In addition to generating personalized advertising content, marketing professionals are using large language models as new personas to shape brand perception and strategy.
CAMBRIDGE, Mass. -- 2024 has been the year of turning AI pilot initiatives into measurable ROI. At MIT Technology Review's EmTech MIT 2024, the potential benefits and challenges of integrating AI were top of mind in many sessions.
From integrating AI into productivity tools like Slack to implementing AI-driven marketing strategies, AI is changing how businesses respond to both old and new problems. But leaders agreed that most businesses are still in the beginning stages of AI adoption, and successful initiatives will require understanding company identity and pinpointing specific use cases to address.
Using AI for more personalized, cost-effective marketing
Rebecca Sykes, a partner at marketing technology company The Brandtech Group, discussed how companies can use emerging technologies like AI in their marketing strategies. In her session, "AI-Driven Marketing in a Post-Advertising World," she highlighted the industry's shift toward personalized advertising, where people-driven campaigns help brands reach consumers in more authentic and targeted ways.
"Traditional advertising methods -- for example, TV ads, billboards and interactive ad formats -- are becoming less effective, or in some cases, even obsolete," Sykes said in her session.
In the context of this shift away from traditional advertising, generative AI offers new avenues for marketing, Sykes told TechTarget Editorial in an interview.
Generative AI can improve efficiency by automating previously manual marketing tasks, such as creating copy, images and video. It can also enable marketers to produce content that would have once been too expensive to create, such as AI-generated backgrounds showing scenes that would be too costly to shoot in real life.
Marketing firms are taking note. In a 2024 report from the Marketing AI Institute, 66% of surveyed marketers said AI was very or critically important to their marketing success over the next 12 months, and 51% were piloting or scaling AI efforts.
Understanding what these models think about you and your brand and how they perceive you is really, really important.
Rebecca SykesPartner, The Brandtech Group
"[Generative AI can] change everything about the way that we make content -- how long it takes, how much it costs, how personalized it can be," Sykes said.
Using LLM personas to understand brand perception
Treating large language models as a target audience can also be a useful tool, Sykes explained during her session.
Because LLMs are trained on vast amounts of internet data, they have become some of the biggest consumers of internet content. Sykes pointed to four major foundation models that can serve as proxies for humans' internet use: Google's Gemini, Meta's Llama, Amazon's Rufus and OpenAI's GPT models. For example, she said, Llama's preferential access to training data from Facebook and Instagram makes it a strong indicator of users' social media activity.
"Understanding what these models think about you and your brand and how they perceive you is really, really important," Sykes said. "The difference between the model recommending your shampoo and somebody else's is going to be a factor of what they understand about you based on the content they've consumed."
In an analysis of which beer brands are top of mind for LLMs, Rebecca Sykes explained how evaluating share of model helped identify LLMs' negative associations with Bud Light to inform future brand messaging.
Sykes referred to this metric as share of model, similar to the concept of share of mind in traditional marketing. Tracking share of model involves repeatedly querying LLMs about certain products and brands, and recording their responses over time. Those responses are then converted into vector embeddings -- groupings of similar themes or phrases -- to help companies understand how LLMs associate positive or negative perceptions with their products or brand.
By analyzing these associations in light of the specific model and its training data, brands can refine their marketing strategies. For instance, Sykes explained, if Gemini associates Bud Light with being "unremarkable," Bud Light could invest in a YouTube Shorts campaign focused on memorable moments, aiming to take advantage of Gemini's access to YouTube data.
"You can ... specifically target a particular perception change in a particular channel to try and change that overall association," Sykes said. Analyzing feedback from this new audience -- LLMs -- can help companies identify gaps and trends in how their brand is perceived.
Finding the right use case for AI
Slack CEO Denise Dresser discussed productivity use cases for AI in her session, "Unlocking Team Productivity With AI at Scale." Capabilities like searching and summarization target common problems that knowledge workers face, she said.
"The way we should think about AI in general -- whether you're struggling to think about how should I bring AI in my organization, or what are the best use cases -- [is] really listening to the use case and the problem to be solved," Dresser said.
In an interview with TechTarget Editorial, Sykes echoed this sentiment when asked if AI-driven marketing is for everyone. "If you can, you always should, but choose your use cases," she said.
"You have to put yourself back in the position of a consumer," she added. What types of AI-generated content would you be comfortable -- or uncomfortable -- with?
Finding the right use case and maintaining transparency in subsequent AI use should be guided by responsible AI principles, Sykes said. Examples of responsible AI in practice include always keeping a human in the loop, taking steps to mitigate bias and watermarking AI-generated content.
Although AI regulation is still evolving, staying ahead of potential requirements benefits both companies and consumers, Sykes said. Responsible AI frameworks and technical training can help ensure that AI-driven marketing initiatives successfully connect with consumers.
Lengths to go
Peter Weill, chairman and senior research scientist at the MIT Center for Information Systems Research (CISR), discussed AI adoption strategies in his session, "Business Impacts: Generative AI." In a recent study of over 750 companies, Weill and his MIT CISR colleague Stephanie Warner found four major stages of AI adoption.
In his session, Peter Weill referenced research performed by MIT CISR that identified where companies fell in four main stages of AI adoption.
Most businesses found themselves in the first and second stages, reflecting what many business leaders noted throughout the conference: Despite lofty goals, AI integration is in its infancy for many companies.
However, in terms of financial success, Weill noted that companies in the third and fourth stages tend to outperform industry averages. While it's impossible to prove causality, the data suggests that more advanced AI integration often goes hand in hand with stronger performance and business success.
The correlation goes both ways, Weill explained: "Companies that use AI more effectively seem to perform better. The companies that perform better are good at lots of things. So, they're good at taking up the new technology. They have good cultures; they have good data. They have ways to stop people worrying about their jobs -- I think those [factors] are all wrapped up in that data set."
In any industry, AI initiatives can serve as a magnifying glass, drawing attention to a business's strengths as well as its weaknesses. For companies in a good state of health, Sykes said, AI integration can drive positive outcomes. But for those in poorer shape, adding AI might only scale existing issues.
Olivia Wisbey is associate site editor for TechTarget Enterprise AI. She graduated with Bachelor of Arts degrees in English literature and political science from Colgate University, where she served as a peer writing consultant at the university's Writing and Speaking Center.
