Forrester published an interview with David Walker, former chief technology officer of Westpac. Walker also held senior technology roles at DBS. He says Westpac did not build a separate approval path for AI. Risk, legal, compliance and security leaders joined a central AI accelerator group from day one. Testing with regulators and the board suggested AI scales familiar risks rather than creating new ones. Walker warns that human review of AI output is a weak control. Reviewers grew complacent after hundreds of correct results and missed errors. He argues controls should be automated and then monitored themselves. On workforce change, he says tools and training alone do not shift behaviour. Westpac gave staff permission to disrupt their own work before running AI training.
What changed
Data migration at Westpac took a squad of six or seven people about six weeks per data set.
What it unlocks
Handing agents an outcome to orchestrate instead of scripting the human process step by step.
- 6-7 people, 6 weeks before
- 4 agents, 2 people, ~4 days
- DBS described as 22,000-person startup
What you need to act on it
- mature existing risk and control framework
- risk, legal and compliance staff embedded from the start
Sources