Under sustained tariff and cost pressure, large retailers are converging on an AI-led inventory orchestration model that uses digital control of stock, sourcing and fulfilment to preserve value and speed.
In Brief
- Across Walmart, Lowe’s, Target and BJ’s, the shared pattern is a shift from local inventory firefighting to centrally orchestrated stock, pricing and fulfilment decisions driven by AI.
- Companies are reconfiguring networks and systems so DCs, stores and marketplaces act as a single pool: automated nodes, store-as-node fulfilment and extended-aisle sourcing all feed one control layer.
- The trade-off is real: lower working capital and faster delivery sit against tariff volatility, margin-dilutive wholesale moves and rising service promises that demand relentless productivity.
The Underlying Pattern and Stakes
Big retailers and grocers are no longer treating inventory as a store-by-store problem. The recurring pattern across recent disclosures is a move to AI-led inventory orchestration: a central digital spine that decides what to hold, where to hold it, how to source it and how to route orders, with tariffs and cost shocks treated as baseline constraints rather than one-off crises.
This is not a generic digital story. It is about how Walmart, Lowe’s, Target and BJ’s are rewiring the way stock, sourcing and last-mile capacity are controlled so they can extend value and speed guarantees while labour, tariffs and input costs climb. The stakes are clear in the numbers: Walmart holding global inventory growth to roughly half the sales rate, Lowe’s flattening inventory despite folding in two distributors and higher tariffs, Target lifting in-stock rates by more than 150 basis points on its 5,000 top items, and BJ’s cutting per‑club stock by 5% while in‑stocks rise.
How Companies are Converging
The first common move is to create a single, AI‑supported view of what stock is where, and use that to actively steer both buying and fulfilment. Walmart has deployed handhelds and computer vision to more than a million associates, mapping store inventory in real time and tying it to automated DC and FC feeds. Target reports machine‑learning systems that forecast, order and position inventory from supplier to shelf, credited with more than 150 basis points of year‑on‑year improvement in on‑shelf availability for the 5,000 items that account for 30% of its unit sales. BJ’s is using in‑club robots and digital twins of each building to spot gaps, prioritise tasks and generate the most efficient pick paths, driving about 40% higher picking productivity while in‑stocks rise.
The second convergence is the way networks are being retuned so different nodes play distinct roles in a coordinated system. Target’s Chicago pilot re‑cast high‑traffic stores as experience-heavy, low brown‑box locations and pushed more digital fulfilment to lower‑traffic stores with large backrooms. That configuration is now being rolled to 35 more markets. Walmart in the US is using a similar logic, with roughly 60% of stores now receiving some freight from automated DCs and more than half of e‑commerce FC volume automated, while stores act as forward nodes serving under‑three‑hour delivery windows. Lowe’s is redesigning freight flow into stores, using Freight Flow 3.0 to sequence inbound loads so critical and Pro‑sensitive SKUs are stocked by overnight teams, with other stock handled later to free daylight labour for customers.
A third shared mechanism is de‑risking owned inventory by leaning harder on extended assortments and marketplaces, while using AI to stop this becoming an uncontrolled SKU explosion. Walmart’s marketplace now carries somewhere around half a billion SKUs, with categories like home décor and toys growing more than 40% year on year, yet owned inventory in the US is up only 2.6% against mid‑single‑digit sales growth. Lowe’s Pro Extended Aisle plugs supplier catalogues directly into ordering, allowing vinyl siding, doors and electrical components to be sold without stocking them in every store, and is being expanded weekly. Tesco adds a European perspective with a marketplace above one million SKUs and a Buy Box to keep those offers price‑sharp, while BJ’s keeps its digital assortment mostly club‑fulfilled but supplements it with AI‑driven content and search to grow non‑owned ranges where it makes sense.
