleague who knew nothing about their client or their voice.

This is the central tension of using ChatGPT at work. The tool is genuinely capable, it can draft proposals, summarize 40-page reports, rewrite cold emails, prepare interview questions, and build slide outlines in minutes. But its output quality is almost entirely determined by the quality of your input. Most professionals using ChatGPT today are leaving 60 to 70 percent of its capability on the table, not because they lack technical knowledge, but because nobody taught them how to communicate with it effectively. That's what this lesson fixes.

The Principle: ChatGPT Responds to Context, Not Just Commands

The Brafton experiment reveals a principle that runs through every successful use of ChatGPT in a professional setting. The AI doesn't read your mind, and it doesn't have access to your industry knowledge, your company's tone, your client's quirks, or the specific outcome you need. When you give it a vague command, it fills in all those blanks with generic defaults, and generic defaults produce generic output. When you give it rich context, it produces something that actually fits your situation. Every technique in this lesson is a different way of providing that context efficiently.

The Four Elements That Make a Prompt Work

Priya Sharma is a senior HR manager at a logistics company in Atlanta with about 800 employees. When her company rolled out ChatGPT Plus for the HR team, Priya's first attempt was to ask it to 'write a job description for a warehouse supervisor.' What came back was technically correct but completely useless, a bland, one-size-fits-all description that could have come from any job board template. It didn't reflect her company's culture, the specific physical requirements of their facilities, or the fact that they needed someone who could manage a bilingual team. She spent 45 minutes rewriting it and ended up thinking ChatGPT wasn't worth the hype.

A colleague showed Priya a different approach. Instead of a single-line command, she structured her prompt around four elements: a role for ChatGPT to play, the specific context of her situation, the exact output she needed, and the audience who would read it. Her revised prompt told ChatGPT it was an experienced HR professional at a mid-sized logistics company, described the warehouse environment and team size, asked for a job description in a specific format, and noted that applicants would include both English and Spanish speakers. The output she got back required only minor tweaks before it was ready to post. Same tool. Completely different result.

Those four elements. Role, Context, Task, and Format, form the backbone of every strong prompt. You won't always need all four in equal measure, and sometimes one or two will do the job. But understanding what each one contributes gives you a reliable framework to diagnose weak output and fix it fast. When ChatGPT gives you something generic, you can almost always trace it back to one of these elements being missing or underdeveloped. Think of it as a checklist you run in your head before you hit send.

Click to enlarge

The Same Principle in a Completely Different World: Financial Services

Marcus Webb is a financial advisor at a regional wealth management firm in Charlotte. His client base includes small business owners and pre-retirees, people who are financially sophisticated but not investment professionals. Every quarter, Marcus needs to send personalized market update emails to about 80 clients, explaining what happened in the markets and what it means for their portfolios. Before he started using ChatGPT Plus, this took him the better part of a day. His early attempts with the tool produced emails that were either too technical (full of jargon his clients wouldn't connect with) or too vague (so hedged and generic they said nothing useful).

The shift came when Marcus started treating his prompts as client briefs. He told ChatGPT exactly who his clients were, what they worried about, what they already knew, and what tone matched his personal brand, measured, reassuring, plain-spoken. He also started specifying length (under 300 words), structure (three short paragraphs: what happened, what it means for them, what we're doing about it), and what to avoid (no doom language, no excessive hedging). Now he drafts a base email in about 10 minutes, then personalizes the opening line for each client segment. His quarterly email open rates went up 22 percent in two quarters. The content got sharper because the instructions got sharper.

Weak Prompts vs. Strong Prompts: A Direct Comparison

Click to enlarge

A Third Perspective: How a High School Teacher Changed Her Planning Workflow

Dana Kowalski teaches 10th-grade English in a public school in suburban Columbus, Ohio. She has five class sections, 28 students per class, and a prep period that rarely runs longer than 40 minutes. When her district encouraged teachers to explore AI tools in early 2024, Dana was skeptical, she'd tried asking ChatGPT to 'write a lesson on The Great Gatsby' and gotten something that felt like a Wikipedia summary dressed up as a plan. She filed it away and went back to her usual prep routine. Six months later, a colleague showed her what a fully-specified prompt looked like, and she tried again.

