The enthusiasm around generative artificial intelligence is undeniable, especially in marketing. However, the gap between imagination and real impact remains significant. The hype has its role: it mobilizes budgets, inspires new possibilities, and places the topic at the center of strategy. But, as with any innovation, the true value appears only when it stops being headlines and becomes operational. Recent data clearly shows this contrast. The adoption of GenAI is already widespread, but the impact on overall results is still rare. McKinsey points out that only 21% of companies using the technology have redesigned at least some workflows; unsurprisingly, over 80% report no tangible impact on the company’s overall EBIT.
In marketing, optimism remains high. About 80% of CMOs are confident, and 71% plan to invest around $10 million annually in GenAI over the next three years. Still, with budgets under pressure and limited to an average of 7.7% of revenue by 2025, the demand for return on investment will only get tougher. This scenario translates into a simple reading: we adopt fast, scale slowly, and measure little. McKinsey itself highlights that tracking GenAI KPIs is among the practices most correlated with impact, but fewer than one in five companies adopt this measure. As I often say, agents make headlines, but it’s the machinery that delivers margin.
This immaturity is evident in recurring signs. Companies with a multitude of pilots without a path to production, teams that don’t consistently measure ROI, and employees who bring their own AI tools from outside the corporate environment show that there’s still more talk than process. Only 1% of executives describe themselves as mature in GenAI deployment. Workflow reengineering is scarce, and AI usage often remains limited to producing cheap content at scale, without integration with CRM, CDP, or DAM—resulting in acceleration but no movement in business metrics. Add to this often-disconnected financial expectations. Some still believe that merely cutting agency or labor costs with AI will be enough. At the end of the day, the rule is simple: AI doesn’t save a disorganized company; it only scales what you already are.
What’s missing, then, for marketing—one of the areas most open to innovation—to strategically scale GenAI? The answer lies in discipline. The first step is structuring a true content supply chain, treating creation, approval, and distribution as an integrated factory, with clear guardrails, DAM, templates, and assured quality. The second is ensuring clean and consented data in a CDP, because without solid taxonomies and proprietary databases, the personalization promised by AI is just rhetoric. The third step is bringing LLMOps practices into the martech stack: standardized prompts, integration with official sources, observability, traceability, and human and automated evaluation. The fourth is end-to-end measurement, connecting impressions, clicks, costs, and results in revenue, margin, and retention. Fifth is investing in talent, creating new roles—from AI product owners to knowledge curators—and promoting continuous upskilling. Finally, the sixth point is abandoning vanity metrics and focusing on proven impact, with A/B testing, Marketing Mix Modeling econometric models, and isolated channel tests. Only then will marketing move beyond rhetoric and prove real business contribution.
But how to differentiate hype from applications that truly generate value? I usually ask four simple questions. The first is about materiality: which business KPI will be impacted, and to what magnitude? Vague gains, without numbers or deadlines, rarely scale. The second involves frequency: does the problem repeat enough to justify automation and investment? The next is about data advantage: do we have proprietary inputs and context that make the solution unique, or are we just using publicly available data anyone can access? The fourth question tackles the last mile: does the solution truly integrate with existing workflows and generate feedback, or is it just a nice demo that doesn’t connect with CRM or ERP? If the answers to these questions aren’t clear, the project is just hype.
In this scenario, the CIO’s role is essential. They aren’t responsible for choosing the best model but for designing the golden path: governance, platform, integration, and discipline to enable impact at scale. This involves Responsible AI policies, modular architecture to avoid the ‘tool zoo,’ evaluation standards tied to real KPIs, and last-mile integration. Without this, AI remains a proof of concept. Strategic realism must prevail over rhetorical optimism. The challenge is no longer proving that GenAI can transform marketing and business—it’s building the conditions for it to stop being a promise and truly deliver results.
