Miller introduces an AI opportunity funnel: a way of evaluating which problems in your organisation or product are genuine AI opportunities versus problems that would be better solved by clearer processes, better training, or simpler technology. An AI opportunity has specific characteristics. First, it involves decisions made at high volume — too many for humans to handle one at a time with consistency. Second, it involves pattern recognition — there is a learnable relationship between inputs and the correct output. Third, it improves with data — the more cases it sees, the better it can get. Problems that fail one of these tests are probably not AI problems. Applying AI to the wrong problems produces expensive, unreliable solutions. Applying it to the right ones produces compounding improvements.