The next 90 minutes will change the way you see every tool you have ever used. We will move from curiosity to real, visible wins with conversational AI — together, live.

• Take 60 seconds. Think — do not type yet.
• Share your answer in the chat in one sentence.
• We will read 5 out loud and map them to AI tools we will cover today.
• Notice: almost every answer is already something AI can do in 2026.

Generative AI is software that creates new content — text, images, audio, video, code — that did not exist before, based on patterns it learned from massive amounts of data.

• Data — billions of pages of text, images, code, audio
• A model — a giant mathematical brain (the LLM) that finds patterns
• Training — months of running data through the model on supercomputers
• A prompt — YOUR instruction that tells it what to create

All Gen AI does one thing: generate. But what it generates puts it in a category.

• Summarise a 60-page chapter into a 1-page revision sheet
• Get a topic explained 5 different ways until it clicks
• Generate practice questions and instant grading on your answers
• Draft, structure, and proof essays (without it writing them for you)

• Inbox triage: reply to 30 emails in 10 minutes
• Turn a 90-minute meeting recording into action items + a follow-up email
• Build slide decks, reports, and proposals from rough notes
• Research a new client, market, or competitor in 5 minutes instead of 5 hours

• Generate full marketing campaigns (copy, images, video) for the cost of a coffee
• Create product mockups before paying a designer
• Build internal tools and automations without hiring developers
• Run 24/7 customer support with branded AI agents

You don't need a PhD — but understanding the pipeline below will make you 10× better at using AI.

An LLM is the brain inside every text-based AI tool you've heard of. "Large" because it has hundreds of billions of parameters (the dials that store what it learned).

Same underlying tech (LLMs) — very different strengths. Knowing which to reach for is half the skill.

A prompt is the instruction you give the AI. Prompt engineering is the skill of writing instructions that make the AI produce exactly what you want — on the first or second try.

The difference between these two prompts is the difference between a junior intern and a senior strategist responding to you.

Drawn from Anthropic's official prompt engineering guide and battle-tested with ChatGPT, Gemini, and DeepSeek.

One formula. Memorise it. Use it for the rest of the week.

12 high-leverage prompt techniques. Export to Excel or PDF and keep it open while you work.

• Explain what Gen AI is, how an LLM works, and why prompts matter
• Pick the right tool (ChatGPT / Gemini / Perplexity / DeepSeek) for the job
• Write a CRISP prompt instead of a vague one
• Export and reuse a 12-technique cheat sheet

• Refine your professional bio using ChatGPT — iterate at least 3 times
• Research one trending topic in your industry on Perplexity, compare with Gemini
• Ask DeepSeek to generate a small automation script and explain it line by line
• Add 3 of your own prompts to your personal cheat sheet

Yesterday we talked. Today we build. This session is 90% live demo, 10% talking. You will watch 5 tools turn an idea into a real deliverable — and you will leave with assignments to repeat each one yourself.

Before we touch a new tool — let's hear from you. Open discussion, 15 minutes.

• Which of the 4 conversational tools (ChatGPT, Gemini, Perplexity, DeepSeek) did you actually use after class?
• Show one prompt where the CRISP formula made a visible difference.
• Where did the AI disappoint you? What did you try to fix it?
• What's one task at work or school you now refuse to do without AI?
• Did anyone export and actually use the prompt cheat sheet? How?

1. Have the tool open in a second tab — follow along, don't just watch.
2. Drop your output in the chat after each demo.
3. Questions in the chat as we go — we batch-answer at the end of each tool.

Yesterday's tools thought with you. Today's tools produce with you. By the end of class you'll have touched a tool for knowledge, writing, marketing, presenting, and designing.

“Your second brain — now with a thinking partner inside every page.”

