According to recent McKinsey insight, while 90 percent of CMOs are currently experimenting with AI use cases, less than 10% have either scaled it or captured value across marketing workflows. Digital marketing specialist, Incubeta, however, says CMOs are facing growing pressure to scale AI initiatives, often without the necessary strategic, data, or measurement foundations in place.
In many organisations, especially in sectors like financial services and retail, AI has moved to a board-level mandate, and CMOs are being asked to do more with their AI tools while still hitting short-term performance targets.
However, the gap between that pressure and the organisation’s readiness often shows up as operational paralysis. Leaders worry about disrupting what already works, introducing new risks and being unable to clearly prove the impact of AI-driven change.
That anxiety is amplified by how AI is often sold, with too many conversations treating it as a magical elixir.
“The first and biggest misconception for me is that AI is just one thing, like it’s a single lever that you pull,” says Courtney van Zyl, Client Solutions Partner, at Incubeta. “In reality, it cuts across creative generation, media bidding, forecasting, audience modelling and more – each with different risks, data requirements and pay‑offs. When brands adopt it as a category rather than asking which specific problem it’s solving for them, disappointment is almost guaranteed.”
From monolithic AI to specific use cases
According to Calvin van Rensburg, Client Solutions Partner at Incubeta, the correct approach is to start with the problems, not the platforms. He explains that Incubeta’s work is organised around measurement, creative, and media, with AI and machine learning weaved across the solutions rather than sold as a standalone cure-all product.
“On the media side, that looks like tools that ingest performance data and use machine learning to decide when to intervene with bid or budget changes. On the creative side, it means using AI to generate and iterate assets at a speed that finally matches digital media’s always‑on cadence,” he says.
Where performance gains actually show up
Van Zyl says the performance gains of this approach don’t live in a single channel or a single tool.
“The most consistent gains we’re seeing aren’t concentrated in one single channel or function. They’re showing up wherever we replace the guesswork with the better signal. That might mean smarter bidding, more context‑relevant creative, or sharper audience definitions. It’s clear that AI works best where it upgrades the signals feeding decisions and where teams have the discipline to test whether that upgrade is real,” she explains.
That discipline is becoming more important as attribution gets messier. Privacy regulation, walled gardens and AI‑driven optimisation inside platforms have made customer journeys harder to observe end‑to‑end. Van Zyl argues that attribution is shifting from something that could be measured with some certainty, to something that now has to be modelled with confidence.
Incubeta’s answer is triangulation: Media mix modelling (MMM) for long‑term, cross‑channel impact; day‑to‑day analytics for operational optimisation; and structured experiments in areas like creative and predictive audiences. Van Zyl says the goal now is measurement you can trust directionally, act on, and stand behind – confidence you can defend, not certainty you can’t.”
Keeping things private
For CMOs in privacy‑heavy sectors, like financial services, data is as important as measurement when it comes to using AI. Van Rensburg explains that clients only ever share information that has been properly consented, and that his team is clear upfront about the specific data they need and whether the client has permission to provide it. These signals are securely hashed or encrypted to platform standards and anonymised beyond typical cookie levels.
However, both are clear that the bigger constraint inside large brands is rarely data volume. Financial institutions in particular often hold a lot of data that isn’t necessarily aggregated or put together in a manner that speaks between the operational silos. “Governance and hygiene, including connecting, cleaning and mapping datasets to defined use cases, is the real foundation,” Van Zyl cautions.
AI as an extension, not replacement
While many teams still treat AI as a magic replacement for human work, it functions far better as an add‑on than a substitute. It still needs human judgement to shape and refine the output, and there are hard cost limits on how much can realistically be automated.
Both specialists agree that the real advantage lies in how well organisations structure their data, processes, and decision‑making around AI, and in using it selectively where it genuinely adds value rather than deploying it everywhere for the sake of appearances.
And the benefits of the correct approach could seriously move the needle for many organisations. As McKinsey points out, “Get it right, and our experience with multiple companies indicates that 4 to 7 percent revenue growth, two- to threefold improvements in productivity, and 60 to 70 percent savings in execution-related tasks are possible with AI.”