Dana's revised approach was to treat ChatGPT like a highly capable student teacher who needed a thorough briefing. She told it her grade level, the specific text and chapter, what students had already covered, the learning objective for that day's class, the time available, the format she preferred, and even the two or three misconceptions her students typically had about the material. Her output went from generic to genuinely useful, differentiated discussion questions, a structured Socratic seminar format, an exit ticket aligned to her state standards. She now estimates she saves about 90 minutes of planning time per week. That's meaningful in a job where time is the scarcest resource.

Putting the Framework Into Practice

The Role-Context-Task-Format framework isn't a rigid formula you fill out like a form. Think of it as a mental checklist you run before submitting any prompt that matters. For quick, low-stakes requests, 'give me five synonyms for efficient', you don't need all four elements. But for anything you'd spend 30 minutes or more doing manually, a proposal draft, a performance review, a client-facing report, running through all four elements will almost always produce better output on the first try than a quick, vague prompt followed by three rounds of frustrated follow-up.

The most common place professionals get stuck is the Role element. Telling ChatGPT to act as a specific type of professional, 'you are a senior account manager at a B2B software company' or 'you are an experienced executive coach', shapes the vocabulary, the assumptions, and the perspective it brings to your task. It's the equivalent of assigning the right person to a job rather than just posting it on a general board. You wouldn't ask your finance team to write your company's origin story for the website. Assigning the right role to ChatGPT works the same way.

The Format element is where many professionals also underinvest. ChatGPT defaults to flowing paragraphs unless you tell it otherwise. But most professional work products have specific structures: bullet-pointed briefings, three-paragraph executive summaries, numbered action plans, tables, slide-by-slide outlines. Specifying format isn't a minor detail, it's often the difference between output you can use directly and output you have to reformat entirely before it's usable. A simple addition like 'format this as a bulleted list with a one-sentence header for each point' can save 10 minutes of reformatting on every single output you produce.

Key Points From Part 1

1. ChatGPT's output quality is almost entirely determined by the quality of your prompt, the tool is capable, but it needs sufficient context to produce professional-grade work.
2. Vague, one-line prompts produce generic output because ChatGPT fills in missing context with defaults. Specific prompts produce specific, usable results.
3. The Role-Context-Task-Format (RCTF) framework gives you four concrete levers to pull when building any prompt that matters.
4. Assigning a professional Role to ChatGPT shapes its vocabulary, assumptions, and perspective, this single change often produces dramatically better output.
5. Specifying Format explicitly, bullets, numbered lists, tables, structured paragraphs, saves reformatting time and produces output you can use directly.
6. The Brafton agency case study, Priya's HR experience, Marcus's client emails, and Dana's lesson planning all show the same pattern: better instructions produce better results, regardless of industry or role.
7. Strong prompting is a clarity skill, not a technical one. Professionals who write good project briefs, clear meeting agendas, or precise client instructions already have the core capability, they just need to apply it to ChatGPT.

When the Brief Is Vague, the Output Is Useless

In 2023, a mid-sized consulting firm in Chicago. Navigant's commercial strategy team, ran an internal experiment. They gave the same ChatGPT prompt to twelve consultants: "Write a summary of our client situation." Every single person got back something different. Some responses were three sentences. Some were three pages. A few were written for a technical audience. Others read like a press release. The AI wasn't broken. The prompts were. Each consultant had a different client, a different context, and a different idea of what "summary" meant, and none of them told ChatGPT any of that. The outputs reflected exactly what they put in: almost nothing.

The team lead, frustrated with the inconsistency, did something smart. She created a prompt template for the whole team, a standard format that included the client's industry, the purpose of the summary, the intended reader, and the desired length. Within a week, the outputs were consistent, professional, and actually usable in client decks. The quality of the AI's work hadn't changed. The quality of the instructions had. That shift, from vague to specific, from informal to structured, is the single most powerful thing a non-technical professional can do to improve their results with ChatGPT.

This is the core tension in AI prompting for business professionals: the tool feels conversational, so we treat it like a quick chat with a colleague. But ChatGPT isn't a colleague who already knows your company, your client, your tone preferences, or what "good" looks like in your context. It knows nothing about you unless you tell it. The consultants who got poor results weren't lazy, they were using the wrong mental model. Once you understand that ChatGPT is more like a brilliant freelancer on their first day than a long-term team member, the way you communicate with it changes completely.

