An Evolution in Needs-Based Innovation
In a groundbreaking exploration, Applied Marketing Science and Dr. Artem Tomaschenko use a custom-trained generative AI model to revolutionize voice-of-customer (VOC) analysis, making it faster and more accurate while preserving the nuanced insights essential for innovation.
“It takes as much time to solve a bad problem as it does a good problem…Customer needs, or jobs to be done, really is the way to make sure that you have that appropriate focus.”
Actionable Takeaways:
- Prioritize needs, not solutions: Focus on understanding the core need, rather than prematurely adopting technologies.
- Enhance with AI: Leverage fine-tuned AI for complex VOC data analysis, improving efficiency without sacrificing quality.
- Iterate and evolve: VOC and AI partnerships are ongoing; regularly refine the model to stay aligned with evolving customer needs.
A New Chapter in VOC Analysis: Beyond the Surface of Customer Needs
The world of VOC research has been transformed by the dynamic entrance of generative AI—specifically, by Applied Marketing Science and Dr. Artem Tomaschenko’s custom-trained model. Kristen Corrigan, an experienced VOC practitioner, shared how this customized GPT model not only captures but also interprets nuanced customer feedback, a feat traditionally handled by human analysts with extensive training.
The excitement in the room was palpable as Corrigan recounted her early inspirations for the project, drawing parallels to the pitfalls of past innovations. “The Segway, for example,” she noted with a smile, “while a technological marvel, didn’t fulfill an unmet need.” Through this anecdote, she reminded innovators of the importance of focusing on needs rather than letting the allure of technology lead the way.
A Deep Dive into Generative AI’s Potential
Corrigan outlined how the journey began: her firm’s desire to understand what a tailored generative AI model could do for VOC. VOC analysis has long been a time-consuming, highly detailed process, with human analysts painstakingly reviewing customer interviews, reviews, and social media comments. Their aim? To extract the “jobs to be done”—that is, the fundamental needs underlying customer comments. But could AI really capture the subtleties that experts spent years perfecting?
This was the question that launched Corrigan and Dr. Tomaschenko’s journey. Initially, they tried using out-of-the-box ChatGPT for VOC analysis, with limited success. While ChatGPT could recognize broad topics in categories like oral care, it missed the deeper needs. The researchers found that the model often presented superficial interpretations of the customers’ concerns, lacking the detail required to truly serve innovation efforts.
“We realized that a custom solution was needed,” said Corrigan. She and Dr. Tomaschenko set out to train a model using decades’ worth of VOC data. The result was a tailored GPT capable of drawing deeper insights, focusing not just on what customers explicitly said but also on what they implied or assumed.
From Functional to Nuanced Needs: A Case of Wood Stain Products
To illustrate the model’s accuracy, Corrigan highlighted a specific case study involving wood stain products. Using feedback from product reviews, the model identified not only that customers wanted an even application of stain without brush strokes but that they needed to know how to sand surfaces effectively without damaging previous layers. This example underscored the model’s ability to read “between the lines,” discerning functional needs and latent desires that are crucial for product developers.
An experiment confirmed the AI’s capacity to produce VOC insights that met human-level detail and accuracy. In fact, when compared to a team of human analysts, the AI model outperformed or matched their results on several metrics. Not only did it excel in identifying actionable insights, but it also surfaced rare, infrequently mentioned needs that human analysts might overlook. This was a game-changer; AI could provide a comprehensive overview while still allowing analysts to concentrate on strategic implementation.
Challenges and Future Potential: Building a Dynamic VOC Tool
Yet, Corrigan was candid about the challenges in training this model. Building the custom GPT required ongoing refinement to ensure that its interpretations aligned with customer language and sentiment. “This isn’t a one-and-done effort,” she said, emphasizing that they expect to continue fine-tuning and evolving the model as new customer insights arise. The project, which took over a year to get off the ground, was as much about creating a sustainable framework for continuous learning as it was about the AI’s initial performance.
This proactive mindset is crucial, she argued, given the rapidly evolving landscape of customer expectations and generative AI technology itself. By continually feeding new data into the model and refining its processes, they aim to keep it relevant and accurate. The ongoing collaboration between Applied Marketing Science and the Kellogg School of Management reflects a commitment to understanding and addressing these complexities in VOC analysis.
Where AI Augments Human Intuition, Not Replaces It
While some in the room raised concerns about AI replacing human analysts, Corrigan offered a reassuring perspective. “AI can’t replace empathy,” she said. The model serves as a valuable tool for speeding up the initial data analysis, allowing analysts to bypass time-consuming tasks and focus on strategic thinking. Rather than taking jobs, this custom GPT model complements human intuition and empathy, expanding the possibilities of what VOC can accomplish.
Corrigan ended with a preview of the next steps. The team’s current focus is on refining the model’s deduplication process and categorizing needs into actionable hierarchies. They are also exploring ways to develop an automated prioritization of needs—a traditionally manual, time-intensive task. Although this part of the project is still in its infancy, Corrigan expressed optimism about its potential to further elevate the efficiency and depth of VOC research.
AI’s Role in Customer-Centric Innovation
In sum, this session showcased the transformative potential of custom generative AI models in VOC research. By combining AI’s computational power with human insight, Corrigan and Dr. Tomaschenko are opening up new avenues for understanding and meeting customer needs. This evolution in VOC analysis underscores a powerful shift: AI is not just a tool for automation but an enabler of deeper, more informed customer engagement and innovation.