Marketers are spending record hours building and fixing AI tools that rarely appear on dashboards. New research shows promised productivity gains often vanish as teams trade real output for invisible maintenance.
Marketing teams are pouring hours into AI projects that never make it onto the official roadmap. The promise was speed and scale, but what’s actually piling up is a backlog of internal tools, constant upkeep, and a steady drain on the work that matters most.
In 2025, METR tracked 16 experienced developers across 246 real-world tasks. Half used AI tools, half didn’t. The expectation was a 24% speed boost from AI. Instead, the AI group was 19% slower. Even after seeing the numbers, most still believed they’d worked faster. This is the AI productivity paradox: work feels quicker, but the hours shift from doing the job to prompting, checking, and fixing. A Search Engine Land analysis points to the extra time spent on prompt engineering, verifying, and correcting AI output as the root of this disconnect.
“Despite being slower with AI, participants in the METR study still believed they were working about 20% faster after the experiment, highlighting a persistent gap between perception and reality.”
- Search Engine Land
For marketers, these costs are even harder to spot. HubSpot reports that 91% of marketing leaders say their teams use AI, and 66% are building custom internal tools. But these projects rarely show up in trackers or dashboards. The hours spent learning, building, and maintaining these tools come out of the time that would otherwise go to content, digital PR, and community work-the things that actually build brand value.
One person’s shortcut is another’s time sink. BetterUp Labs and Stanford surveyed 1,150 full-time U.S. workers about “workslop”-AI-generated output that looks finished but isn’t. Forty-one percent received workslop in the past month, and each instance took nearly two hours to clean up. At a 10,000-person company, that’s over $9 million a year in wasted time. The sender saves 20 minutes, but someone else spends two hours untangling the mess. Output metrics reward the sender, but the real cost is buried elsewhere. According to the BetterUp Labs and Stanford Social Media Lab study, a single episode of workslop was estimated to cost about $186 per employee per month, showing how low-quality AI output quietly eats into team budgets.
Workday’s research puts numbers to the rework: for every 10 hours AI saves, companies lose 4 to fixing and rewriting. Upwork’s survey of 2,500 leaders and workers breaks it down: 39% of AI users spend extra time checking and fixing output, 23% on learning the tools, and 21% just handling more work.
“The METR study found that experienced developers using AI tools completed tasks on average 19% slower than those working without AI, making this one of the most cited empirical findings on the hidden costs of AI-driven workflows.”
- Search Engine Land
Every new AI workflow becomes a maintenance job. The day a custom tool launches, the clock starts ticking: models update, integrations break, and the original builder becomes the default owner. When that person takes time off, the process reverts to manual. None of this shows up on dashboards unless you hire a dedicated content or marketing engineer-which most teams don’t.
Executives see efficiency gains. Individual contributors see broken outputs and constant revision. Strategists watch as time for brand-building and organic search shrinks. Teams are building tools to work faster, but the work that actually earns citations, mentions, and reviews gets pushed aside. Lost months of brand authority can’t be reclaimed.
Learning and pressure-testing AI is now part of the job for growth marketers and organic search strategists. But this is extra work-the old tasks remain, and the new ones stack on top. Teams are building custom AI workflows even as they pay vendors for similar tools, chasing the feeling of instant productivity. The catch: every workflow needs ongoing care, and the hours spent are invisible until missed opportunities and delayed ROI make them obvious.
When marketing attention is split, the first to suffer are the slow-burn projects-publishing deep content, earning mentions on sites that feed AI answers, and showing up in the forums and threads where buyers actually look. These are the efforts that build lasting brand authority, but they’re the easiest to cannibalize for the meta work of AI tool-building and cleaning up workslop.
As reported earlier, the fight for organic visibility is already shifting as AI overviews and chatbots change who gets seen. Teams that lose focus on foundational brand work in favor of endless AI tinkering risk falling behind in the only race that matters: being the source that AI systems and search engines actually recommend.
Building in-house AI tools isn’t always a mistake. I’ve built and tested dozens for myself and clients, and some have delivered real gains. But the lesson is clear: homebrew AI collapses some tasks and quietly slows others, and from a distance, both look the same. The skill that matters now is knowing when an AI workflow truly delivers efficiency and when it just drains time from the work that builds lasting brand value. Buy what you can, build only what you must, and don’t lose sight of the work that actually earns your brand a place in the answers that matter.
HubSpot, founded in 2006, reported annual revenue of $2.2 billion in 2025 and serves over 184,000 customers worldwide. The company’s push into AI-powered marketing tools has made it a leading vendor for both off-the-shelf and custom workflow solutions, with adoption rates among enterprise marketing teams reaching record highs in the past year.
