Editor's note: Neil Dixit is the Founder and CEO of Glimpse and Adam Bai is the Chief Strategy Officer and Chief Client Officer of Glimpse.
From one point of view, the generative AI mania has already crested. But in the research, marketing and content creation spaces, the real drive for adoption at the team and organizational level is about to begin.
Organizations will soon begin making important decisions based on recommendations from large language models (LLMs). They’d better be sure that those recommendations are based on sound foundations and proven techniques. One problem: LLMs are simply not configured right out of the box to help researchers, marketers, advertisers and creators develop actionable insights or differentiated campaigns.
Why? Generative AI is a predictive technology and its training data sets in our industry are sometimes outdated or limited. Outdated because the training data isn’t updated frequently enough to meet the needs of professionals. Limited because training a model on publicly available internet chatter means ignoring all the opinions, emotions and behaviors that aren’t easily scrapable from public sites and platforms.
Without additional relevant first-party data to pull from, it can lead to unrepresentative recommendations, strategically misleading suggestions and even biased or discriminatory outcomes. When drawing on the immense power of LLMs, Glimpse always layers in high-quality, first-party, representative and real-time data, using a set of proprietary (and extensively tested!) techniques to build context around the prompt.
Our clients benefit from this added context as they use the platform to analyze data, create topics and nuanced summaries, generate key messaging (tailored to the needs of particular segments and personas) and generate personas to help them answer the questions that matter.
Client case
Wells Fargo Bank + National Foundation for Credit Counseling (NFCC)
Wells Fargo Bank and NFCC recently launched a study about housing insecurity to 2,000 American low- and middle-income renters.
The use of generative AI to analyze thousands of open-ended responses revealed that the dominant emotion associated with eviction was sadness. But it also revealed a strong undercurrent of optimism about financial recovery and housing ownership, particularly within communities of color.
Generative AI adoption tips
Here are four principles – based on a lot of Glimpse’s generative AI-powered client work – for any organization:
Focus on inputs and outputs.
Generative AI capabilities are shockingly impressive. But it’s important to remember that successful adoption will also depend on high-quality, representative training and testing data and the careful application of gen AI outputs to specific business or research challenges.
Think longitudinally (to track change over time).
In the age of generative AI, historical data is even more valuable than ever before. It allows us to train our models to become more nuanced and effective within the context of our own business challenges.
Strive for a holistic approach.
Generative AI can be applied to any data source to find patterns, spot opportunities or warning signs and help develop insights. It can help discover relationships between data sources, like social listening, first-party customer or sales data and survey data.
Ask the right questions about your teams, talent and processes. Starting right now!
- Are there urgent business or research challenges that generative AI can uniquely help us solve? Or are we following a trend without considering its value?
- Do we have the skills/capabilities/talent on our team right now to use generative AI tools effectively? Do we have people adept at providing context to generative AI processes and then intelligently applying its outputs to business challenges?
- Does process or team structure need to evolve? The answer is almost certainly, yes!