Complete AI Mastery Course

{"contentFormat":"slides.v1","completion":{"requireAllSlides":true,"requireQuiz":true},"slides":[{"kind":"title","eyebrow":"Module 2: Practical Prompt Engineering","title":"Using Personas and Context Effectively","body":"Two prompts can ask for the exact same thing and get wildly different answers — because one told the model who to be and what situation it's in, and the other didn't. Master these two techniques to give your AI outputs professional relevance and precision.","outcomes":["Assign a persona to shape the model's tone, vocabulary, and depth","Supply rich context to tailor answers to your audience, industry, and region","Combine persona and context for expert and relevant outputs","Set permanent persona and context in system prompts or custom instructions"],"narration":"Welcome to Module 2. In this lesson, you'll learn how to use personas and context to get AI outputs that are not just correct, but perfectly tailored to your situation."},{"kind":"content","heading":"Personas: Telling the Model Who to Be","body":"A persona is a role you assign the model at the start of a prompt — for example, \"You are a patient primary-school maths teacher\" or \"You are a cautious legal reviewer.\" Because the model has learned from vast amounts of text written by every kind of professional, naming a role shifts its vocabulary, depth, tone, and assumptions to match.\n\nAnalogy: It's like calling the right specialist. You wouldn't ask a poet to audit your accounts or an accountant to write your wedding speech. A persona picks the right voice from the model's enormous range.\n\nHow to write an effective persona:\n- Name the expertise and its depth. \"You are a supply-chain analyst with 15 years in fast-moving consumer goods.\"\n- Add relevant traits. Patient, concise, sceptical, encouraging — whatever the task needs.\n- State the audience the persona is serving. \"…explaining to a small-business owner with no finance background.\"\n- Set behavioural rules. \"Ask a clarifying question if key details are missing. Flag any assumption you make.\"\n\nA persona shapes how the model responds; it does not grant new facts or real credentials. A model told \"you are a doctor\" is still just a language model — treat its output as informed drafting, never as licensed professional advice.","callout":{"variant":"insight","title":"Persona ≠ Credential","text":"A persona shapes tone and vocabulary, not factual accuracy. Always verify critical information from authoritative sources."},"narration":"A persona tells the model what role to adopt. This changes the style and depth of the response, but doesn't give the model real expertise. Always treat outputs as drafts."},{"kind":"content","heading":"Context: Giving the Model the Facts It Needs","body":"Context is the specific background information the model needs to tailor its answer to your actual situation: your goal, your audience, your constraints, your data, and your region. Without it, the model fills gaps with generic assumptions — often defaulting to a Western, English-speaking, single-currency worldview that may not fit you at all.\n\nKinds of context worth supplying:\n- Situation and goal — what you're doing and why. \"I'm pitching to a micro-lender for working capital.\"\n- Audience — who reads the output, their level, their language comfort.\n- Source material — documents, data, or notes the model should use (wrap these in delimiters like \"\"\" so they aren't mistaken for instructions).\n- Local and cultural specifics — country, currency (₦, ₹, R$, €, KES…), regulations, customs, and units. Stating these prevents wrong defaults.\n- Constraints — length, banned words, must-keep facts, formatting.\n\nAnalogy: Persona is who's talking; context is what they know about your situation. A brilliant consultant who knows nothing about your business still gives generic advice. Persona and context together produce answers that are both expert and relevant.","table":{"headers":["Context Type","Example","Why It Matters"],"rows":[["Situation & Goal","\"I'm pitching to a micro-lender for working capital.\"","Aligns output with your objective"],["Audience","\"Explain to a small-business owner with no finance background.\"","Adjusts complexity and tone"],["Source Material","\"\"\"Use this sales data\"\"\"","Grounds answer in your data"],["Local Specifics","\"Country: Kenya; Currency: KES\"","Prevents wrong defaults"]]},"narration":"Context provides the specific facts the model needs to tailor its answer. Without it, the model makes generic assumptions that may not fit your situation."},{"kind":"content","heading":"Persona + Context Together","body":"The strongest prompts combine both. Notice how role, audience, and local context stack up:\n\n

