Lesson 181 · 2 min
Which jobs need AI and which do not
AI is the right tool when the input is messy and a good-enough answer is fine; for exact, repeatable rules, ordinary software is cheaper and safer.
Adding AI to everything is tempting. But a model costs money on every use, can be slow, and is sometimes wrong. A simple rule written in code is free to run and gives the same answer every time.
AI shines with messy input: free text from customers, photos, voice notes, documents in many formats. It is also good when there are many acceptable answers, like drafting a reply or suggesting product names.
Ordinary software wins when the rule is clear and the answer must be exact: calculating tax, checking a date is in the future, sorting orders by price, sending a reminder at 9am. Using AI for these adds cost and risk for no gain.
Many good products mix the two. AI reads the messy customer email and pulls out the order number. Plain code then looks up the order and calculates the refund. Ask of each step: is the input messy, and is an approximate answer acceptable?
Do it now
Sort the steps
- 1. Pick a process at your work, like handling an order or marking homework.
- 2. Write each step on its own line.
- 3. Mark each one AI, plain software or person, using the two questions from this lesson.
Here are the steps of a process I run: [LIST STEPS] For each step, tell me whether AI, simple software rules or a person is the best fit, and why, in one line each. Be honest when AI is not needed.
Quick check
Words from this lesson
- Algorithm
- A set of step-by-step instructions a computer follows to solve a problem or make a decision.
- Automation bias
- Our habit of trusting a machine's answer too readily, even when our own judgement or other evidence says otherwise.
- LLM
- An AI trained on huge amounts of text to predict the next word, which lets it write, summarise, translate and chat.