Your marketing team is working harder than ever. They're also lying to themselves about why.
Here's the scene: a senior marketer spends Tuesday morning building a custom GPT to automate campaign briefs. Wednesday, she's debugging it because the output keeps hallucinating brand guidelines. Thursday, she's explaining to leadership why the "AI-powered efficiency gains" haven't shown up in the quarterly numbers yet. Friday, she's back to writing the brief manually because the deadline won't wait.
Nobody logged those hours. Nobody asked where they went. And when the next planning cycle rolls around, someone will propose another AI tool to "save time."
This is the productivity paradox hiding in plain sight.
The Gap Between Feeling Fast and Being Fast
A study from METR ran 16 experienced developers through 246 real tasks, half with AI tools and half without. The developers expected AI to make them 24% faster. After using it, they still believed they'd been about 20% faster. The actual result? They were 19% slower.
Read that again. The people using the tools were convinced they were speeding up while the clock said otherwise.
Marketing teams are especially vulnerable to this illusion. We're building our own workflows, stitching together prompts and automations and internal tools that don't appear in any project management system. Kevin Indig calls this "meta work": working on how to get work done. It's invisible labor that gets paid from somewhere, and organic visibility work often foots the bill.
The Numbers That Should Make CMOs Nervous
An NBER survey of nearly 6,000 executives across the US, UK, Germany, and Australia found that while about 70% of firms actively use AI, more than 80% report no impact on employment or productivity over the past three years. Executives who do use AI average only 1.5 hours per week with it. A quarter of them don't use it at all.
Meanwhile, Gartner's 2026 CMO Spend Survey shows CMOs allocating 15.3% of marketing budgets to AI initiatives. Seventy percent say becoming an AI leader is a critical goal. But only 30% report mature or fully developed AI readiness capabilities.
So we have executives who barely use the tools themselves, directing significant budget toward AI transformation, while their organizations lack the foundations to make it work. That's not a strategy. That's a hope dressed up in a slide deck.
Where the Hours Actually Go
The Work AI Index from Glean puts a number on what many of us have felt: workers now spend an average of 6.4 hours per week on what the researchers call "botsitting." That's feeding AI missing context, checking outputs, debugging mistakes, rerunning prompts, and cleaning up confident-but-wrong answers.
Six hours. Almost a full working day, every week, just maintaining the system that was supposed to save time.
And here's where it gets uncomfortable: 69% of AI users admit to "botshitting," which is shipping AI-generated work they haven't fully reviewed, don't completely understand, or couldn't defend if asked. When the overhead of verification exceeds the time saved, people start cutting corners. The efficiency gains evaporate, but the risk compounds.
A.Team's research on the CMO time trap found that marketing leaders spend 80% of their time on data mechanics: pulling reports, reconciling attribution models, chasing creative performance data from agency partners. Only 20% goes to brand strategy, consumer insight, and cultural positioning. AI was supposed to flip that ratio. Instead, it often adds another dashboard to reconcile.
The Accountability Vacuum
A bookstore's holiday ad campaign went live with garbled text and a swapped product photo. Nobody on the marketing team had touched either one. Meta's ad AI altered the approved creative after launch. The team only found out because the photographer whose work got rewritten started fielding messages calling it "AI slop."
This is the accountability question nobody wants to answer: who owns what AI changes after approval? For decades, that question had an easy answer. Now the thing most likely to alter your creative without asking is the same system you installed to move faster.
Guy Hanson at Validity suggests a fix that sounds almost quaint: lock creative from automated modification once it's approved, or run a scheduled audit within 24 hours of launch comparing live assets against approved files. Print production teams have used version locks for decades. What's new is a workflow where the verification step got skipped because everyone assumed the machine had it handled.
The Real Cost Isn't in the Budget
Abhi Yadav's diagnosis cuts to the core: AI made marketing better at doing things. It has not made marketing better at knowing what to do. Content velocity is up, cost is down, campaigns go live in hours. That part worked. But the harder problem, reasoning from your business, your audience, and what your company has already learned, remains unsolved.
Cecilia Weckström frames it as three converging forces: social collapsed the funnel, AI is collapsing it further, and the customer may soon not be human at all. When someone asks ChatGPT for a recommendation, they skip the awareness stage entirely. They don't visit your website. They don't see your ads. If your brand isn't part of that answer, you don't exist.
The hours your team spends building internal AI tools are hours not spent on the work that makes your brand the answer.
What Actually Helps
The organizations pulling ahead aren't the ones with the most AI pilots. They're the ones who finally connected their data, got specific about workflows instead of throwing AI at broad problems, and measured business impact instead of adoption metrics.
Robert Rose at Content Marketing Institute suggests the mid-year review is the moment to stop pretending. The slide that says "AI transformation: on track" next to a robot shaking hands with a person? Everyone in the room knows it's fiction. What if this year, someone said so?
Start by making the invisible visible. Track the hours spent on AI maintenance, prompt engineering, output verification, and tool debugging. Put them in the same system where you track campaign work. When leadership sees that your team spent 40 hours last month building a content automation that saved 10 hours, the conversation changes.
Then ask the uncomfortable question: is this tool solving a problem we actually have, or a problem we think we should have? The best AI implementations I've seen are boring. They automate a specific, repetitive task that was already well-understood. They don't require constant tinkering. They don't promise transformation. They just work.
Marketing is still a team sport. AI is a player, not a coach. And right now, too many teams are spending their practice time teaching the new player the rules instead of running the plays that win games.
