Don't ask whether AI can replace people — ask which tasks it can take on

"Will this role be replaced by AI?" is a big question, and often the least actionable one. What a company faces isn't a job title but a chain of small, interlocking tasks full of exceptions: gathering data, cross-checking, following up, making judgments, drafting replies, routing approvals, keeping records. Treat the whole job as one lump and AI can barely help; break it into tasks and the openings start to surface.

Slice the work before you pick a tool

Once a task can state clearly what goes in, what the output should look like, and who is accountable for the result, it already meets the conditions to be designed. Then look at each step: which are merely moving and tidying data, which need the key points pulled from a wall of text, which must be judged against rules, and which involve commitments, pricing, or relationships. AI is best suited to take on the first two kinds first; the third needs boundaries; the last should usually stay in human hands.

The value of AI usually isn't removing a person wholesale, but lifting them out of the part they shouldn't have to do over and over.

Delegable doesn't mean hands-off

Between "AI can do it" and "send it without a look" sits a whole layer of control design. AI can draft outgoing messages, while a colleague still confirms the tone and any commitments; AI can flag anomalies in a data cross-check, while someone who knows the business still decides whether to act; document classification can run automatically, but low-confidence cases must return to a human queue. A mature adoption treats review not as a failure but as a division of labor.

Look for the repetitive and high-volume — and the low cost of error

The best first use case isn't necessarily the most time-consuming one; it's the one where mistakes are easy to spot, easy to fix, and limited in reach. For example, a first draft of meeting notes, classifying customer-service messages, filling in form fields, or a text summary of a routine report. Work like this quickly builds up real data, so the team learns where AI gets things wrong and how the rules should be shored up.

Move people up a level

When tidying, rewriting, and first-pass checks are taken off their plate, people don't become redundant — they can finally return to work that matters more: handling exceptions, understanding customers, improving processes, and making the judgments someone has to own. This isn't a romantic notion; it holds only if the company actually reallocates the time it saves, rather than cramming the same person with more odds and ends.

Next time AI comes up, try laying out one workflow and asking three questions: which step is the most repetitive? Which step has the clearest rules? Which step can be caught even when it goes wrong? The answers tend to land closer to the next actionable decision than "can it replace us" ever will.