What Australia's CEOs said in Sydney
On Wednesday 18 November, Commonwealth Bank ran its Accelerate AI summit in Sydney. It put the chief executives of Telstra, Wesfarmers and Origin Energy on stage with the chair of the Productivity Commission and CommBank's own economists. The speakers mostly agreed on one point. Australian organisations have spent the past couple of years giving staff AI tools, and that phase hasn't yet lifted national productivity.
Telstra CEO Vicki Brady said Telstra had concentrated on giving employees access to AI tools and building their skills, and is now starting the harder job of redesigning how work gets done. According to AAP's report from the summit, Brady described personalised learning pathways and an AI learning agent that coaches staff during their normal work. She also said Telstra can already see some areas that will need fewer skills and others that will need new ones.
Wesfarmers CEO Rob Scott described a staff-facing chatbot at Bunnings that gives team members technical and project information and has saved more than 30,000 hours. Scott told AAP that getting teams to help tailor and build solutions in-house has produced much better outcomes for Wesfarmers. Wesfarmers hasn't published what the Bunnings tool cost or how the hours were measured.
Microsoft CEO Satya Nadella joined CommBank CEO Matt Comyn by video. He explained that Microsoft now uses agentic AI to reconcile its books at quarter end, with audit and human checking still in place inside a changed workflow. Comyn said he expects the way organisations are structured, the way decisions get made and the way technology gets used to change over a 2030 timeframe.
Why the economists were less upbeat
The policy panel gave the CEOs' comments their context. CommBank's Helen Mayhew pointed out that Australian productivity fell over the past 12 months and is back at its 2019 level, and summed the problem up in one line: "So the technology has arrived, but it's not showing in the numbers." Productivity Commission chair Danielle Wood said most people use AI for work outside their core tasks. e61 Institute CEO Michael Brennan compared AI with factory electrification, where the productivity gains came decades later, once businesses had reorganised how they produced goods.
The Productivity Commission estimates AI could add about 4% to labour productivity over a decade, and Wood told the summit she considers that figure conservative. In June she said the benefit depends on businesses adopting AI deeply instead of using it only for basic tasks, and that Australia trails the US, UK and Canada on adoption.
How New Zealand compares
The New Zealand numbers look much like the picture Mayhew and Wood painted. Datacom's 2026 State of AI Index, released in August, found that 91% of NZ organisations use some form of AI. Only 4% said AI had transformed their core operations, down from 8% a year earlier. General-purpose assistants were the most common tool at 68%, and 13% of organisations used agentic AI. Only 22% had a dedicated AI leadership role, and 20% had a dedicated AI budget with cost tracking. About a third said their AI returns exceeded the costs.
Put the two sets of findings side by side and most New Zealand businesses are at the stage Telstra now describes as finished. Staff have chat assistants, a few teams have built useful prompts, and nobody can point to a process that runs differently because of AI. Wood's comment about non-core tasks matches the Datacom split between assistants and agentic tools. Drafting emails faster saves minutes for one person. Redesigning how quotes, invoices or service requests move through a business changes cost and cycle time for every transaction.
Nadella's reconciliation example deserves a close look from finance and operations leaders. Microsoft kept the control that matters, the audit, and changed the steps leading up to it. Mid-sized NZ firms can copy that pattern at their own scale. Month-end close, supplier invoice matching and IT service desk triage all have clear inputs, measurable volumes and an existing check at the end. AI can do the preparation while a person keeps responsibility for the sign-off.
Brady's remark about needing fewer skills in some areas is the part Australian boards are handling most carefully, and New Zealand employers should do the same. Scott's approach of bringing the teams who do the work into the design gives staff a say in the change before it reaches them. It also tends to surface the exceptions and workarounds that never make it into a process map.
New Zealand firms don't have Telstra's or Wesfarmers' budgets, and they don't need them to redesign one process properly. A firm needs a named owner, a baseline measurement, and a willingness to change the steps instead of adding AI on top of them. Our AI consultancy work usually starts by picking that first process and agreeing how success will be measured.
Practical next steps
- Choose one core process with volume. Pick work that runs dozens or hundreds of times a week, such as quoting, invoice matching or ticket triage. Avoid starting with individual productivity tasks, because the gains are hard to see and harder to add up.
- Measure before you change anything. Record hours, cycle time, error rates and rework for at least a month. Wesfarmers' 30,000-hour figure carries weight only because someone counted, and your own business case will need the same evidence.
- Redesign around the existing control. Keep the approval, reconciliation or quality check that already protects you. Decide which steps AI prepares, who reviews the output and what gets logged. Our process and automation team maps workflows this way before any build starts.
- Give AI an owner and a budget line. Datacom found most NZ firms have neither. A named person with a tracked budget can stop pilots drifting and report honestly on returns.
- Involve the people who do the work. Bring frontline staff into the redesign from the start, tell them what's known and unknown about role changes, and record what you've told them.
- Put training inside the workflow. Telstra's learning agent coaches staff while they work. Short guidance built into the tools staff already use will stick better than a one-off course.
Sources
- Why AI adoption alone won't solve Australia's productivity problem (CommBank Newsroom, 18 November 2026)
- Workers can navigate the age of artificial intelligence (AAP via The Standard, 18 November 2026)
- AI, investment and regulatory reform key to lifting Australia's flat productivity (CommBank Newsroom, 25 June 2026)
- AI uptake hits 91%, but leadership gap limits AI scale (Datacom, 24 August 2026)
- New Zealand firms struggle to scale AI across business (IT Brief NZ)
Want to know what this means for your organisation? Our team in Auckland can show you your top three opportunities in a free 30 minute assessment.
Get my free assessment How we help