\nYou are an experienced small-business mentor who explains simply and never uses jargon.\nContext: I run a tailoring shop in Accra, Ghana. I have 3 staff and want to accept mobile-money payments. My customers mostly pay cash today. Budget is limited; amounts are in Ghana cedis (GHS).\nTask: Give me a simple 5-step plan to start accepting mobile-money payments this month.\nFormat: Numbered steps, each with one action and one thing to watch out for. Under 200 words.\n

\n\nSystem prompts and custom instructions: Many tools let you set a persona and standing context once so you don't repeat it. In ChatGPT this is Custom Instructions (and Projects); Claude and Gemini offer similar settings, and via API these live in the system prompt. Set your role, audience, language preferences, and region there, and every chat inherits them automatically — a huge time-saver.","callout":{"variant":"tip","title":"Set It Once","text":"Use Custom Instructions or system prompts to store your persona and context permanently. Every new chat will start pre-tailored."},"narration":"Combine persona and context for the best results. You can also set them permanently in system prompts or custom instructions to save time."},{"kind":"content","heading":"Real-World Examples","body":"1. HR officer, Nairobi (Kenya). Wanjiru needs interview questions that respect local employment norms. Persona + context — \"You are an HR specialist familiar with Kenyan hiring practices; we're recruiting a junior accountant; keep questions fair and non-discriminatory\" — yields questions far more appropriate than a generic set.\n\n2. Software freelancer, Kyiv (Ukraine). Andriy asks the model to \"act as a senior code reviewer who explains issues kindly to a junior developer\" and supplies his code between triple quotes plus the context \"this runs on a low-memory server.\" The persona sets the tone; the context makes the review actionable.\n\n3. Community pharmacist, Lahore (Pakistan). Bilal drafts patient leaflets. He sets a persona — \"a health educator writing at a basic reading level\" — and context — \"readers may have limited literacy; keep instructions in simple English; do not give dosages, refer to a doctor.\" The behavioural rule keeps the output safely within bounds.","cards":[{"title":"Nairobi HR","text":"Persona: HR specialist familiar with Kenyan hiring. Context: recruiting junior accountant. Output: fair, non-discriminatory questions."},{"title":"Kyiv Developer","text":"Persona: senior code reviewer, kind to juniors. Context: code on low-memory server. Output: actionable, tone-appropriate review."},{"title":"Lahore Pharmacist","text":"Persona: health educator, basic reading level. Context: limited literacy, no dosages. Output: safe, simple patient leaflet."}],"narration":"Here are three real-world examples from different regions showing how persona and context transform generic outputs into tailored, useful responses."},{"kind":"quiz","heading":"Check Your Understanding","questions":[{"question":"What does a persona primarily affect in the model's output?","options":["The factual accuracy of the response","The tone, vocabulary, and depth of the response","The speed at which the model generates text","The length of the response"],"questionId":"cmrf73k4t000rpd27lpo2bwv0"},{"question":"Why is it important to supply local context (e.g., country, currency) in your prompt?","options":["It makes the model faster","It prevents the model from using generic, often Western-centric defaults","It is required by all AI tools","It increases the model's knowledge base"],"questionId":"cmrf73k4t000spd2710vw0eb0"},{"question":"Which of the following is the best way to avoid repeating persona and context in every chat?","options":["Copy and paste the same prompt each time","Use Custom Instructions or system prompts to set them permanently","Ask the model to remember your preferences","Write a longer prompt every time"],"questionId":"cmrf73k4t000tpd27tog3nzxq"}],"narration":"Let's check your understanding with a quick quiz. Choose the best answer for each question.","quizId":"cmk7lgiiq000dg4p8anh40q16"},{"kind":"summary","heading":"Key Takeaways","takeaways":["Assign a persona to shape the model's tone, vocabulary, and depth — but remember it does not grant real credentials.","Supply rich context (goal, audience, local specifics, constraints) to avoid generic, Western-centric defaults.","Combine persona and context for outputs that are both expert and relevant to your situation.","Use Custom Instructions or system prompts to set persona and context permanently for efficiency.","Always verify critical information from authoritative sources, especially when using professional personas."],"narration":"To summarize: personas shape how the model responds, context provides the specifics it needs, and combining both gives you tailored, professional outputs. Set them once in system prompts for efficiency, and always verify critical facts."}]}