Around 90 percent of executives in a survey of roughly 6,000 across four countries reported no productivity gain from AI over three years. Their budgets for AI went up anyway.
That gap is the most important career fact in marketing right now. Somebody is going to be asked to close it. The person who can will get promoted.
Greg Jarboe made this argument in Search Engine Journal this week, pairing research from MIT Technology Review and Harvard Business School. His conclusion is that SEO, paid media and digital marketing managers who become directors will be the ones who prove AI pays for itself while keeping their teams intact. I agree, and I think the second half matters more than most people realise.
The ledger has two sides
Hyperscalers will spend about $750 billion on data centres this year, against total AI revenue estimated at $150 to $200 billion. Wharton's Jessica Wachter calculated that hyperscaler earnings need to grow by a factor of 2.7 by 2030 just to break even on roughly $1.1 trillion of spending through 2027.
That pressure flows downhill. Every CFO is now asking where the productivity went. And most marketing teams can only answer with one side of the ledger.
They report hours saved. They rarely report hours spent fixing weak output. Workday's research, as reported by Kevin Indig, finds that for every 10 hours AI saves, companies give back about four fixing it. A BetterUp Labs and Stanford study found 41 percent of workers had received AI "workslop" in the previous month, and each instance took nearly two hours to sort out.
I covered the hidden rework cost in edition 23. A manager who reports only the saved hours is making a claim that will not survive the first sceptical question in a budget review.
What the job market is already telling you
The Harvard research, led by Suraj Srinivasan, looked at job postings from 2019 through March 2025 across almost all US vacancies. After ChatGPT launched, postings for roles heavy on structured, repetitive tasks fell 13 percent. Postings for roles built on analytical, technical or creative work grew 20 percent.
Map that onto a marketing team. Recurring reports, bulk title tags, ad copy variations, bid adjustments, search term cleanup: those sit on the exposed side. Strategy, test design, measurement and persuading other departments sit on the enhanced side.
The career path is written into that split. Managers who spend their weeks on the first list are competing with software. Managers who spend their weeks on the second list are the ones software makes more valuable.
The research also found automation-prone postings listed 7 percent fewer skills, while augmentation-heavy roles asked more often for AI skills. The job is not getting simpler. It is moving up.
The tension you will walk into
Here is where it gets tense inside companies. The surveyed executives mostly expect to raise productivity by growing sales while cutting staff. Srinivasan's research argues firms get more from AI when they build human and AI collaboration and retrain people.
Your budget meeting will sit exactly between those two positions. The CFO wants the headcount line to shrink. The research says the value comes from people who know how to use the tools.
The manager who gets promoted is the one who can hold both. Show the productivity gain in hours and outcomes, net of rework, with numbers that survive scrutiny. And show a plan for the people whose repetitive tasks moved to the tools: what they now do, what it produces, and why it is worth more.
I would go further. The second part is what separates a director from a cost-cutter.
Anyone can remove headcount. Very few people can redeploy it into work that grows revenue, and that is a skill boards pay for. I made a similar case in don't automate the work that trains people, because the junior tasks you automate today are the training ground for next year's senior hires.
Why the cushion is getting thinner
There is a reason this pressure will intensify rather than fade. Alphabet posted revenue of nearly $120 billion last quarter and still ran a free cash deficit of about $5.9 billion, its first since going public in 2004.
A company funding AI answers from a thinner cushion than ever will keep changing how those answers look and who gets cited. For search and paid media teams, that means the ground under their channels keeps moving while their own budgets get more scrutiny.
In that environment, the manager who can explain what changed, what it cost and what the team did about it becomes the most useful person in the room. Reporting stops being a chore and becomes the promotion case itself.
Build the case this quarter
Practically, start a simple log now. For every AI-assisted workflow on your team, record time before, time after, and time spent fixing output. Run it for eight weeks. That alone puts you ahead of most of your peers.
Then pick one person whose repetitive work shrank the most, and document what they now do with the time. Tie it to a number: a test launched, a campaign improved, a client retained.
That two-page document is a promotion case. It answers the question every executive is asking and almost nobody below them can.
The tools are the same for everyone. The ledger is not.