Unless you spent 2023 befriending bugs under a rock, you know that marketing is undergoing a seismic transformation, courtesy of AI.
Growing mountains of content already either scream hoarse about the AI takeover or swear miracles on slashed-to-ribbons budgets.
But perhaps a sense of cautious excitement is more apt.
“Moving forward, I don't think any major tech company is going to get away with not having some sort of AI built into their engine. They're just going to get left behind.” —Rachel Bien, SVP of Omnichannel Media Architecture at Assembly Global
AI’s true transformative extent is still evolving. Its adoption demands careful contemplation of data quality, regulatory frameworks, and ethical and social considerations.
At the same time, it’s already changed how we as audiences consume and share media and content, and how marketers should use this data to make decisions.
Marketers can no longer avoid integrating AI in their audience-building processes, even as they hold tight to human connection.
But how and where should AI integrate into the audience building process?
And what guardrails need to be put in place to optimize performance and mitigate risk?
Let’s dive into AI use cases for audience segmentation, personalization, and engagement.
And along the way, I’ll touch upon some potential pitfalls to sidestep through human intervention.
There’s just too much data.
Big data revolutionized the game. A white-hot explosion of complex information expands and accelerates every second from diverse sources: website analytics, social media data, mobile app usage, search, transaction and location-based data, and so much more.
Traditional methods of data collection, analysis, and insight-gathering simply cannot handle this onslaught by themselves.
By the traditional approach, I mean a reliance on manual processes and human scrutiny to conduct market research, customer profiling, and demographic and behavioral analysis.
A white-hot explosion of complex information expands and accelerates every second from diverse sources: website analytics, social media data, mobile app usage, search, transaction and location-based data, and so much more.
Of course, plenty of non-AI software tools help with different aspects of traditional audience building, but it remains impossible to parse these unfathomable volumes of data without AI. An attempt to do so could mean:
The inverse of these issues is integrating AI to process data and find patterns and correlations for effective audience-building.
That said, there’s an important caveat: Audiences are human. Building audience relationships is an intensely complex human process. Outsource this entirely to AI at the risk of depersonalizing communication and losing meaningful connections.
The trick is to leverage AI-driven data analysis and insights to produce audience building strategies that resonate on an emotional, personal level.
AI’s analytical power lends itself to crafting empathetic and engaging messaging that resonates with the target audience, with precision. Depending solely on AI to generate soulless communication and engagement is…well, a waste of AI’s capabilities.
The trick is to leverage AI-driven data analysis and insights to produce audience building strategies that resonate on an emotional, personal level.
Marketers can benefit from expediting endless research and analysis work with AI, and allocate time and resources towards building strategies that resonate with their clients.
Perhaps the biggest use case for AI is creating unique segments based on the data—behavioral or otherwise—that these tools scrape. So says Rachel Bien, Assembly Global’s Senior Vice President of Omnichannel Media Architecture.
According to Bien, AI algorithms can handle complicated behavioral data to make it more structured and predictive.
She points to consumer prediction specialist Foresight Factory, which uses AI capability to create contextual clusters of data points that are scraped from millions of websites, news feeds, and articles.
These predictive clusters can then be used to group together potential audience segments based on mutual interests or behaviors rather than just demographic data.
Rephrasely founder Matthew Ramirez concurs with Bien’s assessment of this use case. His company’s brand team uses artificial intelligence to create models that analyze its audience’s interests and past behavior. It then uses these insights to classify them into different groups.
“AI helps us identify which customers are most likely to use our services and refer us to others,” says Ramirez.
“AI-enhanced processes assist us in finding new groups of customers that we might not have thought of previously. We can utilize AI to locate customers who are likely to be valuable in the long run, even though they are currently inactive. This enables us to figure out strategies to reconnect with these customers and get them involved again.”
Ramirez notes how this predictive analysis allows for smarter audience segmentation, which in turn helps the team create customized experiences that draw in those segments and encourage higher sign-up rates.
It points to AI’s immense capacity for granularity. By combining and analyzing huge volumes of data in all its forms, from behavioral to demographic, AI can develop sophisticated audience subcategories that offer more insightful views of the customer.
AI algorithms and machine learning can help marketers tailor product or content recommendations for audiences, and even distribute that content on the channels they frequent the most.
Walmart’s machine learning models help create personalized recommendations for each user, according to a source I spoke with from their data science team. The team uses their historical data of customers interacting with different aspects of the website to develop affinity models.
Data points include actions like clicks, add-to-cart, purchasing – these are all signals that indicate your affinities. Does a user have a propensity for a Samsung TV versus an LG? Do they prefer expensive items or go for bulk buys? The data science team builds and trains models to make predictions based on this data, and then use this to make individualized purchase recommendations.
Di5rupt and YOOF recently published a collaborative report on how AI-driven personalization influences the way audiences engage with news content.
The report examines how news organizations can leverage AI to curate sophisticated news feeds that offer customized updates to individuals based on their interests and preferences. It also explores the possibility of AI chatbots that deliver curated rundowns of current affairs, events and editorials every day.
Through this sort of data-driven personalization, news organizations can enhance user experience by only providing relevant information and thus reducing information overload.
AI insights could help companies execute insightful engagement strategies that reduce customer churn.
For instance, online retailer allbeauty used predictive analytics and AI to determine which customers and prospects within their database are likeliest to churn in 3 months.
Identifying this let allbeauty optimize a campaign to re-activate and convert these customers. This data-driven decision-making raised sales by a whopping 414.6% and revenue by 518.7%. All without even updating the content of their existing email messaging!
Besides predictive analytics, AI can also help reduce churn by identifying trigger events and analyzing customer sentiment—when does a customer leave a business? Is it after poor customer service interactions or price hikes?
Companies can use AI-driven social listening to improve their understanding of dissatisfaction. And they can incorporate those insights in churn-combating strategies, like training customer service or modifying price points.
According to Rachel Bien, the greatest limitation of AI is validation.
As she asks, “Are we just taking the word of this technology that it actually interpreted a data point in the right way and in the right relevant context?”
Ensuring the reliability of a new model is a tricky ask, especially since the model is only ever as good as the data it’s trained on.
Bien recollects when Meta replicated racial biases in its model and ended up deprioritizing black and African-American and minority groups. And she worries about engines that “scarily tell you what you want to…because if I'm building an audience, I want to be accurate and effective.”
The role of human intervention may therefore be to produce the necessary checks and balances so the model does what it’s intended to do, she says—deciding what data points to prioritize, removing as much assumption and bias from the data set as possible, offering the right human prompts.
It’s also to continually test a new model or audience. Bien cautions against a set it and forget it approach, advising marketers to devise a test structure ahead of time so it’s possible to isolate variables in the marketing plan. Test the data to measure results against expectations.
A final piece of advice from Bien: Don’t take the technology for granted.
AI has real implications on the way we serve our clients, use their data and do business. As marketing professionals, secure the guardrails in the technology to do right by your clients. Understand how the technology impacts the way you buy audiences on Meta or create them on your audience builder platform.
Do the due diligence.
