AI has been sold to marketers as both a threat and a miracle. The data tells a more useful, more boring story: it changes the economics of the work. Tasks that were once too slow or expensive to do well, deep research, dozens of creative variants, always-on analysis, are now cheap enough to do routinely. That shift is real and measurable. What it doesn’t change is who’s accountable for the decisions.
Adoption is near-universal. Results are not.
Roughly 91% of marketers now say they use AI in their work (Jasper’s 2026 State of Marketing AI), and McKinsey finds about two-thirds of organisations use generative AI regularly. But the same research exposes the gap: only around a third of companies have scaled AI across the business, and the share of marketers who can actually prove ROI from it has fallen year over year, not risen. Adoption went up; accountability went down.
The prize is real: McKinsey estimates generative AI could unlock $0.8 to $1.2 trillion in annual value in marketing and sales alone. Capturing any of it depends less on which tool you buy and more on whether you redesign the workflow around it.
What actually improves: the speed of learning
The biggest change isn’t quality, it’s cycle time. Research that took days takes hours; ten creative variations that were once a luxury become the default. Because marketing is fundamentally a learning system, the faster you put a real idea in front of a real audience, the faster you learn what works. Semrush reports that 68% of businesses have seen higher content marketing ROI from AI-enhanced workflows, and the operative word is enhanced, not automated.
Personalisation finally becomes practical
Tailoring messages to segments was always a sound strategy; almost no one had the capacity to actually produce and manage that many variations. AI removes the production bottleneck, letting you adapt messaging by industry, role, and buying stage without a proportional increase in headcount. This is where much of that marketing-and-sales value pool McKinsey describes actually sits.
The quality trap
Left unsupervised, AI drifts toward the average, the safe, the generic. Analyses of organic performance consistently find that human-led content still outperforms purely AI-generated content by a wide margin on traffic and engagement. The teams winning with AI treat it as AI-enhanced (human strategy and judgement, AI execution and scale) rather than AI-generated. The difference shows up in the results.
Where the human stays in charge
AI produces options, drafts, and signals; it does not carry accountability. The decision to move budget, the read on whether a bold creative will land or backfire, the judgement of brand fit, the client call when something breaks, these stay with people. Gartner predicts that by 2028 a majority of brands will use agentic AI in customer interactions, which makes human oversight more important, not less: agentic AI without strategic direction is just faster chaos.
The practical stance
Treat AI as leverage on your team’s judgement, not a replacement for it. Let it take the hours, the research, the drafts, the variants, the first-pass analysis, and keep the decisions with people who can explain and defend them. The campaigns that win with AI aren’t the most automated. They’re the ones where good judgement now gets applied far more often, because everything leading up to the decision got faster and cheaper.
