McKinsey Identifies Five AI Capabilities Reshaping Marketing. Elf Beauty's Growth Rate Implies There's a Sixth.
Elf Beauty's revenue grew 40% year-over-year while legacy beauty incumbents posted flat or declining numbers with marketing budgets three to ten times larger. The company's CEO, Tarang Amin, attributed a significant portion of that outperformance to one structural decision: refusing to let marketing operate as a disconnected function.
That same week, McKinsey published its framework for the future of AI in marketing, identifying five core pillars: insights, creativity, personalization, agentic commerce, and orchestration. The framework is well-constructed. It maps the genuine capability terrain. And it has nothing to say about organizational integration.
That omission is not a minor editorial gap. It is the variable that predicts which companies will capture disproportionate value from AI marketing investment and which will see expensive capabilities underperform in structurally broken containers.
The McKinsey Framework and What It Gets Right
McKinsey's five-pillar model is worth taking seriously. The pillars are not arbitrary categories -- they map the actual layers where AI is transforming marketing operations.
Insights capability means AI-driven analysis that moves beyond backward-looking campaign reporting into predictive modeling of consumer behavior. Creativity capability means generative tools that reduce production costs and time-to-market for content at scale. Personalization is the ability to tailor messaging and offers at individual level rather than segment level. Agentic commerce means AI systems that can execute transactions, not just influence them. Orchestration is the connective layer -- coordinating signals across channels in real time rather than running disconnected campaigns in parallel.
Each pillar is substantively correct as a description of where AI is adding value. The problem is that the framework presents these as capabilities an organization can acquire, as if the main barrier to AI marketing performance is access to the right tools.
The evidence from Elf Beauty suggests otherwise.
What Elf Beauty Actually Did
Kory Marchisotto, Elf Beauty's former CMO, operated from a position that is unusual in corporate marketing: she was not running marketing as a separate function that delivered campaigns to the rest of the business. She was embedded in the operational core, with marketing decisions connected directly to product, distribution, and pricing decisions. Tarang Amin described this integration as transformative -- not because it produced better creative, but because it produced better decisions.
That distinction matters. The McKinsey framework optimizes for marketing output quality (better insights, better content, better personalization). Elf Beauty's advantage was marketing decision quality, which depends on how marketing is connected to the rest of the organization, not on the sophistication of its tools.
The two things are related but not the same. A highly capable marketing function that operates as a silo will produce excellent outputs that are frequently misaligned with operational reality. An integrated marketing function with more modest tools will produce outputs that compound across the organization because they are informed by and connected to actual business conditions.
The Neuromarketing Cost Collapse Changes the Comparison
This distinction becomes more acute when you look at what is happening to the cost of behavioral science tools. Roger Dooley's Neuromarketing blog describes what he calls Neuromarketing 2.0 -- the collapse of the cost barrier that previously limited behavioral science to organizations with seven-figure research budgets. fMRI studies that cost $50,000 or more per engagement are being replaced by AI-driven behavioral analysis available at a fraction of the cost. Eye-tracking rigs and EEG equipment that required specialized labs are being approximated by software tools that any marketing team can deploy.
This is genuinely democratizing. The behavioral science knowledge gap between a large CPG brand and a direct-to-consumer startup has narrowed substantially in the last two years.
But here is what the democratization story misses: Elf Beauty was not outperforming because it had superior behavioral science tools. It was outperforming because its marketing function had structural access to operational decisions. Giving every brand the same behavioral science toolkit does not equalize outcomes if the organizational architecture that processes those insights is fundamentally different.
This is precisely what the neuroscience of consumer trust research keeps surfacing: the consumer's brain resolves brand trust questions in under 400 milliseconds using heuristics built from the total organizational experience, not from any single campaign touchpoint. The signal that feeds those heuristics is produced by organizational coherence, not creative quality.
PepsiCo's Relevance Problem Is a Structural Story, Not a Brand Story
PepsiCo's stated challenge -- redefining relevance as consumer preferences shift toward health, functionality, and value -- is usually framed as a brand positioning problem. The commentary focuses on messaging, portfolio, and campaign strategy.
But PepsiCo's relevance deficit is not primarily a communications failure. It is an organizational signal-processing failure. Consumer preferences have been shifting visibly for more than a decade. A company with genuinely integrated marketing -- where consumer insight is connected to product development, pricing, and distribution decisions -- would not be scrambling to redefine relevance in 2026. The insight was available. The organizational architecture to act on it was not.
This is where the McKinsey framework's focus on AI capabilities risks pointing organizations in the wrong direction. PepsiCo can acquire world-class AI insights tools, personalization infrastructure, and agentic commerce capabilities. None of those acquisitions address the structural problem that allowed ten years of consumer preference data to not translate into product and portfolio decisions.
