Why this matters
You have already used AI several times today without being asked to think about it. The useful question is not "is there AI in this?" but "is this the kind of job AI is actually good at?" Getting that wrong is how people end up trusting a system with a decision it should never have been given.
The idea
In daily life, four uses cover most of what you meet:
| Service | What the AI does |
|---|---|
| Recommendation system | Predicts preferences from past behaviour and suggests things |
| Voice assistant | Recognises a voice, understands the command, carries it out |
| Machine translation | Converts text between languages automatically |
| Face recognition | Detects and identifies faces in photographs |
In industry it shows up as image diagnosis and drug discovery in healthcare, harvest prediction and pest detection in agriculture, quality inspection and predictive maintenance in manufacturing, and route optimisation in logistics.
What AI is good at: finding and classifying patterns in complicated data like images and text, recognising and generating images, audio and text, and reasoning probabilistically from data.
What needs caution: ethical judgments, where it can reproduce discrimination; anything involving personal information and privacy; final decisions, because responsibility has to sit with a person; accuracy when the training data was biased; hallucination, where it states something untrue with full confidence; rights over copyrighted work used as training data; and the black-box problem, where nobody can say how the judgment was reached.
The line between the two lists is not about difficulty. AI handles genuinely hard pattern problems. It is about whether being confidently wrong is survivable.
Picture it
flowchart TD
A[A task] --> B{Is it pattern-finding<br/>in data?}
B -- yes --> C{Would a confident<br/>wrong answer cause harm?}
B -- no --> D[Not a good fit for AI]
C -- no --> E[AI can do this]
C -- yes --> F[AI assists, a person decides]Worked example
A hospital wants image-diagnosis AI to find disease in X-rays.
The fit is good on the first test: this is pattern recognition in images, exactly what the technology is strongest at, and it can flag faint things a tired human eye skips. So use it.
But the second test changes the shape of the job. A wrong call here harms a patient, the model may have learned from data that under-represented some groups, and if it cannot explain its reasoning nobody can check the result. Someone must also be answerable for the diagnosis, and that cannot be a piece of software.
So the AI reads every image and flags what looks wrong, and a doctor confirms the diagnosis. The AI widens what gets looked at. The person keeps the decision.
Your turn
Try the practice questions for this lesson. Match everyday services to the AI technology behind them, sort statements about AI's strengths into true and false, and pick out the task that needs human judgment.
Take it further
Pick one industry from this lesson other than agriculture: healthcare, manufacturing or logistics.
- Investigate. Find one real, documented example of AI used in that industry, and record your source.
- Weigh it up. Note one benefit and one caution, using what AI is good at and what needs care.
- Decide. Should that industry use AI for this task alone, or with a person checking the result? Give two reasons.
Ask the two questions from the worked example in order. Is it pattern work? And would a confident wrong answer cause harm?
Recap
- Everyday AI: recommendation systems, voice assistants, machine translation, face recognition.
- Industry uses it for diagnosis, harvest prediction, quality inspection and routing.
- AI is good at pattern finding, recognition and generation, and probabilistic prediction.
- Be careful with ethics, privacy, final decisions, biased data, hallucination and the black-box problem.