Definition
A street light switches on by itself every evening. Is it smart?
Automation is a machine carrying out fixed instructions without a person's help. Artificial intelligence goes further: the machine decides what to do in situations its instructions did not spell out, often by learning from examples.
In simple words
A washing machine runs the same programme every time you press the button. That is automation. A spam filter meets emails nobody has ever seen and still decides which ones are spam. That is AI. Automation repeats; AI adapts.
Why it matters
Many products are called 'AI' when they are only automation. Knowing the difference stops you being fooled by advertising, and helps you choose the right tool: a simple rule is cheaper and more predictable when the situation never changes.
Key terms
- Automation
- A machine or system carrying out fixed steps without a person doing them by hand.
- Fixed rule
- An instruction that always gives the same response to the same situation, such as "at 6 a.m., ring the alarm".
- Learned behaviour
- A response a machine has worked out from examples, not one a person wrote down directly.
- Intelligent automation
- Automation that uses AI for the parts that need judgement, such as reading a messy form, and fixed steps for the rest.
- Edge case
- An unusual situation at the limits of what a system was designed for.
Explained in detail
A short history of automation
Automation is much older than computers. In 1804 the Jacquard loom in France used cards with punched holes to control the pattern woven into cloth, so a complex design could be repeated without a weaver choosing every thread. Water clocks, windmills that turn to face the wind, and the thermostat in a fridge are all automation. The first industrial robot, Unimate, began lifting hot metal parts in a car factory in 1961. Each of these follows fixed steps. None of them learns.
Where automation stops and AI begins
The dividing line is what happens when a situation was not planned for. A traffic light on a timer turns green every 90 seconds, even if the road is empty. That is automation. A smart traffic system that watches the queues with cameras and gives more green time to the busier road is using AI, because it decides from what it observes and can deal with patterns of traffic that no one wrote down in advance.
| Task | Automation version | AI version |
|---|---|---|
| Traffic signal | Changes every 90 seconds | Gives more time to the longer queue |
| Moves mail from one sender to a folder | Recognises new scams it has never seen | |
| Factory check | Rejects items below a set weight | Spots scratches of any shape in a photo |
| Customer help | Menu: press 1 for balance | Understands a question typed in your own words |
Strengths and weaknesses of each
Automation is predictable. You can test every step, it costs little to run, and it does exactly the same thing every time. Its weakness is that it cannot cope with edge cases. AI is flexible and handles variety, but it needs data, it can make surprising mistakes, and it is harder to check. A sensible designer uses plain automation wherever fixed steps are enough, and adds AI only for the parts that truly need judgement.
Most real systems mix both
Think of a bank processing a loan application. Reading a scanned, handwritten form and judging the risk may use AI. Checking that every required field is filled, calculating the monthly instalment and sending the letter are fixed steps. This combination is often called intelligent automation. When you meet a smart system, it helps to ask which parts learn and which parts simply follow rules, because the two fail in different ways.
How it works
The one-question test
Ask: can it deal with a case nobody wrote an instruction for? If it only ever does what was decided in advance, it is automation. If it handles new cases by itself, it uses AI.
| Fixed automation | Artificial intelligence |
|---|---|
| A timer switches the light on at 6 pm | A camera learns when the room is in use and switches lights on and off |
| A lift goes to the floor you press | A lift system learns rush hours and waits where people will need it |
| An ATM gives cash when the PIN is right | A bank system spots an unusual payment and asks you to confirm |
Try it yourself
Fixed rule against a learner
The same messages go to both. The rule never changes, and the learner improves after each answer. Who ends up ahead?
Fixed rule or learned behaviour?
Decide whether each machine follows a fixed rule or has learned from examples.
A music app suggests songs you end up liking
A keyboard predicts your next word
A photo app finds every picture of your grandmother
A water tank pump stops when the tank is full
A microwave stops after 2 minutes
A traffic signal changes every 60 seconds
Common mistakes
- Calling everything automatic 'AI'. Automatic and intelligent are not the same.
- Thinking AI is always the better choice. Fixed rules are often better when nothing changes.
- Thinking AI has no instructions at all. People still decide the goal, the data and the limits.
Check yourself
Summary
- Automation follows fixed instructions; AI handles situations the instructions did not cover.
- Test: can it deal with a case nobody wrote a rule for?
- Both are useful. Choose fixed rules when nothing changes, and AI when things keep changing.
Finished reading?
Sources: Russell & Norvig, Artificial Intelligence: A Modern Approach