• Summarising long notes, meeting transcripts, and research
• Turning bullet points into polished prose
• Auto-generating action items, tags, and next steps
• Asking questions across your entire workspace

1. Open a blank page → type /ai → 'Brainstorm 10 content ideas for a fintech newsletter'
2. Paste a long meeting transcript → highlight → Ask AI → 'Summarise + extract action items by owner'
3. Inside a database → AI Autofill: auto-generate 'Summary' and 'Priority' columns
4. Q&A: 'What did we decide about pricing last week?' (searches whole workspace)

“Make AI-written content sound unmistakably human — and pass AI detectors.”

• Bypassing GPTZero, Originality.ai, Turnitin AI detection
• Rewriting stiff AI prose into natural, conversational tone
• Preserving meaning while changing rhythm and word choice

1. Generate a paragraph in ChatGPT → paste into WriteHuman
2. Choose 'Enhanced' mode → click Humanize
3. Compare before/after → run through an AI detector to verify
4. Iterate: feed the humanised text back if you want a different tone

“An end-to-end AI writing studio — articles, ads, emails, chatbots, in one place.”

• Long-form SEO articles with one-click research
• Ad copy variations (Google, Facebook, LinkedIn) at scale
• Cold email sequences and landing page copy
• Building branded AI chatbots (Botsonic) without code

1. Use 'AI Article Writer 6' → topic + keywords + tone → full draft in ~60s
2. Switch to 'Cold Email' template → paste prospect details → 3 variants
3. Open Chatsonic for live web-aware chat (alternative to ChatGPT for fresh data)
4. Botsonic: upload a PDF/URL → instant chatbot you can embed on a site

“From a one-line idea to a beautiful presentation, document, or webpage — in 30 seconds.”

• AI-generated slide decks with on-brand layouts
• Converting docs/notes into presentations
• One-click theming, export to PDF/PPT, share as a public webpage
• Embedding video, charts, and interactive content with no design skill

1. New → Generate → 'A 10-slide pitch for an AI-powered career coach for African graduates'
2. Pick a theme → Gamma drafts the entire deck with images and structure
3. Edit any card with natural language: 'Make this slide a comparison table'
4. Share as link, export to PDF, or present in full-screen

“Describe a UI in plain English — get production-ready designs and front-end code.”

• Generating multi-screen web and mobile UI from a prompt
• Iterating on designs with natural language ('make the CTA bolder, add dark mode')
• Exporting to Figma or copy-pasting front-end code straight into your project
• Rapidly prototyping before committing engineering time

1. Open Stitch → New project → describe your app
2. Demo prompt: 'A web app called CareerCompass for young professionals to set 90-day skill goals, log weekly progress, and get AI-generated action plans. 4 screens: Onboarding, Dashboard, Skill Tracker, Weekly Plan.'
3. Stitch generates screens → refine: 'Use a navy + gold palette, rounded cards, mobile-first'
4. Export to Figma or grab the HTML/Tailwind code → drop into Lovable / your IDE

One assignment per tool. Pick at least 3 of 5 to submit before Day 3. Export this list and keep it close.

Chain at least 3 of today's tools together on one project. Example: WriteSonic drafts the article → WriteHuman humanises it → Gamma turns it into a deck → Notion stores it → Stitch designs the landing page.

• Saw 5 production-grade AI tools used live, end-to-end.
• Understood when to reach for which tool — and who in your life it serves.
• Walked away with 5 assignments you can actually ship this week.
• Designed the bones of a real web app (CareerCompass) in front of your eyes.

We move from text to image, audio and video generation. Bring one real project where you need visuals or a voiceover — we'll generate them live.

Stop learning. Start shipping. By the end of this 90 minutes you'll have two working web apps— designed, specced and built live with you watching every keystroke.

Vibe coding is building software by describing what you want in plain language and letting AI generate the design, the code, and the infrastructure. You direct. The AI implements.

Every app today follows the same three steps. Same tools. Same order. Once you internalise this, you can build anything.

• UI first forces concrete decisions: what does the user actually see?
• PRD second turns vibes into specs, so the build is predictable.
• Build last is the cheap step now — don't waste it on a fuzzy idea.

A PRD is a Product Requirements Document. It's the bridge between an idea in your head and an app on a screen. Claude is the best model for this — long context, structured output, opinionated.

