Customers will shift from searching for products to describing their needs
AI shopping is still very early, patchy and far from seamless. But the direction of the future customer experience is clear: a shift from customers searching for products to describing their needs.
We are beginning to see the early signs of a behavioural shift. Customers stop being shoppers and become briefers. Instead of searching for taco shells or vegetarian mince, customers describe what they need and the AI interprets the brief, plans the options and builds the list.
This shift from keyword search to intent-based briefing is subtle but significant. It represents a move from product-centric thinking to need-centric planning, where customers articulate outcomes and the AI assembles the pathway to achieve them. For retailers, this opens the door to smarter bundling, personalisation, substitution and optimisation across the entire shop.
The emerging agentic model
Walmart has taken the most decisive step toward this future. Early in 2025, the retailer introduced its GenAI-powered assistant Sparky inside the Walmart App. The company describes Sparky as able to help customers plan, compare and purchase with confidence. It marks a shift in design philosophy. Instead of responding to a series of disconnected search terms, Sparky responds to a brief.
In October 2025, Walmart advanced this strategy again by announcing a partnership with OpenAI that will soon allow shoppers to chat and buy directly within ChatGPT through Instant Checkout. This changes the sequence entirely.
Today it’s search, scroll, pick items and then checkout. Tomorrow, it becomes tell me what you need, the AI plans the list, the items are assembled, and checkout happens inside the conversation. Walmart is reframing the shopping experience away from digital shelf browsing and toward needs-based planning.
That combination—AI-powered need interpretation plus instant checkout; this signals a structural shift in the customer journey:
Today’s Multi-Channel Shopping Journey:
Search for the product → Compare brands on retailer’s site → Select product → Review Basket → Checkout
Tomorrow’s AI Shopping Journey:
Describe your need → Review AI suggestion → Checkout in-chat
This approach represents a structural change. It shifts the cognitive load from the customer to the system. Instead of customers deciding which items fit their plan, the AI does the planning and the customer simply reviews and approves.
Australia is following but not leading
While the direction is clear, the Australian market is not yet close to this level of maturity. Local players are experimenting with personalised recipes and conversational search, but these efforts sit inside siloed environments. There is no major retailer in Australia currently offering true agentic shopping or checkout inside an LLM like ChatGPT.
The experience today remains fragmented. Retailer apps do one thing, their websites do another and external AI tools sit entirely outside the shopping ecosystem. Stock data, images, loyalty programs and delivery systems remain disconnected. This creates friction and forces customers to stitch the experience together themselves.
AI will drive the behaviour shift
When customers move from searching to briefing, the dynamics of the journey change. Instead of typing “taco shells,” a customer might say:
“I’m looking to make a vegetarian taco dinner for 2 adults and 2 kids tonight. Budget $35.”
This matters because it gives the AI context about the user’s constraints: the number of people, dietary preferences, time, budget and goals. The AI can then generate a plan, recommend products, suggest swaps, check for savings and optimise the basket. Walmart has explicitly described Sparky as able to respond to prompts like “What is for dinner?” and then build a complete meal plan and add the ingredients to the cart.
This is not just convenience. It is a fundamental redesign of the shopping behaviour loop.
A real-world test of today’s AI limitations
To understand what the journey feels like today, I tested a simple real-world scenario. I needed to plan dinner for the family and decided to hand the task to GPT. I could not access Walmart’s experience from Australia so I used GPT directly and then selected my supermarket because I already had an account.
I started with a prompt: “I am looking to make a vegetarian taco dinner for 2 adults and 2 kids tonight. Budget 35 dollars. Create the ingredient list and find the cheapest options.” GPT returned a reasonable plan with ingredients, estimated costs and basic product ideas. It could design the meal, but it could not shop inside the interface.
Every part of the journey required another prompt.
Prompt for the recipe.
Prompt for the ingredient list.
Prompt for the price comparison.
Prompt for direct product links.
Prompt again when links were incorrect.
It works, but it remains a sequence of manual instructions rather than an integrated experience. The AI plans. It does not execute.
Experience gaps exposed in the current Australian AI shopping workflow
The experiment highlighted several structural issues that limit the experience today.
No digital images inside GPT
Without product images, customers lack the visual cues that anchor choice and build confidence.
No real-time stock availability
If an item is out of stock, GPT cannot substitute or adapt the basket.
No integrated checkout
There is no ability to complete payment inside the AI conversation. Customers must switch environments to the retailer’s website or app.
No loyalty integration
Points, credits and discounts cannot be applied inside the AI environment.
No delivery or pickup integration
AI cannot check delivery windows, fees or requirements.
The result is an experience that begins intelligently but cannot finish intelligently.
Why GPT cannot complete checkout
The limitation is not technical capability. It is retailer readiness. Most retailers have not opened their APIs to allow LLMs to complete a purchase. The reasons include security, fraud prevention, identity verification, PCI compliance, data governance, integration complexity and the commercial reality that retailers want to keep customers inside their owned environments.
Until retailers allow end-to-end execution, these AI shopping experiences remain partial at best.
The real-world ending: no tacos tonight
After pulling everything into the supermarket website and adding it to the cart, the basket came to $38.10. Then reality kicked in. The supermarket has a $50 minimum for delivery. The order was not eligible and there were no tacos for tonight’s dinner.
This is the gap between the aspiration and the reality of the AI-assisted shopping journey today.
Where is AI-enabled shopping heading
Despite the gaps, the trajectory is unmistakable. Walmart’s move into needs-based planning and Instant Checkout shows where the future is heading. The Australian market will follow, but only when retailers integrate more deeply with LLM platforms, allow agentic actions such as checkout and bring loyalty, stock data, imagery and delivery into the same ecosystem.
The future state looks like a single prompt: “Plan and buy my weeknight meals for under $150, including two vegetarian options.” The AI will not only plan, it will execute. That is the promise. And Walmart has already signalled it is on the way.
Read more articles like this in my Leadership Series: AI-Powered, Qualitative Research-Driven
Customer Experience Insights https://alexallwood.com.au/ai-powered-cx-insights/