McKinsey's Research Found a Gap I Missed in the AI Marketing Infrastructure Argument

AI Marketing Infrastructure Isn't Enough. Here's What's Missing.

McKinsey just published a piece on AI capabilities reshaping marketing and I read it the way I read most major research, looking for where it confirms what practitioners are already seeing and where it surfaces something worth paying attention to that isn't already in the conversation. This one did both.

The confirmation:

Ninety percent of CMOs are experimenting with AI across their marketing workflows and fewer than ten percent have scaled those experiments or captured real value from them. I've been making this argument for over a year. Most companies aren't failing at AI because the technology doesn't work. They're failing because they haven't changed anything fundamental about how they operate. McKinsey's data puts a number on it. So does this: Estimates are that AI could unlock up to $90 billion in improved marketing returns in the US alone. The gap between that potential and what organizations are actually capturing isn't a technology gap. It's a strategy and AI Marketing Infrastructure gap.

I've also written at length about why AI amplifies existing infrastructure problems rather than fixing them, and about what it means for your content to be machine-readable and citable in an AI search environment. Those arguments are holding up. If anything, the this new data makes the case harder.

What the research pushed me to think about differently is the organizational restructuring question and a specific exposure that I hadn't fully framed before. Both are worth unpacking here.

What McKinsey Confirmed

The core argument I've been making is this:

AI is not a strategy, it's an amplifier. If your workflows are broken, AI makes them break faster. If your data infrastructure is weak, AI optimizes against the wrong signals at scale. If your pilots don't connect to anything, AI gives you better-looking disconnected pilots.

McKinsey's survey confirms only 28 percent of organizations are pursuing a fundamental rewiring of their teams and workflows. Everyone else is running point solutions with no defined path to integration.

In B2C, that looks like an AI content production pilot that generates more assets without connecting to how those assets perform, which never feeds back into what gets produced next. Volume goes up. Conversion doesn't.

In B2B, it looks like an AI tool that helps SDRs write better outreach emails. And maybe it does. But it's not connected to intent data, it's not informed by what content is actually influencing pipeline at each stage, and it's not feeding anything that changes how marketing allocates resources. The SDR sends better emails into the same broken funnel.

Pilots stay pilots when they're designed as experiments instead of infrastructure. The value of any single AI capability depends almost entirely on what it's connected to. Insights that don't feed decisions are just reports. Personalization that runs on stale segments isn't personalization. It's just segmentation with a new label.

AI is not a strategy. It's an amplifier. If your pilots don't connect to anything, AI gives you better-looking disconnected pilots.

What the Research Pushed Me to Think About Differently

The trust economy shift hits harder than I'd framed it

I've written before about machine legibility and credibility signals, mostly through the lens of AI search visibility and content infrastructure. What I hadn't fully articulated is how acute this exposure is when it comes to vendor evaluation specifically.

When a procurement team, a CFO, or a department head asks an AI assistant to research vendors in a category, that AI is doing what a junior analyst used to do: pulling from structured sources, summarizing reviews, comparing specifications, and returning a shortlist. The sales team never gets a call. There's no RFP. There's a shortlist, and if you're not on it, the conversation started without you.

McKinsey names this the shift from an attention economy to a trust economy. That framing lands harder than I expected when you apply it to complex, high-stakes purchase decisions. Brand credibility is now being evaluated by machines before any human at the prospect organization has looked at your website. That's not a content problem you can solve with better copy. It's a credibility infrastructure problem: structured data, verified information, detailed specifications, expert-attributed content, consistently updated signals across every surface where AI systems pull from.

They also confirmed that agentic commerce is where organizations feel least ready across the board. The organizations with the most to lose from that gap are the ones where sales cycles are long, deal sizes are large, and being excluded from an AI-generated shortlist is harder to recover from than a bad search ranking.

Brand credibility is now being evaluated by machines before any human at the prospect organization has looked at your website.

The organizational restructuring argument is further along than most realize

This is the one I hadn't written about directly. McKinsey's research points to a structural direction that's worth naming clearly: marketing organizations that get AI right will become smaller, faster, and organized around end-to-end workflows instead of functional silos. Not as a cost-cutting measure. As a design principle.

