Generative AI is pushing beyond what we once thought possible in the creative realms of image and video. However, over the past 18 months, what has truly captured my interest is its ability to finally deliver on the long-held promise of personalisation.
Brands have been under a mountain of first party data but often, have been unable to analyse it, let alone derive insights from it. Now that we can lean on both Discriminative and Generative AI, we can process vast amounts of first-party data to pinpoint who our audiences really are, on an individual human level. Overlay that with other data sources like sentiment analysis from social feeds, and brands are on a path to understanding their fit in peoples’ lives in a deeper sense. We’re talking psychographics at scale.
These datasets and that enriched understanding might deliver eureka moments - but how do we act on them?
I believe these datasets are a cornerstone of what some are beginning to call ‘Brand World Models’ – or at least that’s what I’m calling them. These are multi-faceted, ever-growing knowledge graphs encompassing everything that makes a brand. They allow brands to understand and communicate with their audience at an unprecedented level.
Building ‘Brand World Models’
Aside from audience and sentiment data, the model also integrates a brand’s visual and tonal guidelines. This ensures that every piece of communication, from ads to social posts, aligns with the brand's established identity. This includes tone of voice, visual aesthetics, and even historical performance data. The model also contains assets such as past campaigns, style guides, and product-specific information - allowing the AI to execute creative that is not only personalised, but also consistently on-brand.
Additionally, ‘Brand World Models’ will house content strategy models that define when and how to communicate with audiences across various platforms. These models take into account preferred formats, optimal times of day, and even emotional triggers that will resonate most with each psychographic segment. Other components include competitive analysis, influencer alignment, and cultural localisation to ensure that messaging is not only personal but also contextually relevant.
In the first phase of their lives, they tell us where the fertile ground is and check that we’re on brand when we create for the brand.
Are we ready for brand self-execution?
As AI advances, the potential of these brand world models becomes even more exciting. We’re approaching a time when brand world models won’t just inform strategy - they will execute it. Imagine an admittedly simplified brief of “I want to launch Product X to my audience.” The AI model, powered by all the audience and creative data it has learned, will generate a full set of specified assets - ads, videos, social content - customised for each segment of the audience on a one-to-one level.
These assets are inherently brand-safe, because they’ve been generated from the brand guidelines. We can generate as many creative variations for different psychographic segments as we need.
The road ahead: building blocks in place
While we’re assembling the pieces of Brand World Models, we’re already making strides towards personalisation in other ways.
For example, in the Thai market, McDonald’s recently worked with a partner that analysed millions of social media posts and competitor creative to infer target audience profiles and messaging angles of those competitors. This ultimately revealed white space in the market that they could occupy for stand out.
By avoiding overused audience segments and instead focusing on long-tail interests, the brand was able to increase reach by over 400% while reducing CPM by 85%. This type of A/B testing at scale, combined with the model’s ability to identify new messaging opportunities, is where AI-driven media strategy proves its value.
What about distribution?
Personalised creativity is only half of the equation; the other half lies in distribution. As media shifts toward hyper-personalisation, traditional distribution models are being tested. So, how do we scale media to meet the resource demands of one-to-one personalisation?
By leveraging AI-powered bidding systems, we’re dynamically allocating ad spend across platforms based on the most effective channels and audience segments. This kind of real-time optimisation allows for scalable personalisation without skyrocketing costs. Performance Max might have its detractors but as with any current AI/ML tool, today’s version is the worst one we’ll see of it. There’s a near-future triangle of self-executional brand world models, next gen PMax and AI-driven content delivery networks to get the right content to the right person at the right time.
Looking ahead: the future of performance and distribution
The near future promises self-learning campaigns. We can already monitor and optimise performance in real-time, so it won’t be long before performance data is looped back into the ‘Brand World Model’ to adapt and iterate creative based on live feedback, refining executions in near real-time. Agencies will also employ predictive algorithms that anticipate consumer needs, ensuring that relevant and valuable content is delivered precisely when it’s needed.
The dream of delivering one-to-one personalisation is within reach. And maybe, we marketers will soon be focusing on higher-level tasks during our four-day workweek.
