What ChatGPT Is (and Isn't)
ChatGPT is a general-purpose conversational AI built by OpenAI. Its genuine strength is versatility: it handles writing, brainstorming, analysis, coding, translation, summarisation, and creative tasks within a single conversation. It is the most widely used AI assistant in the world, and for good reason — it does a broad range of things competently, and a narrower range of things exceptionally well. If you've tried it and felt like you weren't getting much out of it, that's completely normal — most people's first few conversations with ChatGPT are underwhelming.
Where ChatGPT genuinely excels: writing and content creation (drafts, editing, tone adjustment, format conversion), conversational problem-solving (talking through ideas iteratively, refining outputs across multiple exchanges), the Custom GPTs ecosystem (thousands of specialised assistants built on top of ChatGPT for specific tasks), and image generation via GPT Image (available on all plans, with usage limits on free accounts). ChatGPT is also remarkably good at adapting to context mid-conversation — you can steer it, correct it, and build on its responses in ways that feel natural.
Where ChatGPT is less distinctive: the free tier uses GPT-5.3 (with a 10-message-per-5-hour cap before falling back to a lighter model) and includes ads as of February 2026. Its training data has a knowledge cutoff, so without web browsing it cannot answer questions about very recent events. Image understanding is available but not its primary strength. For tasks that require live data, deep integration with specific software outside its supported connectors, or highly specialised domain knowledge, other tools may serve you better.
You access ChatGPT at chatgpt.com (desktop) or through the ChatGPT mobile app (iOS and Android). OpenAI offers several models under the ChatGPT umbrella — GPT-5.3 is the default for all users, GPT-5.4 Thinking is the most capable reasoning model (available on paid plans with a model picker), and GPT-5.4 Pro is the deepest-thinking variant for the hardest tasks. Free and Go users also get access to GPT-5.4 mini via the Thinking feature. The model you use affects the quality, speed, and capabilities of responses. This learning path focuses on skills that work across all models, with notes where the model matters.
Get Started
Open chatgpt.com in your browser and sign in (or create a free account). You will see a text input area at the bottom of the screen. No setup or configuration is required.
Type this prompt exactly:
I'm preparing for a job interview at a marketing agency. Act as an experienced hiring manager and give me the 5 most common interview questions for a mid-level marketing role, with a brief explanation of what the interviewer is really looking for with each question.
Read the response carefully. Notice three things:
- Role adoption — ChatGPT shifted its perspective to that of a hiring manager, not a generic assistant. The advice should feel like it comes from someone who has conducted interviews, not someone reading a textbook.
- Structure — the response should be organised clearly, with each question followed by the reasoning behind it. ChatGPT tends to format its responses well without being asked.
- Actionability — the explanations should give you something you can actually use to prepare, not just vague reassurance.
If your first attempt didn't produce anything exciting, that's fine — the goal right now is just to see how ChatGPT responds, not to craft the perfect prompt.
If the response felt useful and specific, you have just experienced ChatGPT's core value: turning a vague need ("help me prepare") into structured, actionable output through a single well-crafted prompt. Everything in this learning path builds on that dynamic.
Core Skill 1: Conversational Prompting
ChatGPT's defining feature is not any single response — it is the conversation. Unlike a search engine, where each query is independent, ChatGPT remembers everything you have said in the current thread. This means your second message builds on your first, your third builds on both, and so on. Learning to use this conversational memory deliberately is the single most important skill for getting good results.
The key principle: start broad, then steer. Your first prompt sets the direction. Your follow-up messages refine, redirect, and deepen. Most beginners try to get the perfect answer in one prompt. If that's been you, you're in good company. Experienced users treat the first response as a rough draft and shape it through conversation.
How conversational memory works
Last verified: March 2026
- Context window — GPT-5.4 supports roughly 272,000 tokens of conversation context. That is approximately 200,000 words. For most conversations, you will never hit this limit.
- Memory across conversations — ChatGPT offers a Memory feature on paid plans (Go, Plus, Pro, Business, Enterprise) that carries facts and preferences between conversations. Free accounts do not have access to Memory. The core detailed context, however, is still per-thread.
- Instruction persistence — if you tell ChatGPT "respond in British English" at the start of a conversation, it will maintain that throughout the thread. Use this to set tone, format, audience, and constraints early.
- Correction works — if ChatGPT misunderstands you, simply say "That's not what I meant. I'm looking for X." It adjusts without needing you to rephrase the entire request.