You are a senior product manager writing a PRD for a vibe-coding platform (Lovable) to implement.

Product name: <NAME>
One-line pitch: <PITCH>
Target user: <WHO>
Core problem: <PROBLEM>

Write a PRD with the following sections, in this exact order:

1. Problem & Audience (3-5 sentences)
2. Goals & Non-Goals (bullet list)
3. Core User Stories (5-7, in "As a __, I want __, so that __" format)
4. Screens & Navigation (list each screen + its purpose)
5. Data Model (tables + key fields, no SQL)
6. Key Interactions / Flows (numbered steps for the 2 most important flows)
7. AI Features (where AI is used, what prompt, what output)
8. Visual Style (colour palette, font, vibe in 2-3 sentences)
9. Out-of-Scope for v1 (bullets)
10. Acceptance Criteria (how we know v1 is done)

Keep it under 2 pages. Be opinionated. Choose, don't ask.

• Be opinionated in the inputs — vague in = vague out.
• Always paste a screenshot of your Stitch UI alongside the prompt.
• If Claude asks clarifying questions, treat each one as a real product decision.
• Edit the PRD ruthlessly before sending to Lovable — every extra sentence is a feature you'll have to maintain.

“A 90-day skills development tracker for young professionals.”

1. Onboarding (pick role + goal)
2. Dashboard (progress + AI weekly plan)
3. Skill Tracker (add skills, log practice)
4. Weekly Plan (AI-generated actions)

• Set 3 skill goals with 90-day target levels
• Log practice hours per skill — weekly streak
• AI-generated Monday action plan based on the past week
• Progress dashboard with a simple radar chart

Build CareerCompass, a 90-day skills development tracker for young professionals.

Audience: graduates and professionals with 0-5 years experience.

Screens:
1. Onboarding — user picks current role, target role, and 3 skills to grow.
2. Dashboard — shows overall progress, current streak, and this week's AI-generated action plan.
3. Skill Tracker — list of the 3 skills with: target level, current level, practice hours log, and an "Add session" button.
4. Weekly Plan — AI-generated list of 5 concrete actions for the week, regenerated every Monday or on demand.

Data model:
- profile(id, name, current_role, target_role)
- skill(id, profile_id, name, target_level 1-5, current_level 1-5)
- session(id, skill_id, hours, note, logged_at)
- weekly_plan(id, profile_id, week_start, actions jsonb)

AI feature:
- Generate the weekly plan using the Lovable AI Gateway. Prompt the model with the user's role, target role, skills, and last week's sessions. Return 5 actions with a one-line "why this".

Visual style:
- Navy #0A1F5C primary, gold #E8920A accents, white background, rounded cards, generous spacing.
- Sans-serif (Inter). Mobile-first.

Use Lovable Cloud for auth (email magic link) and database. Out of scope for v1: teams, payments, mentor matching.

“An automated productivity tracker that learns your work rhythm.”

1. Today (timer + task list)
2. Automations (rules)
3. Insights (charts)
4. Settings (integrations)

• One-click task timer with AI auto-categorisation
• Rules engine: 'When I open Figma for 25 min, log it as Design'
• Daily AI summary of where your time went + a focus score
• Weekly insights: peak hours, biggest distractions, suggested deep-work blocks

Build WorkFlow, an automated productivity tracker for knowledge workers.

Audience: freelancers, remote workers, and small teams who hate manual time logging.

Screens:
1. Today — a running timer, list of today's tasks, quick "Start task" input.
2. Automations — user-defined rules: triggers (keyword, app, time of day) → actions (auto-log, tag, notify).
3. Insights — charts of time by category, focus score, peak hours.
4. Settings — profile, categories, integrations placeholder.

Data model:
- profile(id, name, focus_target_minutes)
- task(id, profile_id, title, category, started_at, ended_at, source)
- automation(id, profile_id, trigger jsonb, action jsonb, active boolean)
- daily_summary(id, profile_id, day, summary_text, focus_score)

AI features:
- Auto-categorise new tasks based on title using the Lovable AI Gateway.
- Generate a daily summary at end of day: where time went, focus score 0-100, and one suggestion for tomorrow.

