Most professionals work with a mental model of AI shaped by marketing hype and science fiction. Over the next 20 minutes, you'll build a working model that stands up to reality — and learn the one structural fact that prevents most AI mistakes.
If you don't know what AI actually is, you'll either over-trust it (and ship its mistakes to clients) or under-use it (and miss real productivity gains). Both are expensive. This module gives you the foundation to do neither.
When you type a question into a chatbot, the system isn't understanding you. It's looking at your input as a sequence of patterns, then predicting — based on billions of examples it was trained on — what the most likely next word should be. Then the next. Then the next.
It's an extraordinarily sophisticated autocomplete. The output often feels like reasoning because human language is full of patterns that look like reasoning. But the mechanism underneath is statistical prediction.
This isn't a limitation to overcome. It's the architecture. And understanding it tells you exactly where AI works brilliantly (pattern-rich tasks) and where it doesn't (genuinely novel judgement calls).
Most workplace tools blend all three, but knowing which one is doing what tells you what to expect — and what to check. Tap each card to expand.
Creates new content — text, images, code, summaries — by predicting what should come next.
→ ChatGPT writing your draft
→ Image generators
→ Code completion tools
→ Meeting summary generators
Forecasts outcomes by finding patterns in historical data — risk scores, demand, churn likelihood.
→ CRM churn predictions
→ Credit risk scoring
→ Demand forecasting
→ Fraud detection alerts
Follows defined rules to handle repetitive, structured tasks — often with ML for edge cases.
→ Invoice matching
→ Email routing rules
→ Document classification
→ Workflow triggers
If a tool is generative, you should always verify its output. If it's predictive, you should check the model's confidence and the data it learned from. If it's automation, you should review the exceptions it flags. Different types, different responsibilities.
Below is a real-style AI response to a business question. Switch between the three views to see what's actually inside it: solid facts, reasonable guesses, and complete fabrications — all delivered in the same calm, authoritative voice.
Notice how impossible it is to tell from the raw text. The fabrications use the same vocabulary, the same sentence structure, the same authoritative tone as the verified facts. This is not a bug. It's the architecture. Pattern-matching produces plausibility — not accuracy.
Three quick questions. No grade — just a chance to make sure the concepts have landed before moving into the practical exercise.
Drag each scenario from the pool into the correct bin. You'll get instant feedback as you drop, with an explanation of why. Don't worry about being right first time — the goal is to internalise the distinctions.
Here's how you did. Tap any card to see the explanation again, or reset to try once more.
Reflection is what turns a module into a habit. In one or two sentences, capture one specific thing you'll do differently at work this week, based on what you've learned. This saves automatically and you can return to it anytime.
You now have the foundation. The next module builds directly on this — once you know what AI is, the next question is how to communicate with it effectively. That's where the real productivity unlock starts.