The Sales team worked hard to have customers to renew their digital subscriptions every year, but winning isn't guaranteed. Failing to get a renewal resulted in filling out a dreaded Did Not Renew (DNR) document with specific things the customer said about our products and their decision criteria.
How could we use GenAI to understand this customer feedback, from different regions, different industries and captured in a variety of digital formats to gain insights that might help revive these "lost" accounts in the following year?
Twenty DNRs were received in a variety of formats and with many variations in the quality and completeness, but it was data direct from the customer, so we believed we could still curate and extract value.
After converting these to PDF format I then uploaded them to Arches AI, a platform that accepted unstructured PDF files to create a "vector database"
so that we could later interrogate an "agent" knowledgebase. After formal testing of results to identify "hallucinations", we felt we achieved a high confidence level. In testing, the agents were able to state "I don't have enough data to answer that..." to our satisfaction.
The Sales team saw future value for the proof-of-concept. Some important insights were unlocked because of fact patterns found across global regions that reflected the infrequent communication between Sales teams. Many liked the ability to get easy-to-read recommendations and get actionable product details to give to the Product Team. The POC showed the Sales team an easy and effective way to get untapped value from these otherwise discarded DNR reports across global regions.
Adobe Acrobat, Arches AI