Visual style:
- Dark mode first. Navy #0A1F5C background, gold #E8920A accents, red #D72638 for "stop / overrun".
- Inter font, generous spacing, big numeric displays for the timer.

Use Lovable Cloud for auth and database. Out of scope for v1: real OS-level tracking, team dashboards, billing.

A built app nobody hears about doesn't exist. We close the day by turning your build into a 7-slide launch deck in Gamma — under 5 minutes.

1. Title — App name + one-line pitch
2. Problem — Who hurts, and how badly
3. The user — Audience in one sentence + a face
4. Demo — 3 screenshots from your live Lovable app
5. How it works — The 3 key features
6. Built with AI — Stitch + Claude + Lovable in X hours
7. What's next + the ask — Try it, feedback, collaborate

Create a 7-slide launch deck for [APP NAME] — [ONE-LINE PITCH].
Audience: [WHO].
Use the 7-slide structure: Title, Problem, User, Demo, How it works, Built with AI (Stitch + Claude + Lovable), What's next + Ask.
Tone: confident, founder-energy, no fluff. Visual style: navy + gold + red accents, clean, lots of whitespace.

Today you watched two apps get built. Now you ship. Pick at least 2 of the 5 assignments below before Day 4.

Today is almost entirely hands-on. We'll generate hyper-real images, do AI product photography, build infographics, and finish with a Gamma presentation generated live from a single prompt.

1. Hyper-realistic image prompting — the 7-part formula
2. AI product photography — turn any product into a studio shot
3. AI infographics — text → one-page visual explainer
4. Gamma App — prompt → full presentation, edited live

A 2024 LinkedIn study: posts with custom images get 2× the engagement of text-only posts. Pitch decks with original visuals close 40% faster. People judge your work in 50 milliseconds — before they read a word.

Every photoreal prompt that works follows the same skeleton. Memorise the order — your images get 5× better the moment you do.

Shot on a Sony A7R V with an 85mm f/1.4 lens at f/2.0, ISO 200, 1/200s.

Subject: <WHO / WHAT — be specific: age, build, clothing, expression, action>
Setting: <WHERE — be specific: location, time of day, season>
Lighting: <KEY light source + secondary + mood — e.g. soft window light from camera left, warm rim light, golden hour>
Composition: <framing — e.g. medium close-up, rule of thirds, shallow depth of field, subject on the left third>
Mood / colour palette: <3-4 adjectives + a colour story>
Material & texture detail: <skin pores, fabric weave, surface reflections — what should look real>
Style reference: editorial photography, National Geographic, Vogue Business — pick ONE.

Negative: no plastic skin, no extra fingers, no warped text, no over-saturation, no AI-art look.

We'll run the SAME 7-part prompt through ChatGPT and Gemini side-by-side and compare what each one gets right and wrong.

You don't need a softbox, a backdrop, or a photographer. You need one decent phone photo of the productand a clean prompt. We'll do this live with a real product on my desk.

Studio product photograph of <PRODUCT NAME>, <ANGLE — three-quarter / top-down / eye-level>.

Background: <seamless white / textured concrete / matching brand colour>
Lighting: large softbox from upper-left, subtle fill from the right, soft shadow under product.
Lens: 100mm macro, f/8 for full sharpness, ISO 100.
Surface: <wood, marble, brushed metal, paper>
Mood: clean, premium, e-commerce hero shot.
Detail callouts: <logo legible, texture of material visible, droplets / steam / motion if relevant>.

Negative: no warped logo, no floating object, no double shadow, no plastic sheen.

In Gemini Nano Banana, upload your own product photo and prompt: "Keep the product exactly the same. Replace the background with [scene]. Match lighting and shadow direction." This preserves your real product and just changes the scene.