🔢272,000 tokens
GPT-5.4's context window — approximately 200,000 words or 500+ pages of conversation in a single thread. Most users never hit this limit.
Source: OpenAI, 2026
If you set a rule like "respond in British English" at the start of a conversation, do you need to repeat it for every new question in that thread?
No. Instructions set early in a conversation persist throughout the entire thread. ChatGPT carries forward tone, format, and constraint instructions automatically — which is why setting rules up front is so powerful. You only need to repeat or reinforce a rule if ChatGPT starts drifting from it in very long conversations.
Follow-up patterns that improve output
Last verified: March 2026
- "Make it shorter / longer / simpler / more detailed" — direct length and complexity adjustments
- "Now do the same thing but for [different audience/context]" — reuses the structure while shifting the target
- "What did you leave out?" — forces ChatGPT to surface information it omitted for brevity
- "Challenge your own answer. What are the weaknesses in what you just said?" — triggers self-critique and more balanced output
- "Give me three alternative approaches" — breaks ChatGPT out of its first-response pattern
Exercise: The Refinement Chain
Scenario: You need to write a professional bio for a conference website.
Here's what we'd suggest trying: Start a new conversation and send this prompt:
Write a 100-word professional bio for me. I'm a project manager with 8 years of experience in the tech industry. I specialise in agile methodologies and have led teams of up to 20 people. I've worked at both startups and enterprise companies.
Then send these follow-ups, one at a time:
Good start, but it sounds too generic. Make it more distinctive — what would make someone want to attend my talk specifically?
Better. Now adjust the tone so it sounds like I wrote it myself, not like a PR team wrote it. Keep it confident but conversational.
Almost there. Add one sentence at the end about what I'm currently excited about in the industry — pick something plausible and interesting.
What to observe: Track how the bio evolves across four exchanges. Each follow-up should produce a noticeably different version. Notice that you never had to restate your background — ChatGPT remembered it from the first message.
Reflection: Compare the first version to the final version. Which would you actually use? Could you have gotten to the final version in a single prompt, or did the iterative process produce something better?
Exercise: Set the Rules Early
Scenario: You are preparing internal documentation for a non-technical team.
Here's what we'd suggest: Open a new conversation and send this setup prompt first:
For this entire conversation, follow these rules: (1) Write at an 8th-grade reading level. (2) Never use jargon without defining it in parentheses. (3) Use bullet points instead of long paragraphs. (4) End every response with a one-sentence summary.
Then ask:
Explain how two-factor authentication works and why our company is requiring it for all employees starting next month.
Then ask a completely different question in the same conversation:
Now explain our new expense reporting process: employees submit receipts through the app within 5 business days, managers approve within 3 days, and reimbursement is processed in the next payroll cycle.
What to observe: Did ChatGPT maintain all four rules across both responses, even though the second question had nothing to do with the first? Did you need to remind it of the rules?
Reflection: Setting constraints once and having them apply to an entire conversation is one of ChatGPT's most underused features. Where in your work do you repeatedly need content in a consistent style or format?
Core Skill 2: Writing and Content Creation
Writing is ChatGPT's strongest domain. Not because it writes better than a skilled human, but because it writes faster than any human at a usable first-draft level. The skill is not asking ChatGPT to write for you — it is learning to direct it precisely enough that its output requires minimal editing rather than a full rewrite.
When we first started using ChatGPT for writing, our prompts were embarrassingly vague. The difference between our early outputs and what we get now is night and day — and it came down to learning to specify three variables.
Three variables control the quality of ChatGPT's writing output: tone (how it sounds), format (how it is structured), and audience (who it is written for). Specifying all three in your prompt dramatically improves the result. Omitting any of them produces generic output.
Tone control
ChatGPT defaults to a helpful, slightly formal, mildly enthusiastic tone. That default is fine for some tasks and wrong for many others. Be explicit:
Write this in a warm but professional tone, like an email from a trusted colleague — not a corporate announcement.
Match the tone of The Economist: measured, analytical, slightly dry, assumes an intelligent reader.
Naming a specific publication, author, or style reference is often more effective than describing the tone abstractly.
Format specification
Tell ChatGPT exactly what structure you want. "Write a blog post" will give you a generic five-paragraph essay. Compare that to:
Write a blog post structured as: a provocative opening question, three sections each with a subheading and a real-world example, and a closing paragraph that gives the reader one specific action to take today. Total length: 600 words.
The more structural detail you provide, the less editing you will need to do.
Knowledge Check
You need ChatGPT to write a project update email for your team. Which prompt will produce the most usable first draft?
