AI marketing tools are selling the same efficiency story programmatic buying sold a decade ago. New data suggests the hours saved on the visible task are being quietly spent on setup, repair and maintenance.
The hidden hours inside AI workflows
Search Engine Journal covered Kevin Indig’s Growth Memo column and shows AI has not removed marketing work; it has relocated it. The numbers make that gap concrete.
In a METR experiment, 16 experienced developers tackled 246 real-world tasks with and without AI assistance. They expected roughly a 25% speed boost. Instead they finished about 20% slower, and they still believed afterward that the tool had made them faster.
Marketing has its own version. A BetterUp Labs and Stanford survey of more than 1,000 workers found AI-generated “workslop” — content that looks finished but isn’t — took nearly two hours to fix each time it arrived. At a large company, that cost can exceed $9 million a year. Workday’s research shows that for every 10 hours AI saves, roughly four go back into redoing weak output.
Upwork polled 2,500 leaders and workers and found reclaimed time often disappears into checking and fixing AI output, learning the tools, or simply taking on more work than before.
Programmatic already broke this promise
The pattern should feel familiar. Programmatic buying promised automation would replace manual media buying labor. What it actually did was convert manual buying labor into fraud monitoring, brand safety and compliance labor — none of which appeared in the original pitch deck.
HubSpot data shows most marketing leaders say their teams already use AI, and a solid majority are building internal AI tools rather than buying them. That in-house building doesn’t stop when the tool ships. It becomes a permanent, mostly invisible maintenance job. When the owner goes on vacation, the workflow often reverts to manual until they return.
The real issue isn’t that AI is useless. It’s that efficiency claims in marketing technology keep getting measured on the wrong side of the ledger: the hours saved on the visible task, never the hours spent setting up and babysitting the system.
Three checks for marketing teams
If you manage a marketing team building its own AI stack, these three moves come straight from the programmatic playbook:
- Put a name and a review date on every internal AI tool. Programmatic used standards like ads.txt to make accountability structural. Your homebrew workflows need the same discipline, or they become invisible headcount.
- Track hidden hours, not just saved hours. Ask your team how many hours went into building, fixing or maintaining AI tools. The METR study shows people can report time savings even when the data shows the opposite.
- Protect slow-return work on purpose. Content depth, digital PR and the mentions that actually get cited in AI answers are often the first thing squeezed out by AI tool maintenance. Ring-fence a fixed share of team time before AI claims the rest by default.
The takeaway for performance marketers is simple. Measure the whole system, not the single task. If an AI workflow only looks efficient on the visible step, you’re repeating programmatic’s accounting error with a newer interface.
Source: Search Engine Journal
