AI for Beginners: Use ChatGPT & Claude to 10× Your Daily Pro

Meet your instructor and get the honest 60-second summary of what this course covers, what it skips, and who'll get the most out of it. Quick, no-fluff orientation so you know exactly what you're signing up for.

Quick tour of the four AI assistants you'll actually meet in 2026: ChatGPT, Claude, Gemini, and Copilot. Understand what each does best, who they suit, and which to start with as a beginner.

Step-by-step walkthrough: create your free ChatGPT and Claude accounts in under ten minutes. Includes bookmarks and your first test message in each tool so you leave with everything working.

What actually happens when you send a message and get a reply - tokens, context windows, and why the AI 'forgets' long conversations. The mental model that makes everything else click.

Honest pricing comparison: free tiers vs paid tiers, what you actually get for the upgrade, and when it matters. Includes a strategic answer for which tool to pay for if you pay for one.

Five rules you should never break when using AI in real work, plus a simple framework for deciding when to trust the output and when to verify. Includes the 'say I don't know' pattern that cuts hallucinations.

Why most 'AI is bad' experiences are actually bad prompt experiences. Side-by-side example of a vague prompt versus a clear one producing dramatically different output - same AI, thirty extra seconds.

The five-part recipe for writing good prompts: Context, Length, Examples, Audience, Role. Original to this course. One framework, applicable to any new AI problem you face.

When to give the AI examples versus when to just ask. Three examples often beat a paragraph of description - here's why and when it works.

Get the AI to return exactly what you want: tables, bullet points, JSON, Markdown. Specific prompts for each format, plus the pitfalls of vague requests.

Asking the AI to 'act as a senior copywriter' or 'pretend you're a patient maths tutor' pulls consistent tone and perspective from its training. When it helps, when it doesn't.

The single most useful phrase in prompt engineering: 'think step by step.' How it gives the AI more thinking space and dramatically improves reasoning, math, planning, and debugging tasks.

Every chat has two layers: the visible user prompt and the hidden system prompt that shapes behavior. How to write your own custom instructions so you don't repeat yourself across every chat.

Telling the AI what NOT to do works surprisingly well - removing AI-isms, constraining tone, and preventing hallucination. Plus the 'say I don't know' pattern for research tasks.

The slider that controls how random the AI is. When to crank it up for brainstorming, when to crank it down for code, and the easy shortcut: just ask the AI to 'be more creative' or 'be more precise.'

Seven mistakes I see beginners make most - vague prompts, treating AI like a search engine, trusting the output blindly, and four more. Identifying which one is half the fix.

Apply CLEAR to the most common writing task at work. Templates for drafting new emails, replying to difficult ones, and a four-part structure for pushing back gracefully.

The five-step report workflow: summarise raw material, outline, draft each section, combine, edit. Cuts a typical three-page report from five hours to ninety minutes.

The single highest-ROI use of AI for most office workers. Paste your messy meeting notes, get a structured summary with key decisions, action items, and open issues in under five minutes.

How to outline, draft, edit, and review a 1,000+ word blog post with AI help. Section-by-section drafting, the editing pass, and the 50/50 rule that keeps your voice in the final draft.

Platform-specific templates for LinkedIn, Twitter/X, and Instagram. The '10 in 10' pattern for getting multiple variations at once, plus what AI is bad at (predicting virality, real-time trends).

Two related tasks AI handles well. The right translation prompt, the back-check for natural phrasing, and the proofreading prompt with explanations that help you learn from corrections.

How to make AI sound like you rather than like AI. The voice-extraction prompt, the iteration loop, and when you should deliberately NOT match your voice.

Use AI as your critical first editor. The strict-editor prompt, multi-pass editing (clarity, concision, tone, persuasion), and the 'worst thing' pattern that finds your biggest improvement opportunity.

Where AI helps with SEO content and where it falls short. The five-step SEO workflow, keyword brainstorming prompts, and why you still need to add the expertise AI can't fake.

Pull everything from Sections 1-3 into one reusable template. Build a personal writing-assistant prompt with your tone, length, and constraint preferences. Use it for every piece of writing from now on.

