Woolworths used its full year results to talk about the AI its own staff use, rather than the AI its customers see. Chief executive Amanda Bardwell revealed two tools the company had not discussed publicly before.
The headline number is Team Assist, an employee support tool that fielded 7000 inquiries a week over the last financial year, with escalation to a human removed in nine out of ten cases.
What the two tools do
Team Assist works as a decision making aid for staff questions that would otherwise land on a support desk. The second tool sits in marketing and cut production of the weekly shopper catalogue from a week to a few hours, as reported by iTnews.
Bardwell described both as having a real impact, and framed the wider programme as transforming operations and workflows rather than chasing a customer facing feature.
Two more still in trial
Head of SAP Murtuza Imdadali disclosed two further tools at an industry event this month. One is a training chatbot that walks staff through the company’s promotions app, which he described as complex enough to need a learning curve, and which he expects to go live soon.
The other checks product dimensions for accuracy across thousands of lines, flagging suspect entries with a low, medium or high confidence rating. Wrong dimensions ripple into ranging, selling and delivery, so the payoff is operational rather than visible.

The numbers underneath
Woolworths disclosed the tools inside a strong result, which is the context that makes them credible rather than promotional. Group sales rose 3.6 per cent to $71.5 billion and earnings before interest and tax rose 12.7 per cent to $3.1 billion.
The digital side is where the AI shows up in the figures. Weekly visits to the app and website rose 23 per cent on last financial year, and ecommerce orders placed through the app passed 70 per cent for the first time in the fourth quarter, with ecommerce sales up 16.9 per cent.
Why Australian businesses should care
This is the least glamorous shape AI takes inside a large business, and the most repeatable. None of the four tools is a moonshot. They are a help desk deflector, a production accelerator, a training aid and a data quality checker, each pointed at a specific internal cost.
The pattern is worth copying at any size. Pick a queue that eats staff time, measure how often a human currently has to intervene, then judge the tool on that number rather than on how impressive the demo looks. That is the same logic behind most internal assistant work, and it is why the cheapest wins are usually inward facing. Our guide to what AI work costs covers how to scope one.
The quiet half of enterprise AI
Most of what Woolworths had said about AI until now was customer facing, from its agentic shopping assistant to a basket that predicts what shoppers forgot. Those get the coverage. The staff tools get the savings.
There is a caveat worth keeping. Every number here comes from the company describing its own tools, with no external measurement, and 7000 inquiries a week says nothing about whether the answers were good. A deflection rate is a cost metric, not a quality one.
Westpac made a similar move this month when it opened up its agentic data work. Two of Australia’s largest employers describing their internal AI in the same fortnight is a reasonable signal that the experimental phase is over.