Finally, all four are using these orchestrated inventories to support fast and ultra‑fast fulfilment promises without flooding the network with safety stock. Walmart can now reach 95% of US households within three hours and, in many cases, under an hour. Target offers two‑hour delivery from 97% of its stores and same‑day services growing more than 35% year on year. Tesco’s Whoosh rapid delivery is now in more than 1,600 UK stores and covers over 70% of households, contributing two percentage points to online growth. BJ’s fulfils more than 90% of digital orders from clubs, leaning on robotics‑informed layout and stock visibility rather than duplicative e‑commerce DCs.
Operating Model Mechanics
At the core of this pattern is a change in node roles and decision rights. Stores, clubs, FCs and wholesale branches are no longer treated as independent inventory owners. They are nodes in a directed network where a central orchestration layer decides which node serves which demand based on stock, speed promise and cost.
In practice, this configuration typically requires:
- one demand and inventory spine that can see owned and third‑party stock, by node and by time, and expose that to both front end and planning teams;
- role definitions per node (retail‑heavy, fulfilment‑heavy, mixed) enforced in routing and labour planning;
- clear service‑time tiers linked to inventory and capacity rules, so two‑hour, same‑day and next‑day promises are not made in isolation from supply.
Walmart’s spine is built around automated DCs, FCs and store‑level visibility. Automated regional DCs feed stores on a more predictable, palletised basis. FCs handle the bulk of national direct‑to‑home volume, with more than half of their throughput now automated, cutting handling time and unit shipping costs by double‑digit percentages and shrinking waste in fresh categories. Stores are then given clear roles: some act as heavy digital fulfilment nodes with dense pick operations, others privileged for shopper experience with lighter brown‑box volume but still used to meet sub‑hour delivery in their catchments.
Target is following a similar route but with a market design lens. Its inventory and routing systems now determine which orders in a metropolitan area should go to a fulfilment‑heavy store, which to a standard supermarket and which to a DC. That is bound into labour and space: high‑traffic stores see their brown‑box work dialled down and their checkouts and aisles reconfigured, while low‑traffic, large‑backroom stores add BOPIS staging and higher pick density. AI‑enabled replenishment tools then send those stores a ranked list of critical out‑of‑stocks to fill, making sure the effort spent moving stock in the back translates into on‑shelf availability where the order promise depends on it.
Lowe’s mechanics are more Pro‑tilted. The Pro Extended Aisle creates a virtual, supplier‑owned layer of stock for bulkier and trade‑specific items. The AI‑assisted Pro Companion pulls that supplier‑side visibility into project planning, producing bill‑of‑materials estimates and then routing orders either into Lowe’s stores, cross‑docks or direct drops to job sites. Freight Flow 3.0 and full‑shelf replenishment then knit store‑level execution back to that demand: overnight crews work the highest‑priority Pro and seasonal loads first, and AI systems generate store‑specific restock lists based on real‑time scans of shelves and historic sell‑through.
BJ’s offers a compact illustration of orchestration in a high‑density warehouse format. Robots capture shelf images to maintain a digital twin for each club. That digital twin feeds planning and allocation teams, who now move stock with a better sense of where displays are thin, where pick efficiency is suffering and which items are critical for ExpressPay, BOPIC and same‑day flows. The result is less stock overall, more of it in the right place, and roughly 40% better picking productivity for digital orders off the same club floor.
Underneath each of these is a commercial and sourcing layer being tuned for this orchestration. Marketplace and extended aisles are deliberately used to move long‑tail and tariff‑sensitive ranges into lighter‑capital models. Vendors are brought closer: Macy’s sharing tariff costs with suppliers, Walmart pushing marketplace sellers into its own fulfilment service so they become part of its node network, Tesco building sustainability‑linked incentives with farmers to secure fresh flows. AI is applied not just in forecasting but in range curation: Tesco’s AI‑powered range tools adjust local assortments, Walmart and BJ’s use AI to spot assortment gaps or over‑ranges in enormous catalogues, and Lowe’s merchants get AI decision support so their time shifts from spreadsheet work to trade‑off choices.
