May 12, 2026 by Mark Dingley
AI is starting to shape what goes into the shopping basket.
In Australia, Woolworths has announced plans to upgrade its Olive chatbot using technology from Google. Instead of simply answering questions, AI will help shoppers plan meals, interpret recipes, apply discounts and suggest items to add to their cart.
Customers still make the final decision, but increasingly, AI is influencing what they see, compare and consider in the aisles.
For manufacturers, that changes the game.
Because when product discovery starts to be shaped by machines, the quality of your product data becomes critical. To prepare for AI-led retail, you need clean data, reliable systems and traceable processes.
Image credit: Ismagilov
AI-led retail is retail shaped by systems that do more than just display products. Instead of simply helping shoppers search, AI can interpret intent, personalise recommendations, build baskets, suggest substitutions and even act on a customer’s behalf with approval.
In grocery stores, that might mean turning a handwritten recipe into a shopping list, recommending a gluten-free pasta instead of a standard one, or building a basket for the week’s meals using ingredients such as tinned tomatoes, chickpeas and coconut milk based on budget, dietary needs and local stock availability.
Retail has spent the past two decades optimising for search and shelf presence. Keywords, positioning, packaging and promotions have all played an important role in product discovery.
AI-led retail introduces a different dynamic.
Instead of a customer typing “pasta sauce” and scrolling through options, an AI assistant might build a meal plan, translate a recipe, or respond to a request like “quick vegetarian dinners for the week”. It will then select products based on suitability, availability, preferences and price.
That decision-making relies on structured, machine-readable product information.
The important thing here is that AI doesn’t interpret packaging the way a human does. It doesn’t have the brand familiarity or visual cues that consumers do. It reads data, such as product names, attributes, ingredients, pack sizes and dietary advice, and uses that to determine relevance to the query.
If that data is incomplete or inconsistent, the product becomes harder to interpret and less likely to be recommended.
Image credit: Jian Fan
For many manufacturers, product data has historically been treated as a compliance requirement. It needed to be accurate enough for a retailer listing, a barcode, a shipping carton or a product label.
A breakfast cereal, for example, might have the correct GTIN and allergen statement in one system, but inconsistent nutritional claims, missing fibre or sugar attributes, or outdated pack dimensions in another.
That may have been manageable in a traditional retail environment. In an AI-led one, it can affect whether the product is correctly filtered, compared or recommended.
Today, product data needs to be consistent across channels, structured for machine interpretation, and aligned with how products are actually produced and packaged.
That includes the basics, such as identifiers, descriptions, weights and dimensions, as well as attributes including ingredients, allergens, dietary claims and usage.
For some manufacturers and retailers, this shift is already happening at the packaging level. Retailers are moving toward 2D barcodes, which can carry far richer product data directly on-pack, including batch, expiry and attribute information.
The real challenge is ensuring that data is accurate at the source and remains consistent as it moves from production through to the digital shelf.
Many manufacturers are managing product data across multiple systems – ERP platforms, labelling software, spreadsheets and retailer portals. However, if data is inconsistent across systems, outdated, or disconnected from what’s happening on the production line, it could mean a product is misclassified, excluded from recommendations, or incorrectly suggested.
As product selection becomes more automated, trust becomes more important.
Retailers need confidence that the product being recommended is exactly what it claims to be – that its attributes are accurate, its labelling reflects the current product, and it can be traced and recalled if needed.
This is where traceability comes into play.
In AI-led retail, traceability links the physical product to its digital identity. It connects the physical product to its digital identity, ensuring that product data is grounded in real production, batches and packaging.
Without that link, product data becomes harder to trust. And if the data isn’t trusted, then neither are the recommendations built on it.
This is particularly important in categories such as infant formula, fresh meat and ready meals, where provenance, batch accuracy and recall capability are critical – not just for compliance, but for maintaining trust in automated recommendations.
Image credit: Leylaynr
Coding and labelling are where product data meets the physical product. Any inconsistencies here, such as incorrect codes, unreadable barcodes, or mismatches between label and product data, will flow downstream.
In a traditional retail environment, those issues might surface as scanning errors, compliance risks or rework.
In an AI-led environment, they also affect how products are interpreted and recommended.
For example, if a gluten-free snack bar is incorrectly labelled or inconsistently coded across systems, it may not appear in AI-driven recommendations for dietary-specific searches. If pack sizes or identifiers don’t align, a product may be filtered out of a “family meal” basket or incorrectly substituted.
Reliable coding and labelling ensure that what is produced, labelled and represented in digital systems all align.
This shift is not limited to Australia. Globally, retailers are embedding AI deeper into the shopping experience.
The direction is clear: AI will play a bigger role in how products are discovered and selected.
For manufacturers, the focus isn’t on adopting AI overnight. It’s on getting the foundations right.
That means looking at how product data is created, managed and shared. Manufacturers need to focus on integrating systems through platforms such as iDSnet and ensuring that coding, labelling and traceability processes feed into a consistent, reliable data set.
Platforms that connect these elements – linking product data with production, coding and supply chain information – can help reduce fragmentation and improve data accuracy at scale.
.jpg)
As AI takes on a greater role in how products are discovered and selected, the manufacturers who benefit will be those whose products are easiest to interpret and trust.
That starts with clean, consistent product data, reliable coding and labelling, and traceable processes that connect the physical product to its digital identity.
Talk to our team at Matthews Australia to find out how we can help.
Image credit: Muratdeniz
