Forethought Quantifies the Benefits of Using AI Trained on a Company's Data

Generative AI, or GenAI, is changing the very foundation of the customer support industry. Often implemented to save money and manpower, the technology also has the potential to improve the customer experience and personalize interactions for customers.

But is it working?

According to a new study, the rapid adoption of GenAI in the contact center is having a positive impact on customer support. Forethought, a provider of solutions for generative AI agent for customer support, recently released its first AI in CX Benchmark Report to highlight the various ways companies leverage AI for customer service and the results they’re achieving. The report found that 54 percent of businesses surveyed — including both public and private B2B and B2C companies — have adopted AI in some way, with more planning to do so, signaling that AI is not only here to stay but also being adopted at a rapid pace.

The study found that businesses that implement AI the right way can see immediate results and value for their customers. So…what’s the right way? Forethought has found that bots using AI and retrieval-augmented generation (RAG) trained only on help center or public data perform worse than AI trained on a companies’ own deep data: AI solutions trained on a company’s own historic data posted an NPS of 29 while AI solutions trained on other types of data earned an NPS of -18.

In short, organizations that train AI on their own historic data are nearly 3.5-times more likely to lower cost per resolution, according to the research. At the same time, the study found that dedicated solutions are more effective than help desk add-ons: the report found that the highest overall average deflection rate was nearly double (38 percent) for companies using a dedicated AI point solution for CX compared to companies using a help desk add-on.

The report surveyed mid-market and enterprise, U.S.-based companies and found that the businesses that have adopted AI for CX in the most advanced and effective way have nearly doubled their ticket deflection rate and are three times more likely to report lower costs compared to companies not using AI at all.