← Cases
AustraliaFinance and insurance2021 – ongoing

nib Group

Generative AI on Amazon Bedrock for member service, call notes and claims

nib's nibby assistant handled over four million member queries, and AWS reported more than $22 million in savings

WorkingVendor case studyVendor ↗

Problem

nib, an Australian health insurer, handled large volumes of member chats, calls and claims invoices. Agents spent time writing call notes and searching knowledge articles.

Approach

nib built its nibby assistant in 2021 and later moved generative AI onto Amazon Bedrock, mainly running Anthropic's Claude 3.5 Sonnet. The models summarise calls, chats and knowledge articles for agents, and draft personalised travel-insurance emails. Amazon Textract works with Bedrock on member claims invoices.

Results

over four-millionmember queries handled by nibby since 2021, per AWSVendor case study[1]
60%reduction in chat-based interaction support, per AWSVendor case study[1]
more than $22 millionsavings reported by AWSVendor case study[1]
8%reduction in average call times, per AWSVendor case study[1]

Takeaway

nib applied one model platform to several service and back-office tasks. The savings figure is AWS's and the release does not say how it was calculated.

Categories

Innovation type
Process innovation
Evidence
Vendor case study

Sources

Source summaries

[1]Amazon Web ServicesMar 2025

An AWS release on nib's use of Amazon Bedrock. It lists nibby query volume, a 60% cut in chat support and 15% in voice, more than $22 million in savings, up to 20% less post-call work, 8% shorter calls and about four million invoices a year.

Full details are in the original sources.

Added 1 Nov 2026 · Updated 1 Nov 2026
Open case page ↗