Aided by powerful new AI tools, marketing has never moved faster. Campaigns are launched in days, budgets shift in real time, and dashboards promise instant clarity. But beneath this acceleration lies an uncomfortable possibility: AI may be making marketing decisions worse, and many CMOs won’t realise it until the quarter is over and the spend is gone.
Similar scenarios, higher stakes
We’ve seen this pattern before. When generative AI first entered the marketing toolkit, it looked like an obvious win. Teams that once spent days crafting copy, scripts, and creative concepts could now produce them in seconds. Output surged. Efficiency improved. On the surface, it felt transformative.
Then the cracks appeared. Content started to feel repetitive. Ideas lacked depth. And, more critically, hallucinations exposed a fundamental issue: AI could produce polished, confident outputs even when the underlying inputs were flawed.
You might expect that lesson to have carried over as AI expanded into more consequential areas of marketing. But the same dynamic is now playing out again. And this time, it’s in decision-making itself.
Precision without accuracy
Today, AI is increasingly embedded in analytics, attribution, and media optimisation. It promises to turn vast volumes of data into clear guidance around where to spend, what to prioritise, and which channels are driving growth. The outputs are polished, precise, and delivered at speed. But precision is not the same as accuracy.
Just like in content generation, AI in decision-making can produce compelling answers based on incomplete or biased inputs. The difference is that now, the stakes are far higher. Instead of a weak headline or a generic campaign, the consequences show up in misallocated budgets, missed opportunities, and underperforming growth strategies. And because the outputs look authoritative, teams are more likely to trust them, and act on them with conviction.
From false confidence to eventual failure
This is where the real risk lies. AI doesn’t just make poor decisions possible. It makes them scalable. It creates a false sense of certainty at exactly the moment when scrutiny is needed most. By the time results fail to meet expectations, those decisions have already been operationalised, budgets committed, and strategies locked in.
It’s tempting to see this as a failure of AI. But in reality, AI is simply doing what it is designed to do. It’s identifying patterns and making inferences based on the data it’s given. The real issue sits beneath it, in something far less discussed but far more consequential — measurement.
Measurement mismatch, and modern marketing’s shaky foundations
For years, marketing measurement has struggled to keep pace with how customers actually behave. Many systems still reflect the simpler era of web-first journeys, fewer channels, clearer signals. But today’s reality is fundamentally different.
Customer journeys are fragmented across mobile apps, websites, connected TV, retail media networks, and offline interactions. They are non-linear, dynamic, and often unpredictable. At the same time, privacy changes have reduced the availability of deterministic signals, forcing greater reliance on modelling and assumptions.
In this environment, measurement systems often present a version of reality that looks complete on the surface but is riddled with blind spots underneath. Conversions appear neatly attributed. Journeys seem linear. Performance looks optimised. But much of this is inferred rather than observed. And that’s the data AI is being asked to act on.
Instead of correcting these gaps, AI works to amplify them. When key signals are missing, it fills in the blanks with assumptions. Over time, those assumptions become embedded in recommendations, which in turn shape budget allocation and strategy. The system begins to reinforce itself. The result is a dangerous feedback loop: AI optimises not for true performance, but for what is most easily measured.
Why mobile should be the anchor
One of the most overlooked aspects of this problem is where measurement is anchored. For most consumers, the centre of gravity is mobile. It’s where discovery happens, where engagement deepens, and where intent is most clearly expressed (even if the final transaction occurs elsewhere).
Yet many organisations still treat mobile as just another channel, rather than the connective layer that links the entire customer journey. Measurement frameworks built on web-era assumptions are retrofitted onto mobile environments, rather than designed for them. The consequence is a fragmented view of behaviour that misses critical signals.
Without a strong, reliable anchor that holds up under the constraints of privacy and reflects real user behaviour, omnichannel measurement becomes a patchwork of proxies. And when AI is layered on top of that patchwork, it delivers faster optimisation, not necessarily better optimisation.
The CMO inflection point
This is a critical moment for CMOs. The conversation around AI in marketing has largely focused on tools, capabilities, and use cases. But the more important question is far simpler: can you trust the data that is driving your decisions? If the answer is unclear, then scaling AI will only accelerate uncertainty.
The priority, then, is not to adopt more AI, but to sequence its adoption correctly. That starts with rebuilding measurement as a foundation, not a reporting layer. It means identifying where blind spots exist across channels and devices, and distinguishing between observed behaviour and modelled assumptions. It means designing systems that reflect how customers actually move. Only then can AI fulfil its promise as a true accelerator of growth.
(Sarah Maina is the Regional Manager, Middle East & France, AppsFlyer)