Exercise: Tone Transformation
Scenario: You have a piece of content that needs to work for multiple audiences.
Give this a go: Start with this prompt:
Write a 150-word announcement that our company is switching from Slack to Microsoft Teams starting next month. Write it in a neutral, informational tone.
Then request three rewrites in the same conversation:
Rewrite this for the company's internal newsletter — make it upbeat and emphasise the benefits, but don't be patronising.
Now rewrite it as a brief, no-nonsense message from the IT department with specific dates and action items.
Finally, rewrite it as a FAQ document anticipating the top 5 questions employees will ask.
What to observe: All four versions convey the same core information but feel completely different. Notice how ChatGPT adjusts vocabulary, sentence length, and structure — not just word choice.
Reflection: In your work, how often do you need the same information repackaged for different audiences? How much time would this save compared to writing each version from scratch?
Exercise: The Editing Partner
Scenario: You have already written something and want ChatGPT to improve it, not replace it.
Here's what we'd suggest: Take a real piece of your own writing — an email you drafted, a paragraph from a report, a social media post. Paste it into ChatGPT with this prompt:
Here is something I wrote. Do not rewrite it from scratch. Instead, give me: (1) three specific suggestions to make it clearer, (2) any sentences that are unnecessarily wordy with tighter alternatives, and (3) one thing I did well that I should keep doing. Preserve my voice — do not make it sound like AI wrote it.
What to observe: Does ChatGPT respect the instruction to edit rather than rewrite? Are the suggestions genuinely useful, or generic writing advice? Does the "one thing you did well" feel specific to your text?
Reflection: Using ChatGPT as an editor rather than a ghostwriter preserves your voice while improving your output. This is often more valuable than asking it to write from scratch.
Core Skill 3: Analysis and Problem-Solving
ChatGPT can break down complex problems into manageable components, examine data you provide, and help you think through decisions systematically. It is not a replacement for domain expertise, but it is an effective thinking partner — especially when you need to organise scattered thoughts, consider angles you have not thought of, or stress-test an idea before committing to it.
The key principle: give ChatGPT something to work with. "Help me solve this problem" produces generic advice. Pasting actual data, a real scenario, or a specific document and asking for analysis produces genuinely useful output. You can also upload files directly — all plans support file uploads (free accounts are limited to 3 per day), with paid plans offering higher limits and the full Advanced Data Analysis experience.
Knowledge Check
You want ChatGPT to help you evaluate whether your department should hire a contractor or a full-time employee. What is the most effective approach?
Exercise: Decision Framework
Scenario: You are facing a real decision — professional or personal — and want to think it through more rigorously.
Task: Describe the decision to ChatGPT with full context, then prompt:
I need to decide between [Option A] and [Option B]. Here is the context: [describe the situation, constraints, priorities, and what matters most to you]. Build me a structured comparison: pros and cons of each option, the key risks I should worry about, what I would regret most if each option went wrong, and your recommendation with reasoning. Then play devil's advocate against your own recommendation.
What to observe: Does ChatGPT identify trade-offs you had not considered? Is the devil's advocate section genuinely challenging, or does it just repeat the cons list in different words?
Reflection: The value here is not ChatGPT making the decision for you. It is forcing you to articulate your priorities clearly enough for an AI to analyse them — which often clarifies your own thinking.
Exercise: Code and Data Interpretation
Scenario: You have encountered something technical that you need to understand quickly — a spreadsheet formula, a code snippet, an error message, or a data pattern.
Task: Find a real example from your work or use this sample prompt:
Explain what this Excel formula does in plain English, then tell me if there are any edge cases where it might give wrong results: =IF(AND(B2>100,C2="Active"),B2*0.15,IF(C2="Active",B2*0.10,0))
What to observe: Does ChatGPT explain it clearly enough that you could now modify the formula yourself? Does it catch genuine edge cases (such as empty cells, text in B2, or misspelled status values)?
Reflection: ChatGPT is particularly good at translating between technical and non-technical language. Where in your work do you regularly encounter technical artefacts that would benefit from a plain-language explanation?
Core Skill 4: Custom GPTs and the Ecosystem
OpenAI's GPT Store contains thousands of specialised assistants — called Custom GPTs — built on top of ChatGPT. These are pre-configured versions of ChatGPT with specific instructions, knowledge files, and sometimes external tool connections baked in. Think of them as ChatGPT pre-loaded with expertise for a particular task.