The goal: someone scrolls past, stops for 30 seconds, gets the point, screenshots it, shares it. Tools we'll use: Napkin AI (text → diagrams), Gamma (auto-layout), and ChatGPT image for the final polish.

1. Title — the question or claim
2. The 1 big number — the headline stat
3. 3 supporting stats — context
4. Simple flow / diagram — the "how"
5. One-line takeaway — what to do now

Create a single-page infographic explaining <TOPIC> to <AUDIENCE> in under 30 seconds.

Layout: vertical, mobile-friendly, 5 zones — title, the 1 big number, 3 supporting stats, a simple flow diagram, a one-line takeaway.
Style: flat vector, 2 brand colours + 1 accent, no clip-art, no stock-photo people.
Type: one strong sans-serif. Title 48pt, body 14pt.
Data: use ONLY the numbers I paste below. Do not invent any data.

<PASTE YOUR SOURCE TEXT / NUMBERS HERE>

I'll paste a 600-word article into Napkin AI and turn it into a shareable one-pager in under 4 minutes.

Gamma generates entire decks from a single sentence — then lets you restyle the whole thing with one click. We saw it briefly on Day 2; today we'll go deeper: themes, custom fonts, exporting to PowerPoint and PDF, embeds.

A 10-slide pitch deck for a Ghanaian fintech serving market traders. 
Audience: investors at a regional demo day. 
Tone: confident, warm, data-led. 
Include: problem, solution, traction, business model, team, ask. 
Brand: navy and gold, modern editorial feel.

The headline task: produce two hyper-realistic, professional images for your actual industry and use case. Then attempt the rest at your pace.

Final session. Today we move from photoreal images into real marketing assets — posters, product ads and corporate flyers — all generated with Gemini and explored inside Google AI Studio (Pro). We close with a course recap and the assessment.

1. Why Gemini for design (5 min)
2. Inside Google AI Studio — Pro account tour (15 min)
3. Poster design — live build (15 min)
4. Product ad — 3 ratios from one prompt (15 min)
5. Corporate flyer — live build (15 min)
6. Course recap + closing (15 min)
7. Assessment briefing (10 min)

Gemini 2.5 (Nano Banana for images) is exceptional at three things we need today: editing existing images, holding brand consistency across formats, and rendering legible text inside images — which is exactly what posters, ads and flyers demand.

aistudio.google.com is Google's free playground for Gemini's most powerful models. With a Pro / paid account you unlock higher rate limits, the 1M-token context, image generation, video (Veo) and audio (TTS) — all in one console.

• Higher daily quota — no mid-class throttling.
• Access to Veo (video) and the latest image models.
• Save prompts as reusable "apps" you can share with your team.
• Larger uploads — entire PDFs, long videos, audio files.

Posters fail when they try to say five things. We build with a clear hierarchy: title → hero image → details. Gemini's text-in-image makes this finally possible without dropping into Photoshop.

1. Write the brief in plain English (event, audience, mood, palette).
2. Use the template below — fill in the brackets.
3. Generate at 9:16 or 2:3, then ask Gemini to "tighten the headline".
4. Re-prompt with: "Same poster, swap the colour palette to ___" for variants.

Vertical A3 event poster, photoreal hero with bold typography overlay.

Event: <NAME — e.g. "AI for Business Summit 2026">
Date / Time: <DATE, TIME>
Venue: <PLACE>
Audience mood: <e.g. ambitious young professionals, executive, faith-based, academic>

Hero visual: <describe ONE strong image — e.g. confident speaker on stage, glowing stage lights, hands on a laptop with code reflection>
Lighting: dramatic key light, soft rim light, cinematic.
Colour palette: <2 brand colours + 1 accent>
Typography: bold sans-serif headline top-third, supporting details bottom-third, legible from 2 metres.

Layout: title 20% top, hero image 60% middle, details + CTA + logo 20% bottom.

Negative: no warped text, no fake logos, no extra limbs, no cluttered background.

