AI IN RETAILAI in retail: fewer empty shelves, fewer markdowns, less ad spend that sells nothing

Where does AI pay off in retail?

Seven places, and they are the same whether you run 2 stores or 200: these are the retail AI use cases we would look at first with any retailer. Each uses data most retailers already hold in their point of sale, ERP, ad consoles and inbox, and each has a number you can measure before and after.

McKinsey’s 2026 survey found respondents in consumer goods and retail most often use AI agents in marketing and sales. [5] That is a fair place to look, as long as the agent spends money you can measure, which is why ad spend is one of the seven above.

What does AI in retail look like in Australia?

The sector is large and already well online. Australian retailers reported $660.1 billion in sales and service income in 2024–25, including $78.8 billion from online sales. [1]

Big retailers are moving quickly. In the Adyen Index 2026, 95% of enterprise retailers surveyed were familiar with agentic commerce, where an AI agent shops on a customer’s behalf. Shoppers are slower: 64% of Australians said they feel uncomfortable allowing AI to complete a purchase for them. [2]

So the practical advice on AI for retailers here is to put it where shoppers never have to trust it: forecasts, orders, prices, ad budgets and draft replies that a person checks. Customer-facing autonomy can come later, once the back of house runs on clean data.

And in India?

Quick commerce has changed the arithmetic. Blinkit alone added 308 stores in the September 2026 quarter, taking it to 2,751, and its net order value rose 79.3% year on year to ₹20,940 crore. Eternal, its parent, expects more than 3,500 stores by March 2027. [3] India’s e-commerce market was valued at US$159.25 billion in 2026. [4]

Every dark store is a small market of its own, with its own demand and its own competition for ad space. A brand on Instamart, Zepto and Blinkit is now making a stream of small decisions every day about keywords, bids and stock by location. That is work for software, with a person deciding how far it goes on its own.

How does AI cut wasted quick-commerce ad spend?

We built Ladya for this. It is an AI agent that monitors campaigns on Instamart, Zepto and Blinkit, flags keywords that spend without orders, and shifts budget to where return on ad spend is highest, down to individual dark stores. [6]

It logs in to the ad consoles the way your team would, so there are no APIs to build. Each morning it sends one briefing: what happened, what it did and what needs your call. You choose the mode: human in the middle, where you approve each action in one tap, or full autopilot. [6] Read more at ladya.in.

How is machine learning used in retail forecasting?

A forecast is only useful if it turns into an order. Machine learning reads sales history by store, channel and product alongside promotions, seasonality and stock on hand, then proposes how much to order and where. A buyer approves, adjusts or rejects it, and the model learns from what they changed.

The same idea works one step back in the supply chain. For Mealkraft F&B, we built one kitchen operations backend where the day’s B2B and B2C orders are consolidated at 4:00 PM and broken down into exact raw ingredient indents for procurement. The result: precise procurement and less raw material waste, with batch QR tracking from raw material lot to meal. [7] It is a rules-and-data platform rather than a forecasting model, and it is the foundation a forecast needs: one clean view of demand.

Should retailers use AI agents for customer service?

Yes, for the sorting and drafting. Where is my order, can I return this, is it in stock: most questions repeat, and the answers sit in your order and stock systems.

We have built this pattern in hospitality. For Greenpark Hotels, our AI email agent sorts each message into bookings, escalations, FAQs or priority queries and drafts a reply in the hotel’s own voice from the CRM and property management system. It categorises more than 3,800 emails a day, with rules that set what it may do and an audit trail of each step. [8] A retailer’s inbox works the same way, with order, stock and returns data in place of bookings.

Can small retailers use AI?

Yes, when the AI does the setup work. Millions of micro-retailers run their business on pen and paper or chat apps. For Windo, we delivered an AI-native commerce app that lets a merchant in Africa create a store, list products and manage sales from a smartphone. Product listings are created automatically, the app works in low-bandwidth areas, and transaction data becomes insights and a credit-worthiness profile, a first step towards access to capital. [9]

What about loss prevention and privacy?

This is the one area to slow down. Between June 2020 and July 2022, Kmart ran a facial recognition system in 28 Australian stores to tackle refund fraud. The Privacy Commissioner found it breached the Privacy Act, building on a 2024 determination about Bunnings. [10]

Start with the data you already hold for other reasons: refunds, voids, discounts, stock adjustments and delivery records. AI can find the unusual patterns in those and hand an investigator a ranked list. Get privacy advice before anything biometric.

What have we built for retail and consumer businesses?

• Ladya. Our AI agent for quick-commerce ad spend on Instamart, Zepto and Blinkit, built for D2C and FMCG brands in India. [6]
• Windo. An AI-native commerce app for micro-retailers in Africa. [9]
• Mealkraft F&B. A kitchen operations platform with automated ingredient indents, batch traceability and shelf-life dispatch rules. [7]
• Greenpark Hotels. An AI email agent that categorises more than 3,800 emails a day, the same pattern a retail service team needs. [8]
• Frosty Boy. For this consumer food brand we designed and developed the website and have stayed on as its digital partner. That work is brand and digital, not AI. [11]