You can browse Custom GPTs from within chatgpt.com by searching or navigating the GPT Store. You can search by category (writing, productivity, research, coding, education) or by keyword. Each GPT has a description and user ratings.
When Custom GPTs add value
Last verified: March 2026
- Repetitive specialised tasks — a GPT pre-configured for writing cover letters, reviewing contracts, or generating SQL queries saves you from rewriting the same setup prompt every time
- Tasks requiring specific knowledge — some GPTs are loaded with documentation, style guides, or reference material that plain ChatGPT does not have
- Workflow integrations — some GPTs connect to external services (Zapier, Canva, etc.) to take actions beyond text generation. On Plus and above, ChatGPT also has native app connectors (called "Apps") for Gmail, Google Drive, Slack, Notion, GitHub, and 20+ other productivity tools
- Learning and tutoring — GPTs designed for teaching specific subjects often have better pedagogical structure than asking plain ChatGPT to teach you
When plain ChatGPT is sufficient
Last verified: March 2026
- One-off tasks — if you only need something once, configuring or finding a Custom GPT is overhead you do not need
- Tasks where you want full control — Custom GPTs impose their creator's instructions, which may conflict with what you actually want
- General conversation and brainstorming — plain ChatGPT with a good prompt is usually better than a narrowly configured GPT
- Sensitive or confidential work — third-party GPTs may have unclear data handling. For sensitive content, use plain ChatGPT where OpenAI's data policies apply directly
You write the same type of client proposal every week, each time spending 5 minutes setting up the same context and constraints in ChatGPT. Is this a good case for a Custom GPT?
Yes — this is the ideal use case. A Custom GPT lets you bake your recurring instructions, tone, format, and reference material into a reusable assistant. The test is frequency: if you find yourself writing the same setup prompt more than twice a week, a Custom GPT pays for itself immediately. For one-off or highly variable tasks, plain ChatGPT with a good prompt is faster.
Beyond Text: ChatGPT's Multimodal Toolkit
Last verified: March 2026
If you have been using ChatGPT purely for text conversations, you are leaving some of its most practical capabilities untouched. ChatGPT can generate images (via GPT Image), analyse uploaded files (PDFs, spreadsheets, code), browse the web for current information, and run Python code through Code Interpreter — all within the same conversation. The trick is knowing which mode fits which problem, because each has genuine strengths and real blind spots.
| Mode | What It Does Well | When to Use It |
|---|---|---|
| Image Generation (GPT Image) | Creates original visuals, diagrams, mockups, social media graphics | You need a quick visual and do not have a designer available — presentation slides, concept illustrations, social posts |
| File Uploads (PDFs, spreadsheets, docs) | Reads and analyses documents you provide — summarises, extracts data, answers questions about the content | You have a 30-page report and need the key findings in 2 minutes, or a spreadsheet you want analysed without writing formulas |
| Web Browsing | Searches the internet for current information, verifies facts, pulls real-time data | You need information newer than ChatGPT's training data — recent news, current pricing, live statistics |
| Code Interpreter (Python sandbox) | Runs actual code — data analysis, chart generation, file conversion, calculations on your uploaded data | You have a CSV or Excel file and want charts, statistical analysis, or data cleaning without opening a spreadsheet |
A good rule of thumb: if your task involves something you can see, count, or verify with data, one of these modes will likely get you further than text alone.
Exercise: Multimodal Problem Solving
Scenario: You have a real document — a PDF report, a spreadsheet, or a data file — and you want ChatGPT to do more than just summarise it.
Here is what we would suggest trying:
- Find a real document from your work — a quarterly report, a budget spreadsheet, a project plan, or even a downloaded bank statement. If you do not have one handy, download any free public dataset or annual report PDF.
- Upload it to ChatGPT and ask a specific analysis question — not "what does this say?" but something like "Which three expense categories grew fastest quarter-over-quarter?" or "What are the top 5 risks mentioned in this report, ranked by how often they appear?"
- Request a visual summary — ask ChatGPT to generate a chart, table, or diagram based on the data it extracted. For spreadsheets, Code Interpreter can produce actual charts; for PDFs, ask for a structured visual breakdown.
- Evaluate the result — check at least 2-3 data points against the original document. Did ChatGPT get the numbers right? Did it miss anything important? Where did it add genuine value versus where did it struggle?
Reflection: Most people discover that ChatGPT is remarkably good at pulling patterns from data but occasionally misreads specific numbers. Knowing where to trust it — and where to double-check — is the real skill here.
