On 12 August, Commonwealth Bank of Australia (CBA) reported their FY26 results. Included was about A$200 million in measured gross benefits from AI. Looking ahead, CBA also forecasted that gross benefits from AI would double in FY27, exceeding its investment. CBA’s FY26 results presentation, slide 68.
This A$200 million figure included what CBA referred to as “reinvested capacity” i.e. time freed up and redirected to other work. In accounting terms, this means that the A$200 million that was reported shouldn’t be read as an equivalent cost saving (EAC). It should instead be read in the context of business benefits i.e. the business value that was generated from the time that was saved.
Doing more with the same team
In the results call, the bank’s CFO, Alan Docherty, said CBA had chosen to use the AI driven productivity gains that were made within its technology team to get more work done. Results transcript, pages 19 and 30.
To illustrate what he meant, suppose that AI coding tools enabled an engineering team to finish its planned work ahead of schedule. Management then makes the decision to keep the team intact and to use the extra time to bring forward a customer feature.
In this scenario, the bank’s payroll hasn’t decreased, but the business may still benefit from releasing a feature sooner. To confirm this, I’d want to see how much earlier it reached customers and whether that improved their experience or brought revenue forward.
Why I wouldn’t add these two figures together
CBA also reported about A$400 million in incremental productivity cost savings in FY26. That total already includes some savings from AI. Results transcript, pages 9, 17 and 32.
The bank’s A$200 million gross AI figure also includes cost savings, alongside revenue benefits and time redirected to other work.
Adding A$200 million to A$400 million could therefore count some AI savings twice. We’d need a breakdown showing which benefits appear in both figures before calculating a combined total.
The same issue can arise inside your own business. The AI team might report a cost saving that also appears in the business team’s results. Both teams can show what they contributed, but the saving should be counted once in the organisation’s total.
The person responsible for each benefit should be able to show how it was measured and where it has been reported.
Calculating payback
CBA expected gross AI benefits to exceed investment in FY27. That comparison covers one year; it doesn’t show when earlier investment will have been recovered. Results transcript, page 30.
To calculate payback, I’d need to see the initial investment, then the cash benefits and ongoing costs over time. That would show when the investment was recovered. An ROI calculation also needs a stated period and a clear basis for the costs and benefits included.
If staff time has been valued in dollars, I’d want to know how that value was calculated and what happened to the time. Using it to deliver another feature has a different financial effect from avoiding the cost of hiring another person.
The presentation and transcript don’t give enough detail about the benefit calculations and associated costs to work out ROI from these figures alone.
What I’d want to see before investing further
For the engineering team, I’d follow the time saved through to the feature delivered and the result after release.
Where expenditure has fallen, I’d want to see the change in the relevant budget or accounts. Where a cost has been avoided, I’d want evidence of what would otherwise have been spent.
Forecast benefits should be clearly separated from results already achieved.
With that information, I can judge whether to invest more, change how the tools are used or invest elsewhere.