Modern brands need the SAME ad in at least three places: Instagram feed (1:1), stories / TikTok (9:16), and YouTube / billboard (16:9). With Gemini we keep one master prompt and just swap the ratio + framing.

Premium product advertisement, <ASPECT — 1:1 / 9:16 / 16:9>.

Product: <NAME + 1-line description>
Brand mood: <e.g. luxurious, playful, minimalist, Afro-futurist>
Setting: <studio / lifestyle / abstract gradient>
Lighting: large softbox key, gentle fill, soft contact shadow.
Composition: product hero-centre, generous negative space for headline.
Headline area: top-left, 4-6 words max, sans-serif.
Colour palette: <2 brand colours + 1 accent>.

Include: subtle brand-coloured backdrop, micro-texture (dust, droplets, steam) for realism.

Negative: no warped logo, no double shadow, no plastic sheen, no extra products in frame.

In Gemini: upload your real product photo + the prompt with "Use the product in the uploaded image, do not redraw it." — the product stays real, only the scene is AI.

A corporate flyer must look credible, communicate fast, and print well. We'll build a one-pager that's appropriate for a fintech, a training company, or a consultancy — and we'll do it without opening InDesign.

One-page corporate flyer, A4 portrait, professional and modern.

Company: <NAME>
Offer / Service: <ONE LINE>
Audience: <WHO IT IS FOR>

Layout (top to bottom):
1. Logo + bold headline (8-10 words) over a hero image
2. 1-line subheadline explaining the value
3. 3 benefit blocks with icons (each: 3-word title + 1-line description)
4. Strong CTA button + URL
5. Footer: contact, email, phone, social handles

Style: clean white background, 2 brand colours + 1 accent, photo treatment matches brand.
Typography: one serif OR one sans-serif family, max 3 sizes.

Negative: no lorem ipsum, no fake testimonials, no clip-art icons.

• "Replace the 3 benefit icons with line-style icons in our accent colour."
• "Make the CTA button red and double its size."
• "Add a QR code placeholder bottom-right."

• You stop asking "can AI do this?" and start asking "which tool + which prompt?"
• You ship in hours what used to take weeks — decks, apps, ads, reports.
• Your output looks intentionally designed, not just "AI-generated".

Coming soon from KNUST AI: AI for Business Assistance — a deeper course for founders, operators, and teams who want AI to run the back office. Watch the class group for dates.

Your final assessment is a short open-book exam — 10 MCQs, 3 prompt-writing scenarios, and ONE practical project of your choice. Open the Assessment page from the Home Hub.

1. Event Poster in Gemini (Gemini 2.5 (Google AI Studio))
   Design a portrait A3 poster for a real or imagined event (conference, launch, church program, school open day). Include title, date, venue, 1 strong visual.
   Deliver: 1 PNG poster + the final prompt + 1 sentence on what you fixed between attempts.
2. Product Ad — 3 Formats (Gemini / Nano Banana)
   Pick one product. Generate the SAME ad in 3 formats: 1:1 Instagram, 9:16 status/story, 16:9 billboard. Keep brand colours consistent across all three.
   Deliver: 3 images + the master prompt + the 2 variations you used for the other ratios.
3. Corporate Flyer (Gemini / Canva + AI)
   Produce a one-page corporate flyer for a service (training program, consulting offer, fintech product). Must include headline, 3 benefits, CTA, contact, and an on-brand hero image.
   Deliver: PDF + the prompt + a 2-line note on which AI Studio settings you used.
4. Google AI Studio Exploration (aistudio.google.com)
   Open Google AI Studio, run the same image prompt at 3 different temperature / aspect-ratio settings. Note how output changes.
   Deliver: Screenshot of the 3 outputs + 3 bullets on what you observed.
5. Final Reflection (Any doc)
   In 200 words, describe ONE workflow in your job/life that you will now do with AI from next Monday. Name the tool, the prompt pattern, and the time saved.
   Deliver: Short doc, submitted in the class group.

Thank you for showing up for 5 days. Keep building. Keep shipping. — Kojo Buabeng, KNUST AI.