Every few weeks a senior technology figure gives a date. This time it was Microsoft's Mustafa Suleyman, who said that for white-collar work done at a computer, "most of those tasks will be fully automated by an AI within the next 12 to 18 months." The sentence travels quickly, and an agency owner reading it between client calls could be forgiven for feeling that something is being decided over their head.

On 3 October the Irish Times published a measured reply. Sarah O'Connor's argument is that professional work is a good deal more than the tasks done at a keyboard. She points to a Norwegian study finding that meetings take on average 12% of work hours and 14% of company wage bills, with much of that time spent on planning, problem-solving and coordinating projects. She also cites PwC's analysis of one billion job adverts across six continents, which found that employers of AI-exposed entry-level roles increasingly ask for management, strategy and teamwork, along with more face-to-face presence.

It is worth being clear about what this is. It is an analysis piece, not a new study, and nobody can say who will turn out to be right about the timeline. Taking either side with confidence would be a guess. But the argument has a practical consequence, and agencies are well placed to see it.

Think about what an account director actually does in a week. Some of it is writing: client updates, status notes, first drafts of a creative rationale. Some of it is reading, such as summarising a long brief or a research pack. A lot of it is something else entirely. It is sitting in a room where a client is unsure, reading the mood, deciding what to say and what to leave out, and knowing which of three reasonable options this particular client will trust. Agencies are paid for that second kind of work far more than they are for the typing around it.

That suggests a smaller and more useful question than "will AI take my job in 18 months?" Which tasks in this role could AI take on this quarter, and what would the person do with the time?

A worked example. An account manager spends two hours each Friday turning meeting notes and a tracker into weekly client updates. Claude can produce a good first draft of that from the notes in a couple of minutes. The account manager still reads it against what they know about the client, changes the tone where the week was difficult, and takes out anything that should be said on a call instead. The saving is real, and so is the part that stays with the person. The question for the owner is whether those recovered hours go back into the client relationship, or simply disappear into a fuller diary. That is a management decision, not a technology one.

This is also how I approach the work with teams. We start by finding out where people actually are with AI, so nobody is pitched at the wrong level. Then we work on the team's own tasks, in the tools they already have, rather than on generic examples. A team that has worked on its own tasks tends to ask better questions about what changes and what does not, and that matters more than any prediction about the timeline.

So if a bold timeline has been unsettling your team, there is a calm way to answer it. List the five or six jobs in one role. Mark the ones that are mostly writing, reading or tidying up information. Try one with the team, check the result together, and talk openly about what the freed time is for. That conversation does more for confidence than any prediction, in either direction.

Go wisely.