By James Calloway

Labor's share of the marketing budget climbed from 21.9% in 2025 to 24.5% in 2026, according to Gartner's 2026 CMO Spend Survey of 401 marketing leaders. The number matters because it lands amid an AI spending push that isn't replacing people — and in many cases is creating new work for them.

The same survey found the top barrier to AI-driven efficiency, cited by 38% of respondents, is a lack of internal AI expertise and talent.

"AI is not a shortcut around marketing capability," said Ewan McIntyre, who leads Gartner's marketing research.

How much time does AI output cost to fix?

Plenty. A survey of 2,003 marketing leaders, commissioned by Optimizely and conducted by research firm Savanta, found 76% spend at least three hours a week editing, fact-checking, or correcting AI output. More than half — 54% — said leadership underestimates the human effort required to make that output usable. MarTech senior editor Constantine von Hoffman covered the study in July.

Two more numbers from the same study:

• 39% said they're too busy managing workflows and outputs to think strategically.
• 46% worry heavy AI use will erode skill development among junior marketers.

That second figure carries a warning for anyone counting on AI to cover a skills gap: the tool can slow the very learning that would close it. And checking output takes skill too — someone who doesn't understand segmentation can't catch a bad segment, no matter how fast the tool built it.

Skills gap or headcount gap?

The MarTech analysis argues these are different problems with different fixes. A skills gap is the distance between what the work requires and what the team can do in production with its tools. A headcount gap means the team knows how to do the work and has simply run out of hours — say, a team running 12 campaigns at full quality when the plan calls for 40.

Mix them up and money goes to the wrong fix. Train people who are already maxed out, and they have no hours to use what they learned. Hire into a team that lacks skills, and the new people learn the same workarounds everyone else uses.

The piece offers three diagnostic checks:

• Look at what gets fixed. Judgment errors — wrong audience, mismatched customer records, off-brand claims — point to skills. Typos, missed deadlines, and skipped steps point to hours.
• Look at where work waits. A queue for the one person who can build a journey signals a skills gap concentrated in one person, and a headcount risk. If anyone could do the work and everyone is busy, it's hours.
• Give the team time. Pull one recurring project for a month. Better output means they were short on hours; identical output means they were short on skills.

What should teams do when it's both?

Fix skills first, the analysis argues, because the gaps feed each other: a team short on skills burns hours on rework, which makes a skills gap look like a headcount gap. Once the team knows the work, every hire and AI license added afterward makes it stronger.

That order has a real cost. Training takes months, and volume targets don't pause. In the same Gartner survey, 56% of CMOs said they lack the budget to deliver their 2026 strategy, and 54% reported insufficient resources.

One way to work both gaps at once: make the next hire a skills hire — an experienced person whose job includes teaching the work.

On budget, the article's benchmarks still hold: training at 15% to 20% of annual license cost, and one process and operations role for every three to four core platforms. Applied to AI, a $100,000 license comes with a $15,000 to $20,000 training line.

The practical close for anyone drafting an AI request this cycle: add one line naming which gap the purchase closes — skills or people — and how you know. When the answer is unclear, run the three checks first.