The Missing Pillar: Organizational Integration Architecture
The original contribution this analysis makes is one that does not appear in McKinsey's framework or in any of the source material: organizational integration architecture is the sixth AI marketing capability, and it is the one that determines the yield on the other five.
Here is the specific claim: AI amplifies existing organizational structure rather than compensating for it. A siloed marketing function that adopts all five McKinsey pillars will see AI accelerate its production volume, improve its targeting precision, and expand its channel reach -- while remaining disconnected from the operational decisions that determine whether any of that output actually compounds. An integrated function with more modest AI adoption will see slower capability gains but dramatically higher decision quality yield, because its outputs feed directly into the processes that convert consumer insight into product, pricing, and distribution action.
This means the ROI comparison between AI-heavy siloed marketing and AI-light integrated marketing consistently favors the latter -- at least until AI capabilities reach the point where they can replace organizational integration entirely (which would require AI systems to participate in operational governance, not just marketing execution).
We are not there. The current generation of AI marketing tools makes the capabilities in McKinsey's framework faster and cheaper to execute. They do not create the organizational connections that allow marketing insight to influence product roadmaps, pricing structures, and channel strategy in real time.
The Same Credential, Different Outcomes Problem
Of Dollars and Data published research this week examining why students from lower socioeconomic backgrounds earn less than their affluent peers after graduating from the same universities with the same degrees. The mechanism, per MIT's Anna Stansbury, is structural: affluent students arrive with professional networks, family capital, and behavioral scripts for navigating credentialed environments that lower-income students do not have, regardless of what they study or how well they perform academically.
The parallel to AI marketing is direct. When McKinsey documents five AI marketing capabilities that are becoming accessible to organizations of all sizes, it is describing the equivalent of university access expansion. The credential (the capability) becomes more broadly available. The structural advantages that determine who captures value from the credential remain unevenly distributed.
Elf Beauty has the structural advantage of organizational integration. PepsiCo has marketing scale and AI budget. The graduate outcomes data suggests which one wins over time -- and for the same underlying reason.
This mirrors the substrate argument made repeatedly in analysis of AI enterprise deployment: the capability layer is not the constraint. The organizational and data infrastructure beneath it determines who captures value and who produces expensive outputs that go nowhere.
What the Framework Implies for Investment Decisions
If the sixth pillar analysis is correct, it generates a specific prediction: companies that invest heavily in McKinsey's five AI marketing capabilities without simultaneously addressing organizational integration will see marketing ROI improve in the short term (because the tools are genuinely useful) and plateau or decline in the medium term (because the structural limitation reasserts itself as the binding constraint).
The short-term improvement will make the investment look justified. The medium-term plateau will be blamed on creative quality, channel mix, or macro conditions -- because those are the variables that marketing leaders have vocabulary and attribution tools for. The actual cause (organizational architecture) will remain invisible because it is outside the measurement perimeter of marketing analytics.
This is why brand trust research consistently finds that communication-layer interventions underperform expectations: the trust signal consumers are responding to is generated upstream of marketing, at the operational level. Investing in better marketing tools while leaving the operational connection broken is the equivalent of improving signal quality from an antenna that is not connected to anything.
What Integrated Organizations Should Actually Do
For organizations that have or are building genuine marketing integration, the McKinsey framework becomes a useful investment roadmap. The five pillars represent real leverage points when the organizational architecture allows marketing insight to flow into operational decisions.
The priority ordering changes, however. McKinsey presents the five pillars roughly as parallel capabilities. For integrated organizations, orchestration should come first -- because without the connective layer, adding insights, creativity, personalization, and agentic commerce capabilities creates isolated optimization rather than compounding improvement.
Orchestration in this context means not just channel coordination but decision coordination. The AI system that connects consumer behavior data to product development prioritization, pricing committee inputs, and distribution partner strategy is worth more than any number of personalization engines operating in a closed marketing loop.
The Actual Question for 2026
The AI marketing conversation in 2026 is still mostly about capability acquisition. Which tools, which platforms, which vendors. McKinsey's framework reflects that focus accurately.
The more interesting question -- and the one that Elf Beauty's results quietly raise -- is which organizations have the structural architecture to convert AI marketing capability into business outcomes that compound rather than optimize.
The answer has almost nothing to do with which AI tools they buy. It has almost everything to do with whether marketing operates as a function that delivers campaigns to the business, or as an integrated capability that participates in the decisions that campaigns are supposed to support.
AI does not solve for the difference between those two organizational models. It accelerates whatever model is already in place.
If you want to understand whether your own marketing infrastructure has the structural integration to capture the returns from AI investment, the research frameworks at STI offer a starting point for that diagnosis -- before the next capability acquisition makes the underlying architecture problem more expensive to address.