Ask a room of managers which function AI will change last, and marketing comes up quickly. It looks like the creative one, the human one, the part of the business that depends on judgement and taste.
Fifteen chief marketing officers, interviewed by Kimberly Whitler of the University of Virginia’s Darden School for MIT Sloan Management Review, describe something closer to the opposite. In Whitler’s summary, marketing as a whole, “along with even the most granular of measurement tasks, is perhaps being transformed most deeply by the rise of AI”.
Fifteen interviews is fifteen interviews, and I will come back to that. The value here is not statistical. It is that a group of senior people in one function have named what is changing, and two of the five things they named apply well outside marketing.
The five, briefly
The priorities Whitler reports for 2027 are: establishing marketing as an executive-level growth function; transforming with and for AI; hiring to defuse what they call a talent time bomb; reaching people across a fragmented media system; and developing new measures for zero-click AI platforms.
Three of those are marketing’s own business. Two are not, and those are the ones worth your time.
The measures stop working
Start with the last one, because it is the most concrete and the least discussed.
For twenty years, a large part of knowing whether your work landed has depended on somebody clicking through to you. Visits, sessions, sources of traffic, the path from a search result to your page. Whole reporting habits are built on the assumption that interest shows up as a click.
When people ask an AI assistant instead, and the assistant answers, that click does not happen. The interest was real. Your evidence of it is gone.
For a marketing team this is an emergency, because the numbers in the monthly report stop describing the work. For everyone else it is quieter, but it is the same problem. If your team measures its usefulness by traffic to a page, downloads of a document, or hits on an internal resource, you are counting a behaviour that people are starting to skip.
There is a related observation from Simon Willison, who measured how ChatGPT searches behave and found a large rise in searches restricted to specific named domains rather than the open web. Treat that as his measurement rather than the company’s, and check it before quoting it. The direction is what matters: being the place an assistant already trusts is becoming more valuable than ranking well on a list of results.
The practical move is not technical. It is deciding, before the numbers get worse, what you will count instead. Enquiries that mention something you published. Questions that arrive already half-answered. People who turn up knowing what you do. Those are harder to count and closer to the thing you actually care about.
The constraint is hiring, not tools
The second transferable point is the talent time bomb.
The constraint these CMOs name is not software budget. It is people: the difficulty of hiring for a function whose shape is changing while you recruit for it, and the gap between the skills a team has and the ones the work now needs.
That should be reassuring to anyone who has been told the answer is a platform. It also puts the effort somewhere less comfortable. Buying a tool is a decision you can make in a week. Changing what your team is good at takes a year, and it is the part that cannot be outsourced to a supplier.
If your department is planning for next year, the question is not which tools to adopt. It is which skills you want your team to have by the end of it, and what you will stop doing to make room for that.
Why this particular function goes first
There is a reason marketing is early rather than special, and it is worth understanding because it tells you who is next.
Marketing sits at the point where the organisation meets the outside world, and that outside world now runs largely through machines: search, feeds, recommendations, assistants. When the machines in the middle change how they work, the function pressed up against them feels it immediately, whatever anyone decided internally.
Any function with that shape is on the same path. Recruitment, where candidates now arrive having had an assistant read the job description and prepare them. Customer service, where people turn up already half-answered. Sales, where a buyer has compared you against competitors before the first conversation.
The pattern is not that creative work is at risk. It is that whoever meets the outside world gets the change first, and usually without being consulted.
What to take from it if you are not in marketing
Three things, all of which cost a conversation rather than a budget.
Find your click. Identify the one measure your team reports that depends on somebody taking a step they might soon skip. Most teams have one and have never examined it.
Ask what you would count if it disappeared tomorrow. Not a perfect replacement. Two or three signs that the work is landing, which survive people getting their answers somewhere else.
Look at the function you think is least exposed. By this account, the department that looks safest is worth a second look, precisely because nobody is looking at it. Ask the people doing the work what has changed in the past six months. They will know before the reporting does.
What this evidence is, and is not
Fifteen in-depth interviews with global CMOs. That is qualitative research: no sample, no margin of error, no claim to represent marketing as a whole. It is what fifteen senior people said, and it should be described that way.
The article is also partly behind MIT Sloan Management Review’s paywall, so this piece is built on the headings, the standfirst and the passages available outside it. If you have access and marketing is your world, read the full version before acting on the detail.
Neither caveat undermines the useful part. A group of people running one function were asked what is changing and named measurement and hiring, in that order, over tools.
The department that looks least exposed to AI is rarely the one that is least exposed. It is usually the one where nobody has checked.