Preserving the past, powering the future

For more than 70 years, CommBank's Archives team has been responsible for protecting the bank’s history.

"Our mandate has been the same since the archives team was established in 1954. We look after the permanent historical record of the bank so we can better serve our customers," Jeff Palmer, Group Archives Manager said.

From leadership messages to TV advertisements to major milestones in the bank's history, thousands of hours of audiovisual recordings capture how CommBank has evolved. But much of that material was stored on ageing formats that were slowly deteriorating, putting decades of corporate memory at risk.

A challenge too big for people alone

The Archives team had more than 30 years of audiovisual material to preserve, including over 2,000 hours of video stored on legacy formats. 

"Watching and transcribing over 2,000 hours of video on old formats would require 8,000-12,000 hours of skilled archival effort, making the task impossible with manual human effort alone," Palmer said.  

To meet the challenge of preserving the ageing video footage, Group Archives joined forces with Group Corporate Services alongside technologists in CommBank’s AI Labs and Engineering teams.

Group Corporate Services helped digitise fragile recordings before they degraded beyond recovery. Archivists brought their expertise in records management and historical preservation. Engineers developed the supporting platform, while AI Labs applied AI models to transform thousands of hours of content into searchable knowledge. 

Together, the teams turned ageing audiovisual material into living corporate knowledge that can be preserved, discovered and reused.

Turning video into knowledge

The AI Labs team used two open-source AI models to support the project.

Open-source (or open-weight) models are publicly available models that offer lower costs and greater customisation than closed, subscription-based AI models. “The first model was a speech to text model to transcribe the audio component and second to view the video frames and provide text descriptions of those,” AI Labs Data Scientist Lachy Berry said.

Using open-source models for the project was not only cost effective, but provided technologists with more hands-on experience with developing a customising AI-enabled solutions.  

The digitised footage with text descriptions were then shared with the Archives team to check before being uploaded to a shared platform which can be searched by anyone in the CommBank Group.  

Amplifying expertise

While AI accelerated the work, the Archives team remained at the centre of the project.

“We have a huge amount of experience within the team and we have found that the best outcomes come from pairing AI with that expertise,” Jeff said.

“The project is a great example of how technology and teams are working together to deliver value in a practical way by removing manual barriers, bringing together expertise from across the bank, and focusing on shared goal,” he said.