AI Costs Lead Westpac to Prod Staff Toward ‘Sensible’ Model Use
TL;DR
Westpac Banking Corp. is tightening oversight of AI costs by tracking token use across the company. The Australian lender is steering simpler tasks toward cheaper models instead of defaulting to expensive frontier systems. Staff are being pushed toward „sensible“ model use, matching each task with the level of quality and cost it actually needs. The case points to a wider shift: enterprises are now managing AI by consumption and ROI, not just adoption.
Nauti's Take
Small teams should avoid copying the corporate ceremony, but they should define model tiers by task: drafts, classification, and research often need a different quality level than customer replies or code reviews. First verify whether token costs are visible by workflow; without that view, optimization is mostly guesswork.
Briefingshow
Many companies first rolled out AI as a productivity experiment. Westpac is treating it more like cloud infrastructure: measurable, budgeted and tied to the right model mix. For teams, that means AI use will increasingly need a cost argument, especially when a premium model is involved.
Sources