The Center for Data Innovation spoke with Alyona Medelyan, co-founder and CEO of Thematic, a customer feedback analytics platform based in San Francisco. Medelyan highlighted the importance of customer feedback and discussed the opportunities and challenges of using generative AI in feedback analysis.
Morgan Stevens: What motivated you to start Thematic?
Alyona Medelyan: I was helping different companies to find natural language processing (NLP) solutions. They would come to me with all kinds of problems like finding the core part of an email or mapping wine names to a database. At some point, three companies reached out with the same problem. They wanted to understand what’s driving their Net Promoter Score (NPS). NPS is a customer loyalty metric. Customer feedback has clues into what people like or dislike. I decided to start Thematic to help companies understand how they can improve their products and services. They already have the data, they just need an effective way of making sense of it.
Stevens: How does Thematic leverage AI to analyze customer feedback?
Medelyan: We have developed our own AI approach to discover a set of themes custom to any dataset we analyze. Other solutions focus on specific industries with pre-canned taxonomies that they tailor. We believe that AI should tell you things you don’t already know about your data. And not all companies fit into a particular industry. For example, in tech, every product is unique. And new products and solutions are invented all the time.
Another unique thing about Thematic is that we built an interface that shows customers how AI created themes and lets them change things. For example, users can re-organize and rename themes or ask the AI to discover more themes.
These are just some examples. AI is used in many other parts of our solution.
Stevens: How has the proliferation of generative AI reshaped the landscape of customer feedback analysis and insights?
Medelyan: There are two main challenges with generative AI when it comes to analysis of feedback. It’s still cost prohibitive to use it to analyze all feedback. It also tends to grab onto specific bits of feedback to create plausible but incorrect and inconsistent analysis.
That said, it’s incredibly powerful for understanding natural language queries and summarizing data. So many solutions, including Thematic, are busy integrating AI into theirs. We’ve launched several features powered by generative AI and the feedback has been very positive.
Stevens: What are the biggest challenges to analyzing unstructured feedback data, and how does Thematic address them?
Medelyan: One of the challenges is that language is hard! People express the same thing in a variety of ways. We use language models to decipher the meaning and our unique AI to create themes from data. Also, finding insights in feedback is like finding a needle in a haystack. We use AI to categorize sentences into Issue, Request, or Question. This filters out generic non-actionable feedback.
Stevens: What does the future look like for Thematic?
Medelyan: We are on a mission to democratize insights from feedback. We believe everyone should be making decisions with the customer in mind. This has been difficult for many years because unstructured feedback is so hard to analyze. Companies have been using crutches such as structured surveys with 20 rating questions. These are horrible for customers, return generic insights, and will disappear in the future. The importance of unstructured feedback and utilizing contact center data will continue to increase. Another way in which companies used to solve the challenges of analyzing feedback is by making a central insights team responsible for this task. This team analyzed the data accurately, but could only spoon-feed limited insights back to the business due to resource constraints. We believe that the future of customer insights is in user-friendly AI-powered solutions like ours. We are passionate about breaking down the insights silos, letting anyone in the company get answers from feedback they need any time. This also means that the insights teams can focus on high value work, elevating their impact.