Distinct roles are emerging for people who build AI systems, people who orchestrate human-agent collaboration, and people who provide the judgment and quality control machines can't replicate. Skills development was cited as the number-one barrier to AI adoption in McKinsey's survey. Not budget. Not technology access. The people challenge is the real one, and most organizations are investing in tools at ten times the rate they're investing in the capability to use them.

Less than a quarter of the senior marketers surveyed have a clear, sequenced road map for making this organizational shift. That's not a technology gap. It's a leadership gap.

Productivity gains that go unreinvested are value that disappears

This is a specific failure mode McKinsey surfaces that I want to add to how I think about measurement. It's not just that companies measure activity instead of value. It's that when AI automation frees up time or reduces costs, those gains often go unreinvested because nobody made a deliberate decision about where to put them.

AI might reduce content production time by 40 percent. If that time isn't redirected into higher-value work, the 40 percent doesn't generate returns. It evaporates. This shows up clearly when an initiative gets declared successful on efficiency grounds while pipeline sits unchanged and no one can connect the work to closed deals.

I've written about what real measurement infrastructure looks like for mid-market organizations. The point this research adds is that measurement has to extend to where the freed-up resources actually go, not just whether the AI capability is functioning.

Productivity gains that go unreinvested aren't savings. They're value that evaporates.

What This Means in Practice

The companies that are actually changing their economic outcomes aren't running better pilots. They're building connected systems where each capability learns from and feeds the others, and they're deliberate about where the efficiency gains get redeployed.

The case studies are consistent. A consumer technology company that embedded AI across insights, creativity, personalization, agentic commerce, and orchestration simultaneously saw campaign activation time drop 35 to 50 percent, external spend fall roughly 20 percent, and processes that took ten to twelve weeks run in minutes. A financial services company that invested in structured data and machine-readable content infrastructure saw organic traffic increase sixfold within nine months while improving customer quality and reducing acquisition costs.

Neither outcome came from a single tool. They came from treating AI as infrastructure: connected, compounding, designed from the start to show up in revenue.

If you're in the 90 percent still running disconnected experiments, the question isn't which AI capability to build next. It's whether you have an honest picture of where your infrastructure actually stands, and whether your organization is structured to compound on what it learns. Those are the two questions the research pushed to the front for me, and they're the ones worth thinking about seriously.

---

FAQ

Why are so few AI marketing initiatives scaling beyond the pilot stage?

McKinsey's research found that only 28 percent of organizations are pursuing a fundamental rewiring of their teams and workflows. Most are running point solutions with no defined path to integration. A pilot connected to nothing stays a pilot, regardless of how well it performs on its own terms.

What does the shift from an attention economy to a trust economy mean for marketing?

It means the criteria by which brands get evaluated and recommended has changed. AI systems are increasingly making or informing those evaluations, which means your brand needs to be structured for machine interpretation, not just human discovery. Credibility signals that machines can read and validate matter as much as the content itself.

What is the biggest organizational barrier to AI marketing adoption? 

According to McKinsey's survey of senior marketers, skills development is the number-one barrier, ahead of budget and technology access. Most organizations are investing in tools far faster than they're investing in the people capability to use those tools effectively.

How should marketing organizations restructure for AI?

The research points toward smaller, faster teams organized around end-to-end workflows instead of functional silos. Three types of roles are emerging: people who build AI systems, people who orchestrate human-agent collaboration, and people who apply judgment, creativity, and quality control that machines can't replicate.

What does "productivity gains that go unreinvested" mean in practice?

When AI reduces the time or cost of a task, that freed-up resource needs to be deliberately redirected into higher-value work. If it isn't, the efficiency gain disappears without generating a return. Measurement frameworks need to track not just whether AI is performing, but where the savings are actually going.

---

Sources:

• McKinsey "From campaigns to continuous growth: AI capabilities shaping marketing," June 2026

• McKinsey Global Survey on marketing technology, August 2025

• McKinsey Global Survey on marketers, March 2026

• McKinsey Global Survey on marketing executives, May 2026

• McKinsey AI Discovery Survey, August 2025 "The agentic commerce opportunity," McKinsey, October 2025; "Reinventing marketing workflows with agentic AI, April 2026.