The four-step research workflow: Question, Search, Synthesise, Verify. Why most 'AI research' fails - and how this simple structure produces dramatically better output every time.

Why AI hallucinations are structural rather than random, the verification rules for anything that matters, and the 'say I don't know' prompt pattern that cuts hallucination dramatically.

The chunking problem with long PDFs and articles, the per-chunk summarisation prompt, and the multi-document synthesis workflow for comparing sources.

Side-by-side comparison prompts that produce usable matrices, the three-step comparison template, and where AI is bad at competitive research (live pricing, customer reviews, strategic positioning).

What AI does well with data (read messy text, summarise, suggest charts) and where it struggles (statistical significance, causation vs correlation, cleaning). The interpretation prompt pattern.

Using AI for buyer personas, market-sizing frameworks, and trend analysis. Treat AI's output as a research plan, not the answer - the actual data still has to come from real sources.

Where each tool is strongest for research - ChatGPT for web search and synthesis, Claude for long documents, Gemini for the Google ecosystem, Copilot for Microsoft 365. A two-tool research workflow.

Pick a real work question, run the four-step research workflow, produce a one-to-two-page research brief. Submit in the Q&A for spot-check and feedback.

Five multiple-choice questions on research workflows, fact-checking, summarising long documents, comparing AI tools, and the best use of AI for market research.

Describe what you need in plain English and get the spreadsheet formula. Covers VLOOKUP, INDEX-MATCH, IF, SUMIFS, array formulas, and the always-verify-with-sample-data rule.

Ask in English, get Python or SQL out. The translation pattern, common use cases, and where AI-generated code fails (novel algorithms, security-critical code, performance tuning).

What AI shines at (standardising text, categorising free-text, finding fuzzy duplicates) versus where it struggles (numerical cleaning, mixed date formats). Use OpenRefine or Python for the second.

Pick the right chart for the data and generate the matplotlib or seaborn code. Iterate visually: change colour, thickness, labels - each as a small fast prompt.

Generate formulas at scale with ARRAYFORMULA, batch operations, and Google Apps Script or VBA macros. The always-test-on-a-copy-first rule.

The four building blocks every dashboard needs and the five-step build process. Turns a multi-day project into a multi-hour one - but the judgement calls about what to show still belong to you.

Numbers without narrative don't change behaviour. The four-section report structure: headline, drivers, implications, recommendations. The AI workflow for each.

End-to-end exercise: take a sales CSV, clean it, compute three KPIs, generate two charts, write a 200-word report. Post screenshots and the report in the Q&A.

Five multiple-choice questions on spreadsheet formulas, code generation, data cleaning, chart selection, and report structure.

An honest take on when non-coders should use AI for code and when they shouldn't. The rule of thumb: if you'd be comfortable writing it yourself given an hour, AI is fine. Otherwise, the AI code is probably dangerous.

Describe a layout, get working code. The describe-generate-iterate pattern, useful constraints to specify (framework, accessibility, responsive), and tips for single-file prototypes.

File operations, web scraping, email automation - the three patterns you'll reuse most. Always include error handling in the prompt, and always test on a copy before running on real data.

"Show me all customers who placed orders in the last 30 days but haven't ordered in the last 90." Describe what you want, specify the schema and SQL dialect, and verify on a small dataset.

Use AI as a tireless reviewer. The three things to give it for debugging (the code, the error, what you expected vs what happened) and the patterns for review (bugs, security, refactor).

Build a real, working single-file web page from scratch using the describe-generate-iterate pattern. Three sections (header, about, contact), responsive, one interactive element.

Five multiple-choice questions on when non-coders should use AI for code, HTML/CSS workflow, automation safety, SQL prompting, and debugging inputs.

DALL-E, Midjourney, and Stable Diffusion - three tools, three strengths. Best aesthetic quality, best integration, best for local use. Specific prompts win.

Upload a photo, get a description, extract text. Best uses: OCR, document summarisation, visual Q&A, accessibility. Ask specific questions rather than 'describe this.'

