As leading Australian Investment Strategist, Hasan Tevfik noted this current reporting season in Australia, AI is being used by organisations in three ways:
- organisations are benefiting from AI CAPEX,
- companies are monetising AI by embedding it into their products, and
- organisations are using AI to drive productivity.
At Ignite, we have seen nearly every client ask the same question when it comes to the last point, using AI to drive productivity. That question?
“We’ve rolled out Copilot/ChatGPT, but where is it showing up in our operational metrics?”
For most organisations, the initial Generative AI pilot delivered headline-grabbing efficiency claims like 20% faster coding, draft emails generated in seconds, and meeting summaries synthesised in no time. But, are companies actually gaining any productivity gains? Are labour costs reducing (We know of several companies who have actually hired ‘AI Champions’ which is hardly a reduction), project backlogs reducing, or margins uplifting?
The gap between pilot enthusiasm and bottom-line impact boils down to two simple things:
- Are you simply rolling out AI on top of your systems and processes? The ‘we gave everyone Claude’ model?
- Do you have the right governance and management processes in place?
- And are you measuring output activity instead of enterprise outcomes?
When evaluating early GenAI deployments, data leaders typically rely on self-reported survey data (“I saved 30 minutes today”). The implicit assumption is that 30 minutes saved equals 30 minutes of labour value captured.
In practice, saved minutes fragment into micro-intervals across the workday. This is actually productivity leakage. If an analyst saves 4 minutes summarising a report and that time isn’t reallocated to high-value strategic modelling, it is not a productivity gain.
To gain from GenAI, you need to have a data model and strategy for your business, ensure that the data the organisation is using is structured the right way and feeding the LLMs the right data for you business and you cannot, nor should you, do it alone. The primary barrier to moving GenAI from basic text synthesis to high-value, autonomous decision-making is not model capability. It is data trust.
So, in this Australian earnings cycles, who did show specific, dollar-backed, or margin-expanding ROI metrics directly linked to their AI strategies?
Commonwealth Bank of Australia
Reductions in operational fraud losses and a measurable decrease in customer service handling costs.
CBA highlighted its machine learning and generative AI fraud detection models (built alongside platform partners like H2O.ai). The AI system processes millions of risk signals in real time to catch scams before payment clearing. CBA reported a 76% drop in scams, lowering operational charge-back costs and regulatory remediation expenses.
Pro Medicus
Expansion of operating margins and contract wins tied directly to AI-enhanced diagnostics.
Pro Medicus embedded proprietary AI algorithms directly into its Visage 7 medical imaging platform. Pro Medicus emphasised its implementation capabilities as a key competitive differentiator, with all projects tracking on or ahead of schedule. Combined with operational efficiency, this helped push PME’s underlying EBIT margins to 74.9%.
Telstra Group
Millions saved in customer support workflows and direct workforce re-allocation.
Telstra detailed the rollout of its internal Generative AI assistant (Ask Telstra), used by over 20,000 frontline employees and reported a 20% reduction in average call handling time for complex customer queries and a 70% reduction in time taken to summarize customer history notes. Management linked these efficiencies directly to reduced overheads in contact centre operations, enabling customer service cost structures to shrink without sacrificing customer satisfaction scores (NPS).
QBE Insurance Group
Significant reduction in claims processing times and improved loss-ratio accuracy.
QBE revealed quantifiable returns from automating the initial triage of commercial insurance claims. Using computer vision and natural language processing to ingest damage reports, repair quotes, and legacy policy documents, QBE delivered a 65% reduction in review submission times.
All these companies had a plan, quality data models and systems and governance in place to roll out track the gains in AI. Ensuring that you have these in place is make or break for your investment.
For questions or help with your organisation’s AI investment, please reach out.