91% of marketers now use AI in their work but sadly only 41% can prove ROI from it – and that number dropped from 49% the year before. Adoption went up while accountability went down.

That’s not a tool problem. That’s an infrastructure problem.

McKinsey’s 2025 State of AI report surveyed nearly 2,000 organizations. Of those, only 109 – roughly 5.5% – could attribute meaningful bottom-line impact to their AI investments. The rest are running experiments that never reach scale. Marketing is where this gap is most visible, and most costly.

The reason isn’t capability. The tools are genuinely powerful. The reason is that most teams bolt AI onto broken processes and expect different results. Everyone has their own approach. Outputs vary depending on who ran the prompt. The two people who actually know what they’re doing pull further ahead while everyone else starts from scratch every session.

McKinsey identifies four consistent blockers across the organizations that can’t scale AI.

• First, data architecture: scaling requires clean, integrated, well-governed data, not the siloed spreadsheets and legacy systems most enterprises actually run on.
• Second, workflow rigidity: bolting AI onto unreformed processes produces marginal gains at best. The organizations generating real returns redesign workflows around AI – they don’t add AI to existing ones.
• Third, measurement gaps: only 19% of organizations track KPIs specific to their AI investments. Without clear metrics, pilots can’t make the case for enterprise-wide deployment.
• Fourth, stack fragmentation: only 29% of enterprise applications are actually integrated. A smaller connected stack consistently outperforms a larger disconnected one.

The organizations in that top 5.5% share one pattern. They build shared infrastructure before they build anything else. A common brand configuration. Shared context across every session and standardized prompting frameworks, with clear governance over who owns what. They treat AI the way they treat any other mission-critical system – with ownership, standards, and accountability.

For marketing specifically, McKinsey found that sales and marketing accounts for 28% of the total potential economic value from generative AI. That’s the largest single function in their analysis. The opportunity is real but realizing it requires starting somewhere most teams skip entirely.

It starts with Foundation: everything the model needs to know about your brand, your audience, and your standards before it touches a single piece of work. Not a clever prompt or the latest/greatest new tool. The shared infrastructure that makes consistent output possible – and that makes the gap between your best marketer and everyone else on the team much smaller.

That’s where the 5.5% started. It’s where everyone else should too.

Next: What happens when AI starts running the campaigns itself – and who’s responsible for the story it tells.