HBF faces AI agent to members for first time

HBF Health Insurance is deploying its first generative AI agent for members with a more interactive version for its smartphone app expected to follow in the months after it goes live.

HBF Chief Information and Transformation Officer Sanjeev Gupta (left) alongside Commonwealth Bank Institutional Banking technology general manager, Amilia Wilkinson.
itnews.com.au

The agent's launch is imminent and will mark the first time that it has used externally-facing AI to provide service to its 1.2 million members digitally.

Foreshadowing the launch at Salesforce’s Finance Summit in Sydney late last week, HBF’s chief information and transformation officer Sanjeev Gupta said that the web-based AI agent would provide members with conversational search but be limited to general information.

However, he said that the agent to follow would interact with members as authenticated users, providing a more agentic, personalised experience.

“They will first be able to do what we call ‘unauthenticated conversation’ with our AgentForce, asking questions about product policy and things like that, but unauthenticated means that it’s just generic," he said.

AgentForce is Salesforce's agentic AI platform.

“The next phase that we're going live with, in a couple of months’ time, is where [the agent] will know who you are and will actually be able to do things for you as a member, like change my current details, change my address or give me information about my claims.

“That’s where the richness starts and [then] extend that to voice and so forth,” Gupta said.

Gupta said that the authenticated AI agent is expected to go live “in a couple of months”.

Speaking at the same event later in the day, HBF Member Connect general manager David Beer provided more detail on HBF’s plans for AI, revealing that the fund expected to develop more autonomous agents.

“At the moment we’re using these chat agents and it's generative. It's really helpful at surfacing, but I think for us at the organisation, what we're looking to is how we are going to really power and remove some of that low value task using agents autonomously working end-to-end in Salesforce,” Beer said.

Beer said that the launch of its first customer-facing AI agent, expected around the end of last month, followed HBF’s roll out of a similar generative system to its customer service teams.

Its initial pilot, which began last December, involved 30 customer service staff but, as of last week, Beer said that the fund had switched it on for all 200 of its contact centre team members.

Over recent weeks, the fund had been able to measure the impact that the new agent was having on call handling times, recording an eight percent decrease in average duration.

“That doesn't sound huge on one hand but when you think that an average call is around 17 minutes, it's quite measurable,” Beer said.

However, the biggest improvement that the fund had seen was in the amount of time staff were spending on hold during customer interactions, which reduced by 30 percent.

The fund did strike some challenges in the pilot phase. Beer said that its AI agent was initially too slow responding to queries, prompting the insurance provider to experiment with its large language models (LLMs) within AgentForce.

“I think in hindsight we probably played it too safe on the quality side. What that actually meant was [that] the speed just wasn't quick enough to our front line, so we heard very quickly and very loudly from our working group,” Beer said.

The experimentation period forced the fund to review some of its initial assumptions about its LLM running costs in order to achieve the response times it needed.

“We did compromise some level of cost in order to get that speed but we felt that was a really fair commercial trade-off considering we're seeing a 30 percent reduction in that whole [handling] time,” he said.

The fund also found that its initial testing group was unsuitable for revealing the full potential of the agent as staff with more experience reported not needing it.

“We brought some of our newer recruits in and that's actually truly where we saw the value and we saw the adoption,” Beer said.