Skip to content
← Latent

Industry/

What CMOs underestimate about the AI investment

Gartner asked 401 of them. AI has 15.3% of the budget and 30% are ready to use it. The gap is not the tools.

The question arrives in a procurement format: what does this cost, and what do we save. It is the wrong pair of numbers, and the evidence that it is wrong now comes from the people being asked. Gartner surveyed 401 CMOs and marketing leaders between January and March 2026, most of them at companies above a billion dollars in revenue. On average they had already allocated 15.3 per cent of the marketing budget to AI. Thirty per cent said their organisation was actually ready to use it.

That is the whole story in two figures. The money is not the constraint. It has been assigned. The thing it was assigned to has not been built.

The budget is flat and the allocation is not

Total marketing budgets in the same survey sat at 7.8 per cent of company revenue, up from 7.7 the year before — flat, in any honest reading. So the 15.3 per cent going to AI is not new money. It came out of something else that used to get it.

Which means the investment question has already been answered inside most companies, and answered before anyone worked out what the money would land on. Seventy per cent of the same group called becoming an AI leader a critical goal for 2026. Seventy per cent also said their own processes were not mature enough to implement and scale it. The same people, in the same survey, holding both positions at once.

15.3%

Share of the marketing budget allocated to AI by 401 CMOs. Thirty per cent of them described their readiness to use it as mature.

Ewan McIntyre, who leads research for Gartner's marketing practice, put the mechanism plainly: firms “invest in AI tools faster than they build the data foundations”. The sentence is about sequencing, and sequencing is where the cost hides.

The expensive part is downstream of the generation

Here is what we see from the production side, and it matches the survey rather than softening it. A model does not reduce the amount of work. It moves the work.

One prompt does not produce one option. It produces a spread — realistically a couple of hundred frames or cuts that are all technically finished and all slightly different. Nothing about that is slow. What is slow is that somebody with taste and authority now has to look at two hundred finished things and decide, and then somebody with legal responsibility has to decide again. The bottleneck used to be the shoot. The bottleneck is now the review.

Generation got cheap. Judgement did not, and approval got more expensive, because there is more to approve.

This is the part that does not appear in a tool budget. It is headcount, it is calendar time from senior people who were already the scarcest resource in the building, and it is a governance process that has to be written before it can be followed. None of those three things is a licence fee.

It is also why the time‑saved metric flatters the project. Time saved is real and it is measurable and it is the one number that improves immediately, which is exactly why it gets reported. It measures the step that got cheap. It is silent on the two that got more expensive.

The organisations that are ready spend more, not less

This is the finding in the Gartner data that most directly contradicts how the investment gets pitched internally. The organisations Gartner classed as AI‑ready were not spending less. They allocated 21.3 per cent of budget to AI — well above the 15.3 average — and spent 8.9 per cent of revenue on marketing overall, against the 7.8 average.

So the leaders are the heavier spenders on both counts. Whatever AI maturity is, it did not arrive as a saving. It arrived as a larger commitment that then produced a return, which is a different proposition entirely and needs to be sold internally as one.

The constraint showing up alongside it is ordinary: 56 per cent said their budget could not fund their own 2026 strategy, and 54 per cent reported insufficient resources. A plan that depends on AI paying for itself in‑year is a plan competing against those two numbers.

The personal underestimate

There is a second Gartner survey, of 402 senior marketing leaders, fielded between August and October 2025, and it finds the gap in a sharper place. Sixty‑five per cent said AI would dramatically change the CMO role within two years. Thirty‑two per cent said significant changes were needed to the CMO skill set.

Most expect the disruption. Half as many expect it to require anything of them personally. Gartner's own forecast is that by 2027, a lack of AI literacy will be among the top three reasons CMOs are replaced at large enterprises — and literacy here means something specific and unglamorous: being able to tell a good output from a plausible one, and to know which use case is worth the governance it will cost.

That is not a technology skill. It is the editorial judgement the role always needed, applied to a much larger pile.

What we would tell a brand to budget for

Four lines, in the order they bite.

The review, not the render. Decide who signs off, on what, and how fast, before anything is generated. An approval chain that works at four films a quarter does not work at forty, and discovering that in week three is the common failure.

The cleared‑model question, settled first. Which models a brand is permitted to use at all is a contractual question with a short answer, and it determines what can be made. We have written separately on which models are cleared for commercial work.

Disclosure as a production decision. What a campaign will have to declare depends on the markets it runs in, and some of it lives inside the file rather than on the screen. That is covered in what a Singapore brand has to label.

Somebody else's maturity, rented. The honest case for using a studio is narrow and worth stating as such: the process, the governance and the taste are the part that takes a year to build internally, and they are the part you can borrow while you build them. The tools are not the thing anybody needs help buying.

What this does not mean

None of the above argues that the spend is wrong. Fifteen per cent of a flat budget moving into AI is a rational bet on a cost curve that genuinely fell, and the readiness figure will not sit at 30 per cent for long. The argument is narrower: the investment that gets underestimated is not the licence, the compute or the studio fee. It is the review capacity, the data a model has to be pointed at, and the written process that lets a decision get made twice the same way.

Those three are not AI costs. They are the costs of having an opinion at volume, and a brand that was already slow at having one will be slower, not faster, with a model attached.

Where this came from

The budget, readiness and maturity figures are from the Gartner 2026 CMO Spend Survey, 401 respondents across North America, the UK and Europe, fielded January to March 2026 and presented in London on 11 May 2026. The role‑disruption and skills figures are from a separate Gartner survey of 402 senior marketing leaders fielded August to October 2025 and published 23 February 2026. Both are surveys of large‑enterprise marketers in North America and Europe, so neither is a reading of this region or of smaller brands. The observations about review load and option volume are ours, from production, and are not in the Gartner data.

sillyhumans is an AI creative studio in Singapore. Brand films and campaigns, editorial stills, interiors, pitch decks and brand identity, for brands internationally.

See the workStart a brief

Next — Compliance How brands are rewriting their AI ad policy Read