LESSON 1 OF 6
Find where it actually pays
BY THE END OF THIS LESSON
Identify the three tasks in your business worth automating first.
Most AI projects fail on selection, not execution
Businesses pick the most visible task rather than the most suitable one, get a mediocre result, and conclude the technology is not ready. The technology was fine. The task was wrong.
The three filters
Frequency. Does it happen weekly or more? A monthly task will never repay the setup.
Tolerance. What happens if it is 85% right? If the answer is "a person edits it", good. If the answer is "we lose a customer" or "we breach something", stop.
Bottleneck. Is this actually slowing you down? Automating something that was never the constraint just moves the queue.
All three, or pick something else.
Where it usually does not pay
Final legal or financial sign-off. Anything with regulatory exposure and no review step. Genuinely novel strategic decisions. Work where the relationship is the product — a personal apology from a founder is worth something precisely because a person wrote it.
Being clear about this is what makes the rest credible internally.
WORKED EXAMPLE
VISIBLE BUT WRONG
"Let's use AI to write our quarterly strategy deck." — happens four times a year, high stakes, and the value was in the argument nobody has had yet.
BORING BUT RIGHT
"Let's use AI to turn support tickets into weekly product feedback." — happens every week, an 85%-right summary is genuinely useful, and right now nobody does it at all because it takes two hours.
The second passes all three filters. It is also unglamorous, which is why it gets overlooked — and why it delivers a result you can point at within a fortnight.
YOUR TURN
Audit your own week against the three filters.
RUN THIS
Help me find where AI would actually pay in my business. Interview me about a typical week — what my team does repeatedly, what gets dropped when we are busy, and what people complain about. Ask one question at a time. Then score each candidate task on frequency, tolerance for an 85%-right first draft, and whether it is genuinely a bottleneck. Give me a ranked table, and separately a list of tasks I should NOT automate, with the reason. My business: [describe it]
What a good result looks like
The "do not automate" list matters as much as the ranking. If it is empty, ask "what would you refuse to automate here, and why?"
KNOWLEDGE CHECK
Answer all 3 correctly to complete this lesson.
1. Why do most business AI projects fail?
2. A task is 85% right and a person reviews it before it goes out. What does that mean?
3. What happens if you automate something that was not the bottleneck?
REMEMBER THIS
Frequent, tolerant of an 85% draft, and genuinely a bottleneck — all three, or pick something else.