CMOs spend 15% on AI but only 30% can scale it

By James Dore · 2026-09-11

The numbers that don't add up

Put two numbers next to each other and see what happens. CMOs are allocating an average of 15.3 per cent of their marketing budgets to AI in 2026^[1], but only 30 per cent of their organisations have the readiness to scale it.^[1] That comes from Gartner's 2026 CMO Spend Survey, drawn from 401 marketing leaders, most of them running companies with over a billion dollars in revenue.^[1]

These are large organisations with dedicated budget cycles, strategy teams and procurement functions. They've decided, collectively, that AI gets roughly one dollar in every six. Then three quarters of them admit they can't do much with it at scale. That gap is a planning and governance failure, and more tool spending will widen it.

Gartner's VP Analyst and Chief of Research, Ewan McIntyre, put it plainly: "CMOs recognize AI's potential as a force multiplier for growth, efficiency and transformation, but most marketing organizations are not yet built to capture that value."^[1] McIntyre's second observation is the one worth printing on every budget template: "The risk is that CMOs invest in AI tools faster than they build the data foundations, processes, governance and talent required to scale them."^[1]

Australian data confirms the pattern locally. ACAM's 2026 Australian AI in Marketing Benchmark Report, based on 126 CMOs and senior marketing leaders assessed through July 2026, found 83 per cent of organisations at early-to-mid AI maturity^[2], and none of the 126 reached either of the two highest maturity levels.^[2] In a country where the marketing industry talks constantly about AI transformation, zero organisations have actually completed it by any meaningful measure.

What the seven drivers actually reveal

ACAM's diagnostic framework evaluates organisations across seven maturity drivers: Leadership and Culture; Skills and Capability; Governance, Risk and Brand; Data and Technology Readiness; Use Cases and ROI; Team Design and Workflow; and Roadmap, Planning and Investment.^[3] The framework matters because it shows exactly where money is going relative to where the capability gaps sit.

The weakest driver across the entire benchmark is Roadmap, Planning and Investment.^[2] Roadmap is the capability that determines whether AI spending is sequenced, whether it connects to measurable outcomes, and whether there's a coherent plan for building on what works. When that driver is the weakest one, you're looking at an organisation spending without a map and calling it momentum.

The other drivers fill in the picture. Governance, Risk and Brand covers the guardrails: brand safety, data ethics, regulatory exposure. Data and Technology Readiness covers the foundations every AI model depends on: clean data, integrated systems, reliable pipelines. Skills and Capability covers whether the people using the tools know how to get results from them. These are the load-bearing walls. Spend on tools before you've built them, and the tools don't scale; they generate impressive demos and disappointing quarterly reviews.

The counter-argument: you learn by doing

There's a reasonable version of the opposing case. Governance frameworks built before anyone has hands-on experience are theoretical at best and paralysiing at worst. You can't write a roadmap for a technology you haven't used. The experimentation budget is the learning budget. Organisations that moved early will have built institutional knowledge that cautious ones can't buy back.

That argument holds at a certain scale. A team of five running pilots on a modest budget can learn by doing without much damage from a wrong turn. Gartner's survey covered companies with over a billion dollars in revenue, allocating 15.3 per cent of already-substantial marketing budgets to AI. At that scale, learning by doing without governance isn't experimentation; it's an expensive bet that the problems you create won't cost more than the value you generate.

The issue is the absence of a framework for deciding which experiments are worth running, how to measure whether they worked, and when to stop. That's the Roadmap driver. Without it, every experiment is an island: interesting, possibly useful, and impossible to build on at enterprise scale.

What to cut and what to keep

The answer is not to stop spending on AI tools. Some tool spending is producing real, measurable results, and those investments deserve protection. The ACAM framework's Use Cases and ROI driver exists precisely to identify which AI applications have demonstrated value and which ones are still searching for a business case.^[2] The productive question is how to make AI spending legible, not how to slow it down.

Practically, that means three shifts. First, redirect a portion of the tools budget toward the readiness capabilities holding the organisation back, most likely Roadmap and Planning, then Data and Technology Readiness, then Governance. Second, protect the tool spending that already has a demonstrated ROI: the use cases where the workflow is embedded and the team knows how to run it. Third, build the measurement framework before committing the next round of budget by deciding what scaling AI looks like in measurable terms before spending on the assumption that it will happen.

ACAM is now taking its benchmark diagnostic to the US market, announced in September 2026, which suggests the Australian pattern is considered internationally transferable.^[3] If US results echo what 126 Australian organisations revealed, 83 per cent at early-to-mid maturity and none at the top two levels, then the Gartner gap is an industry-wide structural problem dressed up as an adoption curve. Gartner's survey covering those US organisations was conducted between January and March 2026.^[1]

James Dore

James Dore is founder and editor of Prompt the Market and Strategy Director at Bushnote, a Sydney strategy, SEO and AI search agency. He has a background in strategy, media and financial services, and led consumer product marketing at MONEYME. He writes on AI search, answer engines, technology policy and media economics.