CFC adopts large language models for 'everyday workflow'

CFC adopts large language models for ‘everyday workflow’ | Insurance Business Asia

Insurance News

CFC adopts large language models for ‘everyday workflow’

It puts faith in machine learning

Insurance News

By
Jen Frost

CFC is using large language models, as powered by machine learning, to triage broker emails, CFC CEO Graeme Newman has said.

“Those models are taking all the wonderful emails that you send us, reading them, interpreting them working out, therefore, how we triage that most effectively, where we need to send them,” Newman told broker attendees at the 2023 CFC Summit. “Those models increasingly can read the context of those emails and work out what the coverage requests are, what the limit requests are, and what the client details are.”

Using the models and its APIs, the underwriter has sought to make the process more efficient, regardless of the channel the broker initially used, according to the CEO.

Newman further predicted that conversational artificial intelligence (AI) could help it further crack other markets and overcome language barriers. While CFC trades in 80 countries around the world, the CEO acknowledged that the business is “largely concentrated” in the four major English language speaking countries.

“That’s because language is problematic,” Newman said. “We have that problem in Quebec right now [and] we want to get better at it.”

Google Translate never quite fit the bill in terms of responses and can be “a bit clunky”, Newman said. Looking forwards, he appeared optimistic about the opportunities that conversational AI could open up for the business in non-English speaking markets.

“If you look at what conversational AI can do right now, it can create perfect natural language translation, which means that, ultimately, we can open access to many new markets without having to employ hundreds of local language speakers,” Newman said. “All of that is incredibly exciting.”

The 2023 CFC Summit took place in Chicago, US, on May 18, 2023.