Risk, Constraints and Trade‑offs
Inventory orchestration at this scale is not free. One obvious tension is between working capital and continuity. Walmart is showing that owned stock can grow more slowly than sales when automation and marketplace scale are in place, but Home Depot, which is also leaning into speed, saw inventories rise by $2.3 billion and turns fall as it layered a wholesale distributor and more DFC stock into its network. Lowe’s is holding total inventory flat year on year despite adding roughly $500 million of stock from FBM and ADG and absorbing higher tariffs, but at the cost of lower reported gross margin from those lower‑margin wholesale lines.
Another constraint is the margin impact of wholesale and marketplace roles in the network. Lowe’s and Home Depot are both adding large wholesale distributors to their portfolios. For Lowe’s, FBM and ADG are expected to dilute consolidated gross margin by about 75 basis points and operating margin by about 30 basis points in the first full year, even as they are accretive to earnings per share. Home Depot expects SRS and GMS together to lower its structural gross margin by around 120 basis points and operating margin by about 60 basis points versus a pure retail profile. These moves give both companies deeper control of Pro flows and more stable volume, but they pull blended margins down and increase the operational complexity the orchestration layer must manage.
Tariffs and regulatory shocks are the other persistent headwind. Macy’s sees a 40–50 basis point gross margin hit and $0.25–$0.35 of earnings drag from tariffs despite mitigation. Lowe’s expects a 40–50 basis point impact on gross margin once FBM and ADG are folded in, on top of tariff policy that its CEO openly describes as fluid. Walmart has faced higher costs from tariffs on some categories yet has increased the number of rollbacks in US stores by more than 20% to over 7,000 items, with more than half in grocery. In this configuration, the AI‑enabled cost and sourcing levers are not optional extras; they are what make subsidised price positions and rapid service promises sustainable.
Finally, there is operational load and organisational complexity. Running a network where some stores are fulfilment nodes, others experience nodes; where some SKUs are supplier‑owned, others DC‑owned, and still others club‑only; and where AI agents can originate orders unpredictably through new channels, stretches planning disciplines. Target has already eliminated about 1,800 HQ roles and is reshaping its merchant roundtable so decisions can move faster. Walmart is standardising platforms globally under a ‘build once, scale globally’ principle to limit divergence. BJ’s and Tesco are embedding robotics and AI analytics into planning to keep the human workload manageable as data intensity rises.
Operational Self-check
A configuration like this is invisible on a slide until it breaks under stress. The questions that expose whether AI-led inventory orchestration is real rather than aspirational tend to be uncomfortable:
- Can the network point to a small number of nodes where inventory is deliberately heavier for specific service times, and is that visible in routing rules rather than in anecdotes?
- When tariffs or category shocks land, is the first response a sourcing and assortment move or a margin call, and does that response appear consistently across regions?
- In weekly planning, do digital and store demand compete explicitly for the same stock with clear priority rules, or are they planned as separate streams reconciled after the fact?
What This Pattern Signals
Across Walmart, Lowe’s, Target and BJ’s, the same structural shift is visible: inventory and fulfilment are being run as a globally optimised, AI‑assisted control problem rather than as a collection of local decisions. Tariffs, labour inflation and service promises are baked into that control logic, not treated as episodic shocks. Networks are becoming layered: wholesale branches, automated DCs, FCs, stores and clubs each have defined roles in a unified orchestration model that can direct demand and stock to where service time and economics line up best.
If this pattern persists, network design will continue to decouple assortment breadth from owned inventory through marketplaces and extended aisles; store formats will be specified as much by their digital roles as by their selling space; and automation will be sized not around static throughput, but around the variability injected by AI‑driven front ends and rapid‑delivery commitments. Margin structures will increasingly depend on ancillary profit pools such as retail media and membership to fund the cost of speed and resilience.
This looks less like a temporary response to a bad tariff year and more like the next operating model. Each company is still in early execution, with different emphases, but the direction is consistent: AI-led inventory orchestration is becoming the way large consumer businesses reconcile price, speed and volatility at scale.
This article is based on recent earnings reports and public disclosures from the companies referenced.
