Woolworths pairs agentic AI with offshoring: lessons for NZ firms

By DataForge team22 November 20265 min read

Woolworths' 18 November investor day linked agentic AI shopping with more offshore work. Here's what NZ owners and IT leaders can take from the plan.

What Woolworths told investors

Woolworths Group held its investor day on Wednesday 18 November in Sydney, its first in three years. The main number was a target to add about $10 billion in group sales by FY29, on top of FY26 group sales of $71.5 billion. Grafa reports that the company called the target a medium-term ambition, so it isn't formal guidance. The plan includes roughly $2 billion a year in capital spending and more than 60 new Australian supermarkets.

Artificial intelligence came up in two parts of the plan. For customers, Woolworths said it's the first Australian grocer to use agentic AI, through its Olive assistant, which plans meals and builds shopping baskets. The company also said it expects more shoppers to hand their weekly shop to AI agents. On costs, chief executive Amanda Bardwell said the group is extending its global capabilities across Woolworths India and with managed services to become more efficient. Capital Brief notes that Woolworths confirmed earlier in 2026 that it would move hundreds of corporate roles in people, finance and IT offshore. A day earlier, Bunnings had announced about 350 technology job cuts under a partnership with Tata Consultancy Services.

Woolworths cut above-store costs by $400 million in FY26 and signalled more savings to come. Public coverage of the investor day doesn't show how much of that figure came from AI, how much from offshoring and how much from other programmes.

How Olive reached this point

Olive started life as a customer service chatbot. In January, Woolworths became the first Australian retailer to adopt Google's Gemini Enterprise for Customer Experience, aiming to turn the bot into an assistant that plans meals around a shopper's tastes and finds specials within their budget. By late April, Woolworths had given the agentic version of Olive to its more than 200,000 staff. The tool merges search and chat, adds items to a cart from a text prompt, and can identify a meal from a photo and add its ingredients.

The most transferable detail from the April demonstration is a set of eight in-house "agentic judges" that check Olive's answers before anyone sees them. iTnews described three of them. A "number cruncher" rechecks unit prices, ingredient quantities and serving sizes. A "product detective" tests product descriptions against legal, food safety and compliance rules. A "goal judge" confirms the agent finished the task it was given, so a basket meant to cost $20 that comes to $25 fails and alerts the team.

Where New Zealand fits

When the Google partnership was announced, RNZ reported that Woolworths had no set timeline for a New Zealand launch of the new Olive. Woolworths still hasn't published a New Zealand date. A University of Auckland marketing lecturer told RNZ she expected the rollout to be slow enough that shoppers might barely notice it.

The Woolworths plan still matters to NZ businesses well beyond grocery. Australia's largest employers are now presenting AI to investors as one part of a wider operating model change, alongside offshore teams and managed services. Board papers in New Zealand are starting to look the same, with AI savings sitting next to outsourcing savings. When the figures are pooled, nobody can later tell whether the AI investment paid for itself. That matters because recent ANZ surveys keep finding that most organisations haven't formally calculated what AI has returned or had finance leaders review those returns.

The judges model is also something smaller firms can copy. A mid-sized NZ distributor using a language model to draft quotes doesn't need eight judges. It does need an automated check that the arithmetic adds up, a rule-based check on pricing and compliance wording, and a test that the output does what the request asked. Arithmetic checks can usually run in conventional code, which costs very little and behaves the same way every time.

Woolworths' expectation that shoppers will delegate the weekly shop to agents affects every supplier whose products appear on a retailer's digital shelf. An agent choosing between two products relies on pack sizes, unit prices, ingredients and allergen data. If a supplier's product data is incomplete or inconsistent, an agent filtering for a budget or a dietary need can skip that product without a person ever seeing it. NZ food and grocery suppliers should treat product data quality as a sales issue now, before Olive or a rival agent reaches New Zealand.

Woolworths also gave the agentic Olive to staff before the full customer rollout. Staff testing on that scale finds failure modes that a small pilot group would miss, and it gives the business a large group of people who know how the tool behaves before customers start relying on it.

Practical next steps

  • Split the savings before you report them. For each AI initiative, record which gains come from the AI itself, which from process redesign and which from sourcing changes such as offshore or managed services. If you need a structure for this, our AI consultancy work starts with a measurement baseline.
  • Put automated checks in front of AI output. Verify numbers in code, test outputs against your compliance rules, and check that each task was completed as asked. Log every failure so you can see trends.
  • Test with staff before customers. Give internal users the tool first, collect the edge cases, and fix them before any customer-facing release.
  • Audit the product and pricing data an agent would read. Check pack sizes, unit pricing, ingredients and allergen fields across your catalogue and the feeds you send to retailers. Fixing these feeds is a process and automation job as much as a data job.
  • Map data flows before you outsource alongside AI. If AI and offshore or managed service changes arrive together, document which systems and personal information each new party can reach, and confirm your privacy obligations for data handled outside New Zealand.
  • Watch for a Woolworths NZ announcement. A local Olive launch would be the first large NZ test of shoppers handing a weekly shop to an agent, and suppliers will want their data ready beforehand.

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
Get new articles by RSS