Ulta Beauty, Macy’s and Dollar General are quietly re‑architecting fulfilment around AI, automation and store‑based logistics to move inventory faster with less working capital.
In Brief
- Across beauty, department store and discount formats, fulfilment is shifting from static DC‑to‑store replenishment to AI‑supported, omni‑node networks that include automated DCs and ship‑from‑store.
- Companies are adding large automated hubs, retrofitting existing DCs, and converting thousands of stores into micro‑fulfilment nodes, while layering digital order capture, delivery partnerships and retail media on top.
- This configuration tightens working‑capital and labour constraints, and raises the stakes on inventory accuracy, tariff and freight exposure, and the true cost of ultra‑fast last mile.
The Underlying Pattern and Stakes
The common thread running through recent disclosures from Ulta Beauty, Macy’s and Dollar General is a structural redesign of how orders are fulfilled and how inventory moves. Each is building towards an AI‑enabled, store‑centric fulfilment model that promises faster service and leaner stock, but demands tighter integration between network design, planning, and front‑end execution.
This is not a generic turn to e‑commerce or a broad digital transformation story. It is a concrete operating shift: automated mega‑hubs and retrofitted DCs feeding a dense grid of stores acting as forward warehouses, with AI and modern planning systems orchestrating inventory, labour and transport across nodes. The prize is higher margin through better utilisation, lower shrink and damages, and incremental revenue from delivery and media layered on the same network.
How Companies Are Converging
One visible mechanism is the build‑out of automated regional hubs that serve both direct‑to‑consumer and store replenishment. Macy’s new 2.5 million square foot China Grove facility in North Carolina is described as its largest and most technologically advanced distribution centre, outfitted with automation, robotics and AI. It supports all product categories and is configured to ship more items from a single location so customers receive everything faster and in fewer boxes. Ulta, on a smaller footprint, has completed an automation retrofit of its Dallas DC, adding advanced automation, robotics, a new warehouse management system and warehouse execution system to strengthen inventory flow and capacity.
In parallel, store fleets are being recast as fulfilment infrastructure, not just selling space. Ulta now ships from more than 1,000 stores and reports strong buy‑online‑pickup‑in‑store adoption, with about 80 percent of its sales still coming from physical locations. Dollar General goes further: roughly 18,000 of its nearly 21,000 stores are enabled for delivery through a combination of its own myDG service and partnerships with DoorDash and Uber Eats. More than 80 percent of those delivery orders arrive within an hour. These store‑based flows change the role of the outlet from pure endpoint to dynamic node, with direct implications for backroom layout, staffing and replenishment patterns.
A second point of convergence is the deliberate shift toward lower on‑hand inventory without sacrificing availability. Dollar General has reduced merchandise inventories by 5.7 percent year‑on‑year in its latest quarter and by about 7 percent per store, while in‑stock levels improved by roughly 250 basis points. Ulta has increased inventory 16 percent to support new brands, banners and stores, but is simultaneously investing in DC automation and ship‑from‑store to raise turns and keep gross margins roughly flat despite channel and tariff pressures. Macy’s has kept inventory dollars almost flat, with units down, while heading into peak season with assortments it describes as tightly curated and newness‑heavy.
Third, all three are leaning on AI and advanced planning to coordinate this complexity. Macy’s calls out end‑to‑end planning improvements and tools such as Hold & Flow to manage inventory allocation. Ulta has spent several years modernising ERP, POS, data and supply chain systems and is now using that foundation for marketplace, personalisation and other digital enhancements, even as it accepts near‑term SG&A pressure from cloud‑based platforms. Dollar General is explicit about building an AI operating system for the enterprise, aimed at reshaping workflows, lowering SG&A per unit of work and improving productivity across supply chain, stores and back office. None of these AI layers stand alone; they sit on top of, and steer, the new fulfilment configuration.
Operating Model Mechanics
At the network level, the pattern is a dual‑tier structure: highly automated regional DCs feeding stores, and stores in turn acting as both service points and local shipping hubs. Macy’s China Grove facility is positioned in the Southeast to extend reach and reduce split shipments by holding apparel, beauty, home and toys under one roof. It initially serves Macy’s fulfilment and store replenishment, with plans to bring in other nameplates, concentrating flow to leverage automation and transport density.
Ulta’s automated Dallas DC plays a similar role, supplying stores and direct orders with better throughput and accuracy, while ship‑from‑store spreads last‑mile volume across a thousand nodes. Orders can be sourced either from central hubs or from stores depending on availability and proximity, supported by upgraded order management and inventory visibility.
Dollar General’s network mechanics are different in geography but similar in logic. About 80 percent of its stores sit in communities of 20,000 people or fewer, and roughly half of outbound transport is handled by a private truck fleet that is about 20 percent cheaper than third‑party providers. DCs push inventory into this rural grid, and that grid now handles both walk‑in demand and on‑demand delivery. The same store that takes frequent small‑basket trips from cash‑constrained shoppers is also picking and staging one‑hour delivery orders triggered by the app, DoorDash or Uber Eats.
Inventory design inside this model hinges on fewer SKUs, tighter packs and better allocation. Dollar General has removed more than 1,500 SKUs over recent years and is planning net reductions again in 2026. This rationalisation, combined with more aggressive seasonal sorts in DCs and fewer in‑store floor stands, has simplified backroom and shelf operations, supported higher in‑stocks and reduced shrink. Ulta, after a period of inventory build, is using DC automation and ship‑from‑store to manage higher SKU counts from new brands, a marketplace that added more than 3,500 SKUs without on‑balance‑sheet risk, and new wellness fixtures in roughly 50 stores, while also reducing shrink through fixtures, process changes and associate training.
