For years, the retail industry talked about AI as if it were mainly a forecasting upgrade. Perhaps a smarter dashboard here, a better replenishment model there, a more polished analytics layer sitting above the same old operational machinery. That phase, however, is ending.

The newer wave of deployment is pushing much closer to execution, where systems aren’t just interpreting what happened last week, but helping retailers decide what should happen next, and quickly.

Gartner now lists agentic AI, ambient invisible intelligence and an augmented connected workforce among the top supply-chain technology trends for 2025, while Deloitte says 30 per cent of retailers already use AI for supply-chain visibility, a figure expected to rise to 41 per cent within a year.

Fifty-nine per cent of executives surveyed by Deloitte also expect a positive return on investment from AI-driven supply-chain initiatives within the next 12 months.

That is the realistic starting point. AI isn’t transforming retail because it can produce a flashy chatbot demo. It’s transforming retail where it reduces the latency between signal and action.

In other words, where it helps merchants spot demand earlier, planners position stock better, warehouse teams move faster, stores receive goods more intelligently and delivery networks correct themselves before a delay turns into a customer complaint.

McKinsey’s recent work on supply chains makes the same distinction. The opportunity is shifting from reactive analysis to real-time decision-making, but it’s not a magic bullet. The strongest results are showing up in tightly defined and ringfenced operational use cases.

Walmart is one of the clearest examples of that shift. In April 2025, it revealed Trend-to-Product, a proprietary AI and generative AI tool designed to compress the earliest stage of the supply chain. The journey from cultural signal to actual product. The system analyses trend data from the internet and tastemakers, creates mood boards and product concepts, and helps fashion teams move from concept to shelf in six to eight weeks, shortening the traditional process by as much as 18 weeks.

A few months later, Walmart said its broader AI-enabled supply-chain model was being rolled out internationally, with systems already live in markets including Costa Rica, Mexico and Canada to predict demand, reroute inventory, reduce waste and simplify work for associates.

“At this scale, the only way to move faster is to move smarter,” says Vinod Bidarkoppa, Walmart International’s chief technology officer.

What’s striking about Walmart’s approach is that it stretches across the full chain rather than stopping at merchandising theatre. Trend-to-Product may sound like a fashion story, but it’s really a first-mile supply-chain story. The technology compresses the distance between trend detection, product development and sourcing.

Meanwhile, the international rollout shows where the harder commercial value sits, in self-healing inventory, demand anticipation and AI systems that keep orders moving through the network. This is the version of retail AI that matters most. Not a shiny add-on, but a means of making the chain less sluggish, less wasteful and less dependent on static planning cycles.

Amazon, unsurprisingly, is making a similar bet from the opposite direction.

Rather than starting with design and assortment, it’s leaning into the middle and last mile. Delivery accuracy, predictive inventory placement and robotics are Amazon’s key pushes. In June 2025, Amazon outlined three AI-led logistics tools that say a great deal about where transformation is genuinely happening.

Wellspring, its generative AI mapping system, pulls together satellite imagery, road networks, building footprints, customer instructions, prior delivery data and street imagery to improve delivery accuracy.

When Amazon began testing it in the US in October 2024, it mapped more than 2.8 million apartment addresses to corresponding buildings across more than 14,000 complexes and identified convenient parking at 4 million addresses.

At the same time, Amazon said its new AI forecasting model improved long-term national forecasts for deal events by 10 per cent and regional forecasts for millions of popular items by 20 per cent.

Again, the important point is not simply that Amazon is ‘using AI’. It’s where that intelligence sits, in the handoff between demand and location, between address and doorstep, between forecasting and network flow.

Amazon’s own description of the customer benefit was telling. Better delivery accuracy, faster shipping and improved product availability. That’s a useful corrective to the broader AI conversation.

In retail supply chains, the technology becomes meaningful when it makes the network more precise. The glamour is in the consumer promise of same-day delivery; the real work is in solving the boring operational problems underneath it.

The grocery sector offers an equally revealing example. According to Fortune’s reporting in July 2025, Albertsons is using AI not only for forecasting and optimisation, but also to match inbound shipment volumes with available store labour for receiving and replenishment.

The result is that products move from loading dock to shelf about 15 per cent faster during peak shopping seasons.

Chandrakanth Puligundla, a tech lead and data analyst at Albertsons, also points to another practical use case that deserves more attention: using AI to read unstructured supplier information such as emails and PDFs, extracting changes, risks and commitments that conventional systems often miss.

In his words, AI is particularly valuable at clarifying complexity for frontline teams. That feels exactly right. A modern retail supply chain isn’t short of data, but it is short of coherence.

This is why some of the more convincing industry evidence sits in areas that receive relatively little public attention. McKinsey says generative AI can cut documentation lead times by up to 60 per cent and reduce logistics coordinators’ workload by 10 to 20 per cent through the automation and consolidation of shipping documents.

In one example it cited, virtual dispatcher agents saved a last-mile operator with a fleet of more than 10,000 vehicles between $30 million (£22.5 million) and $35 million (£26.2 million) from a $2 million (£1.5 million) investment.

The same logic underpins Gartner’s emphasis on “connectivity and intelligence”. The value is now being created where systems can sense, decide and coordinate across functions rather than merely report what already went wrong.

So, where is AI actually transforming the retail supply chain? Realistically, in three places. First, in demand sensing and inventory positioning, where retailers are getting better at matching products to localised demand before the sales data is fully visible.

Second, in operational orchestration, where warehouses, stores and transport networks are beginning to respond to real-time signals rather than weekly planning rhythms.

Third, in administrative and decision-support work, where AI is quietly stripping friction out of documentation, supplier communication, labour allocation and exception handling.

All three go directly to availability, markdowns, waste, labour productivity and working capital. NVIDIA’s 2025 retail survey found that 59 per cent of respondents said their supply-chain challenges had grown over the previous year, while 82 per cent planned to increase spending on AI for supply-chain management; by January 2026, NVIDIA’s follow-up survey found 91 per cent were using or assessing AI and 47 per cent were using or assessing agentic AI.

But there’s a reason the most credible voices are still careful. McKinsey warns that gen AI isn’t a cure-all. Gartner frames the shift around specific use cases tied to business outcomes. Walmart’s own agentic AI strategy stresses highly specific, purpose-built tasks rather than generic automation.

In other words, the industry is learning that the supply chain doesn’t need AI everywhere at once. It needs AI in the right places, connected to reliable data, embedded in actual workflows and measured against hard operational outcomes.

That, in the end, is the realistic answer. AI is not yet reinventing every inch of the retail supply chain. But it’s already transforming the parts that matter most, being the moments where retailers either lose margin through delay and waste, or win it back through speed, precision and coordination. Right now, that is where the real transformation is happening.

Click here to sign up to Retail Gazette‘s free daily email newsletter