The Four Elements of a Strong Prompt

From the consulting firm's experience, we can extract a clean principle: every effective prompt answers four questions that ChatGPT cannot answer on its own. Who are you writing for? What do you want the output to look like? What context does the AI need to do this well? And what constraints should it stay within? These four elements, audience, format, context, and constraints, form the backbone of any prompt that consistently produces professional-grade output. You don't need all four in every prompt. A simple task might only need two or three. But when you're getting mediocre results, at least one of these is almost always missing.

Click to enlarge

A Hospital That Learned to Brief the AI Like a New Hire

The HR director at a regional hospital network in the UK, not a tech company, not a startup, started using ChatGPT to draft job postings in early 2024. Her first attempts were disappointing. She'd type "Write a job posting for a ward nurse" and get something generic enough to belong to any hospital in any country. The postings used buzzwords she'd explicitly told her team to avoid. They didn't reflect the hospital's values around patient-centered care. They weren't even formatted the way the hospital's careers page required. She almost gave up on the tool entirely.

Then she reframed the task in her own mind. Instead of thinking "I'm using AI," she thought "I'm briefing a new copywriter who has never been inside a hospital." She started providing the hospital's existing posting format as a template. She included a sentence or two about the department culture. She specified the tone, "warm but professional, not corporate", and the word count. She listed three phrases the hospital never used in job postings. The next batch of postings were so good that two colleagues asked if she'd hired a new communications person. She hadn't. She'd just learned to write a better brief.

The lesson here applies across every role and industry. ChatGPT performs like the quality of the brief it receives. A thin brief produces thin output. A rich, specific brief, even if it takes you an extra two minutes to write, produces output that is genuinely useful, often on the first try. The HR director now spends about three minutes writing her prompt and gets a usable draft in thirty seconds. Her old process, writing from scratch, sending to comms for review, revising twice, took the better part of a day. The math is obvious once you see it.

The Same Tool, Wildly Different Results: A Comparison

A sales manager at a B2B software company in Toronto ran a similar experiment to the Chicago consultants, but more deliberately. She asked two members of her team to use ChatGPT to draft a follow-up email after a discovery call with a prospect. One rep used a one-line prompt. The other used a structured prompt with context about the prospect, the pain point discussed, the next step agreed on, and the desired tone. The results were so different they might as well have come from different tools. The comparison below shows exactly what changed, and why it matters for any professional using ChatGPT in client-facing work.

Click to enlarge

Teachers Who Stopped Getting Generic Lesson Plans

Secondary school teachers in a Melbourne school district started using ChatGPT for lesson planning in 2023 after their district provided access through a Microsoft Copilot pilot. The early feedback was mixed. History teachers complained the AI gave them lesson plans that were "textbook but lifeless." English teachers said the suggested discussion questions were "too obvious." A math teacher said the word problems were too abstract for her Year 8 class. All of them were prompting with a single sentence: subject, topic, year group. That was it. And all of them were getting exactly what that minimal prompt deserved.

A curriculum coordinator noticed the pattern and ran a half-day workshop on prompt structure. By the end of the session, the history teacher was including the specific unit her class was studying, the prior knowledge students already had, the learning objective for that lesson, and a note that her students responded better to case studies than to timelines. Her next prompt produced a forty-five-minute lesson plan built around the 1973 oil crisis, with a primary source analyzis activity and a structured class debate. She used it almost unchanged. The shift wasn't the technology. It was learning to communicate with it the way you'd communicate with a talented substitute teacher who needed a proper handover.

Putting the Framework to Work in Your Own Role

The four-element framework, audience, format, context, constraints, works across every professional context, but the way you apply it shifts depending on your role. A marketer writing a campaign brief will front-load context: the product, the target customer, the campaign goal, the brand voice. A manager writing a performance review template will front-load constraints: what the company's HR system requires, what language to avoid, what rating scale is in use. An executive preparing for a board presentation will front-load audience: who's in the room, what they already know, what decisions they need to make. The framework is the same. The emphasis shifts based on what ChatGPT is least likely to guess correctly.

One practical approach is to build your own prompt library, a simple document or Notion page where you save prompts that worked well. This is especially useful for tasks you repeat regularly: weekly status reports, client proposals, meeting agendas, interview question sets, social media posts. Once you've written a strong prompt for a recurring task, you don't need to reconstruct it every time. You update the context-specific details, the client name, the meeting topic, the role you're hiring for, and keep the structure that already works. Within a month of consistent use, most professionals have five to ten reusable prompt templates that save them hours every week.