Exercise: Find and Evaluate a Custom GPT
Task: Go to the GPT Store and search for a Custom GPT relevant to your work. Pick one with good ratings and try it with a real task. Then open a new plain ChatGPT conversation and do the same task with a well-crafted prompt.
[After testing both] Which gave the better result — the Custom GPT or plain ChatGPT with a detailed prompt? What did the Custom GPT do that you couldn't replicate with prompting alone?
What to observe: Did the Custom GPT save you time on setup? Was the output quality noticeably different? Did the GPT impose constraints that helped or hindered?
Reflection: The best Custom GPTs save you from writing the same complex prompt repeatedly. The worst ones add constraints without adding value. Your evaluation skill matters more than the GPT's rating.
Exercise: Build Your Own Custom GPT
Prerequisite: ChatGPT Go, Plus, Pro, Business, or Enterprise subscription (Custom GPT creation is not available on the Free plan).
Task: Think of a task you do repeatedly at work that requires specific context every time. Create a Custom GPT for it:
In chatgpt.com, navigate to the GPT creation interface (the exact location may vary as OpenAI updates the UI — look for "Create a GPT" or "My GPTs" in the sidebar or profile menu). In the instructions, describe: what this GPT does, the tone it should use, the format it should output, and any background knowledge it needs. Upload any relevant reference documents.
Test it with three different inputs to see if the instructions hold consistently.
What to observe: Does the GPT maintain its instructions across different inputs? Where does it deviate? Do you need to refine the instructions?
Reflection: Building a Custom GPT teaches you more about effective prompting than any other exercise, because you are forced to write instructions that work without your real-time guidance.
Challenge Exercises
These exercises combine multiple skills from this learning path. Each one simulates a realistic work scenario that requires conversational prompting, writing, analysis, and critical evaluation together.
Challenge 1: The Content Pipeline
Scenario: You need to produce a weekly newsletter for your team or audience on a topic you choose.
Task: Use a single ChatGPT conversation to:
- Brainstorm 5 topic ideas for this week's edition based on a theme you provide
- Select one and ask ChatGPT to create an outline with three sections
- Draft each section, providing feedback and requesting revisions after each one
- Ask ChatGPT to write a compelling subject line and a 2-sentence preview for email
- Request a final review: "Read the complete newsletter and suggest three improvements"
Deliverable: A complete, polished newsletter ready to send.
Success criteria: The final product should require no more than 5 minutes of your own editing before it is ready to use. The tone and style should be consistent throughout.
Challenge 2: The Strategy Session
Scenario: Your team needs to decide whether to build a feature in-house or buy a third-party solution.
Task:
- Describe the feature requirements and constraints to ChatGPT in detail
- Ask it to generate a build-vs-buy analysis framework tailored to your situation
- For the "build" option, ask for a rough timeline, resource requirements, and risks
- For the "buy" option, ask ChatGPT to identify evaluation criteria and potential vendor questions
- Ask ChatGPT to argue forcefully for "build," then argue forcefully for "buy"
- Finally, ask it to identify the single most important factor that should drive the decision
Deliverable: A one-page decision brief you could present to your team.
Success criteria: The brief should be balanced, specific to your actual situation (not generic), and should make the decision factors clear without making the decision for you.
Challenge 3: The Learning Accelerator
Scenario: You need to learn a new skill or topic quickly for a project starting next week.
Task:
- Tell ChatGPT the topic, your current knowledge level, and your deadline
- Ask it to design a 5-day learning plan with specific daily objectives
- For Day 1, ask ChatGPT to teach you the foundational concepts, then quiz you on them
- Based on your quiz answers, ask it to identify your knowledge gaps and adjust the plan
- Ask it to generate a "cheat sheet" — the 20% of knowledge that covers 80% of practical use
Deliverable: A personalised learning plan and a reference cheat sheet.
Success criteria: After completing Day 1 with ChatGPT as your tutor, you should feel measurably more confident in the topic than when you started. The cheat sheet should be something you would actually pin to your desk.
Quick Reference
Prompting Patterns That Work
- Role assignment: "Act as [specific role] and help me with [task]."
- Constraint setting: "For this conversation, always [rule 1], [rule 2], and [rule 3]."
- Iterative refinement: Start broad, then steer with "make it more [X]" or "less [Y]."
- Format specification: "Structure this as [format] with [specific structural elements]."
- Self-critique: "Now challenge your own answer. What are the weaknesses?"
- Audience targeting: "Write this for [specific audience] who [specific context about them]."
- Example anchoring: "Match the style and tone of [specific publication, author, or example]."
