Almost nobody actually decides this.
What happens is that the marketer leaves in March. You think about replacing them, the tools are good now, and you decide to see how it goes for a quarter.
By June the output looks about the same. Maybe more of it. The posts go out, the emails go out, the deck for the sales team got made in an afternoon.
By the following March somebody asks why the pipeline is thinner than last year and nobody in the room can answer, because the person who would have known the answer left fourteen months ago and was never replaced.
So: no, and not because the tools are bad. They are genuinely good at the thing they do. The problem is that you would be replacing the half of the job that was never the bottleneck.
What was actually scarce
Think about what a decent marketer in a small company spends a year doing, and sort it by how hard it was to get.
Making things was always the easy part to buy. Before AI you could buy a writer, a designer, a freelancer in a timezone that suited you. Production was a market with a price, and if you had money you had output.
What was hard to buy was the set of decisions. What we sell and to whom. Why anybody would pick us over the obvious alternative. Which of the nine things we could do this quarter is the one that moves anything. What we stop doing.
And one more, which gets forgotten because it is not a skill: somebody whose name is on the number. A person who notices in week three that it is not working, says so out loud, and changes it.
A tool makes things. It does not hold a position, and it cannot be accountable, because accountability requires somebody who experiences a consequence.
The two quarters where this looks like it worked
Here is why this decision is so easy to get wrong, and it is a measurement problem rather than a judgment problem.
Marketing output is visible immediately. Marketing results arrive one to two quarters later, through a pipeline that has other things wrong with it, in a number that moves for six reasons.
So for roughly two quarters after you stop having a marketer, the dashboard looks fine. Volume is up, cost per piece is down, the obvious metrics all point the right way. You have removed judgment from the system and the system has not noticed yet.
The bill arrives as a slow drift. Positioning gets a bit vaguer because nobody was defending it. The offer stops being rewritten as the market moves. Nobody kills the campaign that stopped working in month two. None of those show up as an incident. They show up as a year that was somehow flat.
A thin marketing function and a healthy one look identical for about six months. That is the whole trap.
The cost did not disappear, it moved
There is a finding on this I keep coming back to, with a caveat attached.
A survey by BetterUp Labs and Stanford’s Social Media Lab, written up in Harvard Business Review in September 2025, named a thing called workslop: AI-generated work that looks finished and is not, so the person who receives it has to verify, correct or redo it. Their reported figure was that around 40% of US desk workers had received some in the previous month, and each instance took close to two hours to sort out.
The caveat: I could not get behind HBR’s paywall to the method, and the widely quoted dollar cost per employee is a model built on self-reported survey answers, so I would not quote it at you as a fact. Treat the magnitude as soft.
The finding I do trust, because it does not depend on the modeling, is the one about reputation. People who received workslop rated the sender as less capable and less trustworthy. That is a judgment about a colleague, and it transfers directly to how a prospect reads a company.
Which is the actual mechanism. The production cost fell and the verification cost rose, and the verification moved onto somebody who was not budgeted for it. In a small company that somebody is usually you, in the evening, fixing a draft you did not plan to touch.
What the people spending the money have found
Worth knowing what is happening at companies with resources, because they tried this first.
Gartner’s 2026 CMO Spend Survey found marketing leaders putting an average of 15.3% of their budgets toward AI, while only 30% said their organization was ready to scale it. The sample is 401 senior marketers at companies above $1 billion in revenue, so read it as a report on large firms rather than a benchmark for you.
Still, the shape of it is informative. The money went in first and the capability to use it came second, which is the ordinary pattern for every tool I have watched arrive in this industry. The firms getting value are the ones with somebody senior enough to decide what not to automate.
Nobody is reporting that they fired the marketing team and it worked.
The honest split
Being specific, because “AI changes everything” is not a plan and neither is refusing it.
Genuinely moved: first drafts, variants of something already decided, research summaries, reformatting one asset into four, transcription, translation, the tedious half of reporting. That is a real chunk of hours and you should take it.
Did not move: deciding the offer, setting the price, knowing the buyer well enough to predict the objection, choosing what to stop, and carrying the number. Also, interestingly, the taste to tell whether a draft is good, which is the one thing the tool cannot supply and the one thing you need most if you are using it heavily.
Notice the pattern. The tools made execution cheap, which raises the value of judgment rather than lowering it, because judgment is now the only scarce input in the chain.
So what would I actually do
If a marketing role just came open, I would not refuse to refill it and I would not refill it with the same job description.
Write the role around the decisions. Fewer hands, more responsibility. One person who owns positioning, the offer, and the number, who uses the tools heavily for production and is measured on outcomes rather than output. That person is more expensive per head and cheaper per result, and you need one of them rather than three of the old kind.
And if you have already been running without a marketer for two quarters and the dashboard looks fine, go and check the one number that is slow. Not impressions, not pieces published. Inquiries from people who had heard of you before they needed you.
That is the number that decays quietly, and it is the one nobody notices until the year is already flat.
If you’d like to think through what the role should actually own before you post it, drop me a line on email, WhatsApp, phone or LinkedIn and we can have a quick chat, and my notes on the work itself cover what the job involves in practice. I’m taking on a small number of contracts at the moment, so I’ll tell you straight whether it’s something I could help with.