In practice, this kind of configuration typically requires:
- A single view of inventory across DCs and stores so orders can be routed to the most efficient node without undermining local availability.
- Case‑pack engineering and slotting that reflect both shelf‑replenishment and pick‑from‑store patterns, to avoid double handling and out‑of‑stocks.
- Integrated transport planning that links store delivery schedules, private fleet routing and third‑party last mile capacity.
Pricing, tariffs and margin management are intertwined with these mechanics. Macy’s has absorbed 40–50 basis points of gross‑margin impact from tariffs while still nudging margins up ex‑tariffs by negotiating vendor discounts, sharing costs and selectively raising tickets. Ulta is seeing more brand‑driven price increases that temporarily raise merchandise margin as lower‑cost inventory sells through at higher tickets, but those gains normalise as replenishment arrives at higher cost. Dollar General is holding its price gap of roughly three to four percentage points versus mass competitors by engineering costs out elsewhere: shrinking and damage reductions, supply chain productivity, and a mix shift into higher‑margin non‑consumables and private brands.
Risk, Constraints and Trade‑offs
The most obvious trade‑off in this model lies between speed and cost. Dollar General’s promise that more than 80 percent of delivery orders in its network arrive within an hour, often in rural markets, drives incremental sales and basket size but adds complexity and handling cost in stores that were not originally designed as micro‑fulfilment centres. The economics are currently favourable enough that delivery contributed around 80 basis points to its latest quarter’s same‑store sales, and management characterises delivery as profitable, but sustaining that profit as volumes grow will depend on labour efficiency, picking discipline and continued use of lower‑cost private fleet where possible.
Inventory tightness versus resilience is another tension. Macy’s is carrying only slightly more inventory dollars with fewer units, and Dollar General is committing to inventory growth below sales after two years of reductions, even as they both face weather events, SNAP timing shifts and tariff volatility. These choices improve working capital and shrink but leave less buffer if demand spikes or inbound disruptions hit. Ulta has taken the opposite tack in the short term, lifting inventory by 16 percent to support new brands, spaces and international banners; the risk is future markdown if demand does not justify the breadth, which is why its DC automation and app‑driven replenishment mechanisms are critical to keep turns high.
Technology investment versus SG&A leverage is a third constraint. Ulta’s SG&A rose more than 20 percent in the latest quarter, with cloud‑based platforms, labour and supplies for merchandising and events all contributing. Management is clear that FY2025 is an investment year and expects to be more selective in FY2026, effectively entering a harvest phase where the Dallas DC, ship‑from‑store and marketplace need to generate more operating leverage. Macy’s is experiencing similar dynamics as it absorbs the cost of China Grove and digital tools while leveraging SG&A by 90 basis points through store closures and expense discipline. Dollar General, by contrast, is already booking SG&A benefits from work simplification and remodels, but notes that SG&A will not positively leverage until comps are slightly above three percent.
Finally, trade policy and fuel costs cut across all three. Macy’s assumes current tariffs persist into 2026 and is explicit that 70–100 basis points of Q4 gross‑margin pressure is pure tariff. Ulta is managing tariff‑driven weakness in styling tools, and Dollar General reports LIFO charges from low single‑digit cost inflation. Each is using vendor negotiations, origin shifts, and selective price movement to contain these headwinds, but changes in tariff regimes or shipping lanes still represent structural risk to the AI‑led fulfilment configurations now being built.
Operational Self‑check
A few questions expose whether this pattern is present or absent in a given network:
- Is there a clear, tested logic for when orders are sourced from automated hubs versus stores, and how that changes by category, season and service promise?
- Have SKU counts, case packs and floor fixtures been redesigned to reflect ship‑from‑store and BOPIS picking patterns, not just traditional shelf replenishment?
- Do planning and finance models explicitly quantify how much gross‑margin expansion is expected from shrink and damage reduction, supply chain productivity and digital monetisation, relative to tariff and fuel exposure?
What This Pattern Signals
The evidence across Ulta Beauty, Macy’s and Dollar General points to a durable structural pattern: automated hubs plus store‑as‑node fulfilment, steered by increasingly capable AI and planning systems. Stores remain central, but their job has expanded from point‑of‑sale to multi‑role nodes handling walk‑in demand, click‑and‑collect, ship‑from‑store and, in Dollar General’s case, one‑hour delivery deep into rural territories.
If this pattern persists, network design will continue to favour a small number of highly automated, multi‑category DCs feeding a wide grid of flexible store nodes, with fewer pure e‑commerce warehouses and more mixed‑use assets. Inventory will be held in tighter bands, with SKU decisions driven as much by operational simplicity and shrink risk as by range breadth. Commercial models will increasingly treat digital delivery and retail media as economic overlays on existing logistics and store assets rather than separate P&Ls.
This looks less like a temporary response to pandemic‑era demand shifts and more like a new equilibrium for consumer‑facing networks where speed, variety and capital discipline must coexist. AI‑led fulfilment in this sense is not about replacing people with algorithms; it is about orchestrating a more intricate mesh of nodes, flows and costs in ways that would be unmanageable with legacy systems and siloed planning.
This article is based on recent earnings reports and public disclosures from the companies referenced.
