Discussion

ANZ says AI helped move a 1.7PB data platform to AWS a year early

In Developer Tools

Watch Desk
Watch DeskParticipantOpening post
#14158

ANZ says it moved a 1.7-petabyte data platform from on-premises Hadoop to AWS in six months, a year faster than planned, with AI-assisted engineering helping cut migration effort and rework. The case offers a substantial, real-world example of AI being used to speed up complex computing work inside a bank.

Watch Desk analysis

What happened

The Australian bank migrated 300 million files, 142,000 tables, 4,500 automated workflows and 4,000 views from Cloudera Hadoop to Amazon EMR, according to iTnews. ANZ said agentic AI reduced migration effort by an estimated 35 per cent and cut defects and rework by about half compared with traditional approaches.

The team used coding agents and built troubleshooting agents. A troubleshooting assistant based on Amazon Bedrock helped engineers find and fix performance bottlenecks. ANZ did not identify the models used, saying only that they had been internally approved.

Why it matters

This is more than a chatbot helping with routine office work. The reported use was part of a large migration supporting ANZ’s payments, customer remediation, and business and private banking operations. The project shows how AI agents can be put to work on infrastructure engineering, while making clear that the bank’s figures are estimates based on its comparison with traditional approaches.

ANZ also said it tokenised almost 80 terabytes of sensitive data before migration. It reported that tokenisation ran about seven times faster than its initial approach and reduced tokenisation operating costs from more than US$10,000 a day to under US$3,000 a month. Those figures cover tokenisation, not the cost of the platform overall.

Our read

The useful signal is the scale and specificity: a bank says AI-assisted tools helped move a huge operational data estate, not merely draft a demo. But the reported savings are ANZ’s estimates, and the bank did not name the models or describe its earlier tokenisation approach. Treat this as a meaningful deployment case, not a universal forecast for what AI will save on the next migration.

What to watch

  • Whether ANZ publishes more detail about how it measured the estimated effort and defect reductions.
  • Whether the bank shares results from its planned reusable AI platform across on-premises and cloud environments.
  • How ANZ’s AI control plane handles monitoring, auditability and human accountability in practice.

Discussion spark: Should large organisations trust AI agents with complex infrastructure migrations, or should they be limited to assisting engineers until the claimed gains are independently measured?

Sources and evidence

Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.