Reading contracts, summarising reports, querying across documents. The single-doc workflow with four useful queries, plus multi-document comparison patterns.

Three categories: built into AI assistants, Whisper (open-source, runs locally), and specialised meeting tools (Otter, Fireflies). Reliable for everyday work now.

Useful today: summarising long videos, searching within videos, transcription-to-summary, captions. Still hype: text-to-video for production, realistic avatars. Use AI to analyse video, not yet to create it.

Pick one media task you do regularly, document the steps from input to output, identify where AI helps. Post your workflow and the time saved per week.

Five multiple-choice questions on text-to-image tools, image analysis prompting, PDF workflows, Whisper, and video generation capabilities.

A decision framework for picking between ChatGPT, Claude, Gemini, and Copilot. Four questions to ask before each task: where's the data, how long is it, how careful, who sees the result.

Save, organise, and version your best prompts. The 'if you've used it twice, save it' rule. Three storage options from simplest (Notes app) to most powerful (version-controlled text files).

ChatGPT memories, Claude Projects, and the case for exporting the best exchanges to your own archive. Review your library monthly - what's working, what to drop.

Multi-step processes and handoffs: research in ChatGPT, analysis in Claude, draft in one, format in another. Where humans must stay in the loop (judgment, sensitive content, verification, feedback).

When to build a custom assistant (recurring tasks, packaged knowledge, team consistency) and the three layers to build well: instructions, knowledge files, conversation starters.

Where AI lives in the apps you already use: writing, email, meetings, code, search. The three questions to ask before adding any new tool to your stack.

What gets stored by default (history, prompts, metadata, sometimes human review), what to always keep private (passwords, financial details, regulated data), and paid tiers with no-training guarantees.

Document a complete weekly workflow: 5 to 10 recurring AI tasks, the tool and prompt for each, the handoffs, the human checkpoints, and the hours saved per week.

Five multiple-choice questions on tool selection, prompt library discipline, workflow hand-offs, custom assistant design, and privacy considerations.

What agents actually are - a loop of plan, act, observe, repeat. When to reach for agents (multi-step, success criteria, verifiable output) and when to skip (one-off, simple). The realistic version versus the hype, with a clear-eyed look at token cost, infinite loops, and fragile verification.

When the API earns its keep: batch jobs, integrations, custom cost control, server-side automation. The API is just HTTP calls - endpoint, POST with JSON, bearer token auth, per-token cost. Plus the auth, key rotation, and spend-cap safety basics you want in place before you ship.

Web search bridges the model's knowledge cutoff. How it works (search + read + synthesise + cite), when to use it (current events, recent research), and why you still need to verify the citations. Includes a 30-second verification routine and the search-plus-API patterns that turn it into a pipeline.

Vision shines for reading text in images, describing visual content, chart analysis, visual Q&A, and OCR. Struggles with fine text, spatial relationships, counting, and low-quality images. Image prompting tips, the whiteboard-to-notes workflow, and vision-plus-API patterns for receipts, forms, and dashboards.

Voice mode for hands-free thinking, interviews, accessibility, and language practice. Practical tips: speak in short complete sentences, name the format up front, ask for a saveable summary. Where voice shines, where text is still better, and why accessibility is the unsung win.

Trends worth watching (multimodal, long context, tool use, cheaper tokens) versus hype to ignore (AGI timelines, single-job replacements, product-launch obsessions). The durable skills - clear writing, critical thinking, workflow design, comfort with ambiguity - will outlast every current AI product.

Five multiple-choice questions on agents, the API, web search, vision, voice mode, and what skills will travel across model changes.

The five traps most beginners fall into: relying on the first answer, never building a prompt library, treating AI like a search engine, ignoring privacy, and not measuring time saved.

From 'I tried it once' to 'I use it daily.' The 2-week starter plan, the two-minute daily check-in, and the four-week review cycle that turns AI from novelty into infrastructure.

Where to keep learning: official docs, practitioner newsletters, two communities worth joining, and the curated reading list for the next 90 days.

A short sign-off, an encouragement, and a reminder that the CLEAR framework will outlast every AI product on the market today.