The other move worth making immediately is learning to iterate rather than abandon. When a ChatGPT response misses the mark, the instinct is often to delete it and start over, or worse, to conclude that AI isn't useful for that task. The better move is to treat the first response as a draft and tell ChatGPT exactly what's wrong with it. "This is too formal, rewrite it in a warmer tone." "The second paragraph is too long, cut it to three sentences." "Add a specific example from the retail sector." This follow-up prompting, sometimes called chaining, is how professionals consistently get from a decent first draft to something genuinely excellent, without starting from scratch every time.

Key Principles from This Section

1. Vague prompts produce vague outputs, the AI reflects the quality of the brief it receives, not the quality of the tool itself.
2. The four-element framework (audience, format, context, constraints) covers the information ChatGPT needs to produce professional-grade output.
3. Think of ChatGPT as a brilliant new hire on day one, capable and fast, but needing a proper brief to do good work in your specific context.
4. The emphasis you place on each framework element should shift based on your role and what ChatGPT is least likely to guess correctly.
5. Save prompts that work, a personal prompt library for recurring tasks is one of the highest-ROI habits you can build as a regular ChatGPT user.
6. Iterate rather than abandon, follow-up prompts that correct specific issues are faster and more effective than starting over from scratch.
7. The gap between a weak and a strong prompt for the same task can represent twenty or more minutes of editing time per output.

Section Takeaways

• Inconsistent ChatGPT results are almost always a prompting problem, not a tool problem.
• Four elements, audience, format, context, constraints, are the foundation of any prompt that reliably produces usable output.
• Real professionals across consulting, healthcare, sales, and education have transformed their results simply by writing better briefs.
• Prompt libraries turn one-time effort into ongoing efficiency, write the prompt well once, reuse it forever.
• Follow-up prompting (chaining) is how you get from good to excellent without starting over.

Giving ChatGPT a Role, a Reader, and a Reason

In 2023, the marketing team at Patagonia's retail division faced a familiar problem: they had one product, a new line of recycled fleece jackets, but four completely different audiences to reach. Outdoor enthusiasts. Corporate sustainability managers buying bulk gifts. Parents shopping for teenagers. And eco-conscious millennials on a budget. Writing four separate campaigns from scratch would take weeks. A junior marketer on the team had been quietly experimenting with ChatGPT. She didn't write four separate prompts. She wrote one prompt structure, and changed three variables each time: the role she gave ChatGPT, the reader she described, and the reason the message mattered to that reader.

The results were striking. The same AI tool produced copy that felt genuinely tailored, not just tonally different, but structurally different. The sustainability manager version led with supply chain data. The parent version led with durability and value. The millennial version opened with a climate statistic. Her manager thought she'd hired a freelancer for each segment. She hadn't. She'd simply learned that ChatGPT performs like a skilled professional when you treat it like one, giving it a clear job title, a clear audience, and a clear purpose before asking it to produce anything.

The principle she stumbled onto is one of the most transferable in all of prompting: context is not decoration. It is instruction. When you tell ChatGPT who it is, who it's talking to, and why the output matters, you are not being verbose, you are being precise. Vague prompts produce generic output. Specific context produces specific, usable work. This applies whether you're writing marketing copy, summarizing a legal document, building a training module, or drafting a performance review.

Click to enlarge

How Context Changes Everything. A Hospital Story

St. Joseph's Health System in Michigan rolled out an internal communications initiative in 2024. HR needed to send a policy update about hybrid work to three groups simultaneously: frontline nurses who couldn't work remotely, administrative staff who could, and department heads who had to manage both. One HR business partner used the same ChatGPT prompt template for all three drafts, swapping audience context each time. The nurse version acknowledged the unfairness directly and focused on what benefits they were gaining instead. The admin version was enthusiastic and action-oriented. The manager version was structured like a briefing with FAQs embedded.

None of those tonal choices came from the HR team dictating exact sentences. They came from clearly describing the audience's situation and emotional context in the prompt. ChatGPT inferred what a nurse facing this news would need to hear versus what a manager implementing this policy would need to know. The HR partner's prompts didn't just say 'write an email to nurses.' They said: 'Write an email to ICU nurses who cannot work from home and may feel this policy is unfair. Acknowledge their situation honestly. Focus on what hasn't changed for them.' That specificity is the skill.

This is what separates professionals who get 60% of ChatGPT's value from those who get 95% of it. The tool can read emotional nuance into a situation, but only if you put that nuance into the prompt. You are not programming a machine. You are briefing a very capable colleague who knows nothing about your workplace, your audience, or your constraints unless you tell them.

