How GitHub harnesses AI to transform customer feedback into action

At GitHub, customer feedback via support tickets has been continuously analysed using AI to better understand the user base and innovate. The company used the BERTopic open source AI model to generate high-quality document embeddings, interpret customer feedback, and create dynamic topics around various product features. The multilingual capability of BERTopic allowed for the analysis of global user feedback. However, the outputs needed to be easily understandable and made into actionable insights. GPT-4 provided a solution in summarising these topic clusters using its advanced natural language processing capabilities. The process enabled GitHub to deliver tailored customer feedback insights in an easily understandable form, which was incorporated into an internal web application. This transformed how GitHub interacted with and responded to its customer base.