Why grocery is becoming the next test case for agentic AI

When Woolworths unveiled its latest Google Gemini AI integration, branded as Olive, the announcement was framed as a convenience upgrade. But the shift is more structural than cosmetic.Olive does not simply recommend products; it can place suggested items directly into a customer’s shopping basket, drawing on past purchase behaviour, seasonal patterns and contextual cues to shape the shop. In doing so, Woolworths is formalising what grocery shopping has long been, repetitive, habitual and time-poor. The innovation lies not in replacing choice, but in pre-empting it, therefore positioning AI as the first actor in the transaction, rather than the last.Habit, acceleratedAccording to professor of marketing and consumer behaviour at the QUT Business School, Dr Gary Mortimer, the behavioural leap is smaller than it appears. “If you’re a shopper that does a lot of online shopping, most of us are kind of doing that already,” he explained to Inside Retail. “When you go in, you can simply edit last week’s basket.” Olive extends the logic of “edit last week’s basket” across a much longer memory. Mortimer said an AI agent could look beyond last week to baskets across the last 12 to 24 months, learning from what customers remove and retain and adjusting future suggestions accordingly. Over time, Mortimer noted, they learn along in unison, refining future baskets based on what shoppers remove or retain.What differentiates agentic AI from earlier recommendation engines is contextual understanding. Legacy systems could surface similar products or discounted substitutes, but they lacked comprehension. “What it didn’t do is understand what you’re doing with those products,” Mortimer said. Under Olive’s logic, ingredients become signals. “If I bought arborio rice, chicken stock, chorizo and garlic, an AI agent would go looking at those ingredients – he’s making risotto.” The next suggestion might not be more rice, but garlic bread or dessert. But the system is the same, moving from replenishment to intent, reshaping how value is inferred.Does AI have a narrow choice?Concerns that algorithmic baskets might entrench incumbents or disadvantage niche brands are common but, in Mortimer’s view, overstated. “I actually don’t think it’ll have any impact on ranging or choice,” he said “I think it’ll just simply increase the efficiency and speed and frictionlessness of grocery shopping.” High-penetration products such as Cadbury block chocolate or Kellogg’s Cornflakes will still dominate where habits are strong. But niche products, whether a South Australian raspberry jam or regional cheese, remain visible precisely because they are habitual for a smaller cohort. The product will still appear into a consumer’s basket, because that’s a product they would normally buy.A longer memory than the shopperThe strategic advantage of AI is not persuasion, but recall. Shoppers can easily replicate last week’s order; they cannot intuitively parse two years of seasonal behaviour. “We don’t have the ability nor the time to look at our last two and a half years worth of purchases,” Mortimer said. Olive can, and does so instantaneously. It can anticipate winter stews from past chuck-steak purchases, or suggest casserole recipes as temperatures drop. Crucially, it offers ingredients as a set, reducing forgotten items which is one of grocery’s most persistent friction points.Beyond Woolworths, grocery is slowly becoming the category where consumers first learn to let software act on their behalf  and the psychology is already shifting. Recent global surveys show convenience is now the dominant driver of digital grocery behaviour, a 2025 PwC consumer study found more than 9 out of 10 shoppers are willing to share personal data if it meaningfully reduces effort. And on delegation itself, a Contentsquare survey in late 2025 found 30 per cent of consumers were willing to let an AI agent complete purchases, a meaningful minority, but not yet a majority. This year, grocery fits that threshold neatly. If an algorithm gets your yoghurt wrong, the cost is minor but if it saves you time on a Tuesday night, the reward is immediate. Unlike fashion or beauty, grocery doesn’t rely on aspiration but on rhythm and that makes it the first place shoppers begin to trust AI to decide.As AI mediates more decisions, transparency becomes unavoidable. Mortimer is pragmatic. “I think you will always get consumers that will question what is being suggested, is the supplier paying the retailer a rebate?” Mortimer’s view is that this scrutiny is not new, but newly visible. Consumers have long understood that grocery recommendations are rarely neutral whether they arrive via a shelf ticket, a catalogue or a loyalty app. Olive does accelerate that influence. Importantly, the power will rest with consumers, who can decline suggestions, remove items or ignore promotions altogether.The first step, not the endgameSeen in context, Olive is less a radical leap than the first formal expression of agentic commerce in Australian grocery. “All Olive will do is expedite what’s already happening,” Mortimer said. The longer-term horizon is more consequential, shoppers authorising AI agents to transact autonomously within set limits. With Mastercard already trialling authenticated AI-led purchases and voice assistants normalising delegated checkout, the basket-building phase may soon feel conservative. For now, Olive signals a recalibration of retail labour, from searching and remembering to reviewing and approving, and a reminder that in grocery, speed remains the most persuasive feature of all.Further reading: How agentic AI is reshaping retail efficiency, innovation and competitiveness.