When I ask our AI-powered insights platform to search for news and media narratives around “AI” from the last three months, the platform aptly reports that: ‘the AI revolution is gaining momentum’ and is sparking ‘concern and excitement.’

From here, the platform enables us to click into each narrative (e.g. ‘momentum’ or ‘concern’) and see which news articles, or social media posts, form the story. We can then use the insights from those articles to form a new, more specific search across social media channels and news – for instance, a more targeted exploration of reviews of AI-based apps, or an investigation into people’s thoughts on the OpenAI commercial shown at the Superbowl. Artificial intelligence proves to be a fantastic tool for researching itself.

As a cultural consultancy, Sign Salad sees AI-expertise as integral to cultural expertise. While we use AI to help decode culture, we use brand-centric cultural insight to help clients understand ‘AI’ as a lived phenomenon. Understanding machine learning is increasingly essential for achieving brand relevance in a tech-mediated world.

How we think about Artificial Intelligence

Navigating AI might seem to require a constantly updated dictionary. When we talk about AI in insight, we’re typically talking about one of the below sub-categories:

Traditional AI: Also known as Narrow AI. Works as a set of specialised tools used for specific tasks (e.g. image analysis, data processing), and always requires human input.

Generative AI: The creative member of the AI family creates new content (e.g. text, images, music) by learning patterns from data and generating outputs that imitate, but do not replicate, human creativity.

Artificial General Intelligence (AGI): The most science fiction-adjacent AI. Artificial General Intelligence with human-like cognitive abilities, capable of understanding, learning, and applying knowledge across a wide range of tasks without specific programming. On the horizon according to some, but not a reality yet.

Across insight and marketing conferences dealing in AI knowledge, most research agencies emphasise the combination of “the human and the machine.” That combination sounds reassuring and rigorous, but as outlined in the above definitions, it’s also simply a practical reality. Even the most advanced AI tools require human guidance, interpretation, and oversight, and it’s well-known that the results we get from AI are as good as the prompts we write. However, those prompts still need to form straightforward, highly specific questions, and the researcher will need to be content with prompts that don’t require complex creative thinking on the part of the machine. AI-powered research models, like Open AI’s Deep Research, will often generate erroneous responses to questions that require a more lateral, probing way of thinking, as The Economist pointed out last week.

Of course, sometimes AI is spot on:

At Sign Salad, our approach is about combining artificial intelligence with human cultural intelligence. In terms of our methodology, we primarily use AI to enhance Social Data Analysis and Language Analysis, and we broadly categorise our approach as covering Topics (key words) and Panels (experts or specialists).

Topics analysis can entail organizing thousands of reviews or transcripts into themes, ‘topics’, and emotions. Alternatively, we might track certain keywords across Instagram and TikTok to see how breakfast is being depicted – one of our earlier Sign Salad articles shares findings from exactly that search.

Panels, on the other hand, can entail using AI to help us identify influential TikTok accounts or subreddits we should be tracking for a project (expert dermatologists, bartenders, automotive afficionados, etc.), forming a Panel of experts or specialists.

AI helps us deliver robust insight by identifying patterns, themes and narratives in large amounts of data, and we carefully cross-reference all of this with what we see in culture at large, ensuring each project accurately captures consumer realities and reflects the cultural changes beginning to surface. For instance, AI can tell us that “amaro” is trending and can group words used in relation to amaro to identify what people like about the herbal liqueur (e.g. depth of flavour, associations with laidback aperitivo moments, versatility).

However, AI alone can’t perform an analysis of the changing ways people are socializing, celebrating and rewarding themselves, from which to decipher what might come next in the cocktails category. AI gravitates to the median and enables us to see where we are right now very clearly, but not what we’re moving towards at the leading edge, in the fringes of culture.

A measured approach to machine learning

Sign Salad is also cautious and deliberately sceptical when assessing AI’s capabilities. At times, AI can be an industry buzzword. From a semiotic perspective: talking about the super-intelligence of a machine implies that the research itself is ‘super-intelligent’, which can be an alluring promise. At Sign Salad we find targeted, intentional uses for AI (as with any of our insight approaches). AI should never be used for the sake of it, the same way we wouldn’t force-fit depth interviews into a brief that didn’t merit them. The decision-making around how, when and if to use AI tools is about working backwards from the most fitting, ideal output to each brief. A brief seeking to understand football fan communities globally aligns perfectly with AI-powered social data analysis, enabling us to track forums and social media posts across regions and organize conversations and commentary into themes, likes, dislikes, and so on, as well as grouping. A semiotic analysis of the meanings communicated by 5 proposed pack design routes wouldn’t require our AI tools, however, because there’s no ‘data’ we need to pull in beyond in-house expertise.

Where AI is being used to replace the human in aspects of qual research, applications to cultural insight are less straightforward. We use AI-powered platforms to detect themes and narratives across news, social media and reviews at scale, but it’s vital to remember that vast amounts of culture and daily life don’t show up online. For every person who posts online about their snack routine, there’ll be thousands of people who don’t. So we test, validate and develop what we see on social channels by looking at life on the ground, on supermarket shelves and streets, and in the historic and immediate cultural contexts that AI is not attuned to. We uncover tangential links between categories, creating a deeper understanding of influences on consumer opinions, brand perceptions, and purchasing decisions. We are rigorously researching IRL, while the AI is quietly rigorous URL.

A cultural view of AI

The nebulous category of “AI” is itself a cultural and social object, one that tells us a great deal about where we are now. Artificial Intelligence has become a prism through which to view a vast range of issues, from geopolitical relations (see the announcement of DeepSeek), to creativity (with debates over human vs. AI ‘art’, intellectual property, and suggestions our cognitive function might be lessened by AI), to diversity, to equity and inclusion (with AI systems displaying racial bias and discrimination), to job security (see the rise of customer service chatbots). Google searches featuring the terms “AI” and “ChatGPT” have risen exponentially since c. 2021.

We can also discern a great deal about the cultural moment from where AI doesn’t show up. Gen Z have been championing analogue activities and craft clubs, and much of the leading edge of luxury continues to centre the promise of going ‘off the grid.’ The prominence of digital is evidenced in the resistance to digital. LEGO consumer sales grew by 14% last year, and one record label is going “100% vinyl” after a stark increase in vinyl sales. Brand success now depends on staying attuned to the cultural shifts happening offline, in the intentionally crafted non-digital spaces so many people are gravitating to. 

Key Takeaways

  • AI provides speed and scale, while humans provide context and creativity. We use AI to deliver rapid results – but we use cultural intelligence to delve into the messier, deeper, contextual and more lateral thinking across categories that tells clients not just what we’re seeing in the data, but why.

  • AI is changing at an extraordinary pace, and we need to meet its new capabilities with a careful assessment of which tools best answer each client’s questions. AI is not yet at the stage where it can challenge or push the questions it has been asked.

  • Technological advancement often sparks countercultural movements that seek to preserve experiences characterised by their tangible physicality, imperfection, and human touch.

  • AI can tell you what most people agree on and are already comfortable with. However, it can’t independently generate meaningful insights from the outliers and the fringes. The future is often not in what everybody thinks – but in what a small number of people are beginning to think.

 Recommended listening:

Katrina Russell, Director

Meeting Artificial Intelligence with Cultural Intelligence: How to use AI intelligently