ChatGPT's Strengths
Last verified: March 2026
- Versatile general-purpose assistant across writing, analysis, coding, and creative tasks
- Strong conversational memory within a thread — effective for iterative work
- Excellent at tone control and writing for specific audiences
- Large ecosystem of Custom GPTs for specialised tasks
- Built-in image generation (GPT Image), web browsing, and Advanced Voice (free tier: 15 min/month preview; paid plans: higher daily limits)
- File uploads and basic analysis on all plans (free: 3 files/day); full Advanced Data Analysis on paid plans
- Native app connectors (Gmail, Google Drive, Slack, Notion, GitHub, and 20+ others) on Plus and above
- Available on web, iOS, Android, and desktop apps
📈 ChatGPT Capability Strength by Task Type
Source: AI Tutorium internal testing, March 2026
ChatGPT's Limitations
Last verified: March 2026
- Free and Go tiers display ads and have tighter usage caps
- Can produce confident-sounding but incorrect information (hallucination)
- Knowledge has a training cutoff — without browsing, it cannot answer about very recent events
- Long conversations may lose coherence as context fills up
- Memory is not available on the free tier; native app connectors and Agent Mode (formerly Operator) require Plus or higher
⚖️ Free vs Go vs Plus
Tier Features Price ChatGPT Free GPT-5.3 (10 messages/5 hrs, then falls back to lighter model), includes ads, no memory $0/month ChatGPT Go GPT-5.3 with 10x more messages, memory, custom GPTs, includes ads $8/month ChatGPT Plus GPT-5.3 default + GPT-5.4 Thinking via model picker, Agent Mode, app connectors, no ads $20/month
Before-You-Send Checklist
- Did you specify the audience, tone, and format you want?
- Did you provide enough context for ChatGPT to give a specific (not generic) answer?
- Are you planning to refine the response, or expecting perfection on the first try?
- For factual claims — will you verify them before using them?
- Is this something ChatGPT is good at, or would a different tool serve you better?
You've got the foundations — the rest is practice. Every conversation teaches you something new, and the gap between your first prompt and your hundredth will surprise you.
Practice Project
If you've made it this far, you already know ChatGPT can do impressive things with a single prompt. But the real shift happens when you build something you'll actually reuse — and that's what we're doing here.
Your Personal AI Toolkit
Time: 45-60 minutes
What you'll build: A set of 5 reusable prompt templates for your top weekly tasks — each with a clear scenario, a tested prompt, and notes on what to adjust for different situations.
Why this matters: Most people start from scratch every time they open ChatGPT. That means re-explaining context, re-specifying format, and getting inconsistent results. A personal toolkit eliminates that friction. We've seen people cut their prompt-writing time by 70% just by having 5 good templates ready to go.
Steps
- Identify your top 5 repeating tasks. Open your calendar or task list from last week. Pick 5 things you did more than once — writing emails, summarising meetings, drafting social posts, creating agendas, reviewing documents. Write each one down with a sentence describing the typical scenario (e.g., "Every Monday I write a project status update for my manager covering what shipped, what's blocked, and what's next").
- Draft a prompt template for each task. Open a new ChatGPT conversation for each one. Write a prompt that includes: the role you want ChatGPT to play, the specific context it needs, the format you want the output in, and the tone or audience. Test it with real data from last week. If the output isn't quite right, refine — add constraints, specify what to include or exclude, adjust the length.
- Document what to customise. For each template, mark the parts that change every time (the variables) versus the parts that stay fixed (the structure). For example, your status update template might always use the same format but swap in different project names and accomplishments each week. Add a note about common adjustments — "For executive audience, cut to 3 bullet points. For team audience, include technical details."
- Test each template twice more. Run every template with a different scenario to check it holds up. Does the email template work for both good news and difficult conversations? Does the summary template handle short meetings and long ones? Fix any that break under variation. A good template works 80% of the time without changes.
Deliverable: A document (Google Doc, Notion page, or even a note on your phone) with 5 prompt templates, each including: the scenario it's for, the full prompt text with variables marked in [brackets], and 1-2 notes on when to adjust it.
Stretch goal: Save your best template as a Custom GPT (Plus plan) or pin it as a conversation starter so it's one click away next time.
Reflection: Which template surprised you most — the one that worked better than expected, or the one that needed more iteration than you thought? That gap between expectation and reality is where the real learning happens.
Building this toolkit isn't just an exercise — it's something you'll actually use on Monday morning. And once you've felt the difference between starting from scratch and starting from a tested template, you won't go back.