A Sales Team That Stopped Writing Proposals From Scratch

A mid-sized management consulting firm in Chicago had a business development problem. Senior consultants were spending four to six hours writing first-draft proposals for prospective clients. The proposals were good, but the process was exhausting, and it bottlenecked the pipeline. A BD manager started building what she called 'prompt templates', not code, just saved text files with structured prompts that captured the firm's voice, their typical client profile, and the proposal's purpose. She'd fill in the specifics for each new pitch and hand it to ChatGPT. First drafts that took half a day now took twenty minutes.

The key wasn't automating the thinking. The consultants still made all the strategic decisions, which case studies to reference, which pricing model to propose, which risks to address. ChatGPT handled the drafting, the structuring, and the language. The BD manager described it as 'having a brilliant writer who already knows our style and never needs to be told how a proposal is structured.' That's an accurate description of what a well-primed ChatGPT session delivers. The prompt did the briefing. The AI did the drafting. The humans did the judgment.

The Teacher Who Differentiated an Entire Lesson in 15 Minutes

A high school history teacher in Austin, Texas used ChatGPT to solve a problem that had frustrated her for years: differentiation. Her class included students reading at a 6th-grade level and students in the school's gifted program, all studying the same content. Writing two versions of every handout was theoretically ideal and practically impossible given her workload. She started using ChatGPT with a simple role-and-audience prompt: 'You are an experienced curriculum designer. Rewrite this passage about the New Deal for a student reading at a 6th-grade level. Use short sentences, define difficult words in context, and include one analogy to modern life.'

Then she ran a second prompt for her advanced students: same role, different audience description, different purpose. Both versions took under five minutes each. She now differentiates every major reading and assessment this way. The lesson here isn't specific to education, it's universal. Any time you serve multiple audiences with different needs, ChatGPT can draft a version for each one, as long as your prompt describes each audience with enough specificity that the AI understands what 'useful' means for that person.

Putting It Into Practice

The most common mistake professionals make when starting with ChatGPT is treating it like a search engine, typing a short question and hoping for a complete answer. The better mental model is a talented new team member on their first day. They're capable. They're eager. But they need context. They need to know what you're trying to achieve, who the work is for, what success looks like, and what constraints matter. A good prompt is a good briefing. It doesn't have to be long, it has to be specific.

Start with your most repetitive writing tasks. The emails you send every week. The reports you format the same way every month. The meeting summaries you dread writing. These are ideal first use cases because you already know what good output looks like, which means you can evaluate what ChatGPT gives you and refine your prompt quickly. Iteration is part of the process. If the first draft isn't right, don't start over. Tell ChatGPT what to fix: 'Make the tone less formal' or 'Shorten this to three bullet points' or 'Add a call to action at the end.'

The professionals who get the most from ChatGPT are not the ones who know the most about AI. They're the ones who are clearest about what they need and most willing to describe it explicitly. That's a communication skill, not a technical one. You already have it. The task below gives you a structured way to practice the full prompt-building method right now, using a real piece of work from your own job.

Key Lessons From This Section

1. Context is instruction: telling ChatGPT who it is, who the audience is, and what the output must accomplish is not optional detail, it directly determines output quality.
2. The three-variable rule works across every profession: Role + Audience + Purpose transforms generic AI output into specific, usable work.
3. Emotional and situational context belongs in your prompt: if your audience has concerns, constraints, or a specific mindset, describe it explicitly so ChatGPT can reflect that understanding.
4. Iteration is normal and fast: a follow-up message refining tone, length, or structure takes ten seconds and often produces better results than rewriting the original prompt.
5. Your most valuable AI asset is a prompt library: saving your best prompts turns one-time effort into reusable infrastructure that saves hours every week.
6. The skill is clarity, not technology: professionals who describe what they need precisely will always outperform those who type vague requests, regardless of which AI tool they use.

Takeaways

• ChatGPT performs like a skilled professional when briefed like one, give it a role, an audience, and a purpose before asking for output.
• Specificity in your prompt is the single biggest lever for improving output quality without any technical knowledge.
• The same prompt structure works across industries: marketing, HR, consulting, education, sales, and operations all benefit from the Role + Audience + Purpose framework.
• Follow-up messages are prompts too, use them to refine, not restart.
• Build a Prompt Templates document now. It is the highest-ROI habit you can develop as an AI-assisted professional.