Rising volatility in demand, tariffs and channels is forcing operators to hard-wire AI and data into how inventory is planned, routed and repriced across sprawling physical networks.
In Brief
- AI-led inventory orchestration is emerging as a structural shift: decisions about what to buy, where to place it and how fast to move it are being recoded in software rather than left to periodic manual planning.
- Companies as different as On, Bath & Body Works and TJX are wiring AI, automation and data-rich feedback loops into buying, allocation and fulfilment so that inventory acts as a live lever, not a quarterly outcome.
- This pattern brings a new constraint set: higher data and systems dependency, tighter links between marketing and supply, and constant tension between capital discipline, tariff exposure and service promises.
The Underlying Pattern and Stakes
Beneath very different brands sits the same operating preoccupation: how to keep networks responsive without flooding them with stock. The emerging pattern is AI-led inventory orchestration, where granular data and machine-led decisions shape buys, placement and flow in near real time.
This is not about generic digital dashboards. It is about using AI and automation to collapse process steps, sense demand from live signals, and then push that intelligence deep into sourcing, manufacturing, allocation and store operations. The stakes are clear in the numbers: On is holding gross margins above 62 percent while absorbing new tariffs, Bath & Body Works is trying to rebuild margin while cutting inventory, and TJX is expanding inventory per store yet still widening merchandise margins.
How Companies Are Converging
One shared move is to treat customer interaction data as a planning asset rather than a marketing by-product. On has deployed a conversational AI layer across service platforms, positioned explicitly as a way to have personalised, large-scale conversations and then extend that AI blueprint from design into the global supply chain. That same logic appears in Bath & Body Works’ shift to richer digital content and a loyalty programme handling more than 80 percent of owned-network transactions, as well as TJX’s use of marketing mix modelling to understand where traffic and baskets really come from. In all three cases, demand is being sensed from direct interactions and fed back into what gets made or bought.
A second convergence is the move from fixed, batch-heavy manufacturing and fulfilment to more automated, step-reduced processes that can flex with demand.
If AI systems are sensing demand earlier and more precisely, the physical network also needs the ability to respond just as quickly. That is where manufacturing innovation becomes critical, because demand signals only matter if production and fulfilment can adapt to them.
On’s LightSpray facility in Busan is the most explicit expression of this shift. A robotic arm spins a 1.5 kilometre filament into a shoe upper in about three minutes, collapsing roughly 200 assembly steps into one and scaling capacity thirty-fold versus the prior year. In practical terms, this turns a traditionally fragmented manufacturing stage into a programmable node in the network. Instead of relying on a web of cut-and-sew suppliers, production can be adjusted more quickly in response to demand signals flowing through the system.
Bath & Body Works is making a quieter version of the same move by exiting a third-party fulfilment centre, investing in logistics upgrades and using a domestic supply chain that can ‘chase into demand’ for a new hand soap that is running at twice the productivity of the item it replaced.
Third, channel mix is being managed as a supply chain configuration question, not just a revenue one. On has pushed D2C share to about 42 percent of sales, with owned retail and e-commerce carrying more than 60 percent of fast-growing apparel and accessories. Bath & Body Works is adding Amazon as a curated wholesale node while trimming SKU count in stores by 10 percent and lowering the free-shipping threshold. TJX is expanding store count past 5,000, building e-commerce assortments and using remodels and relocations as continuous tuning of the physical network. In every case, inventory orchestration is inseparable from channel strategy: where inventory sits and how it moves is being rethought alongside where demand shows up.
Finally, there is a shared discipline around margin as a live output of these configurations. On has brought airfreight down and widened the spread between purchase and selling price while raising average selling price through mix, not list increases. Bath & Body Works is upfront that tariffs will take about 150 basis points off gross margin in the first quarter, and that a 250 million dollar cost programme will be used primarily to fund product and logistics changes rather than to bank short-term margin. TJX is using its 21,000-vendor base and non-committed buying model to lean into a glutted merchandise market, raising merchandise margin and using hedging and opportunistic buys to offset tariff and freight noise.
Operating Model Mechanics
AI-led inventory orchestration shows up first in how origin and plant roles are defined. On’s Zurich labs and foam competence centre sit outside the traditional Asian sourcing orbit, anchoring critical materials knowledge in-house and reducing dependency on external tech clusters. The Busan LightSpray plant then becomes a specialised node in the network: it is not just another factory, but an automated origin for uppers feeding multiple shoe lines. That allows production planning to treat uppers almost as a configurable digital component, with capacity staged centrally and final assembly potentially closer to demand.
Bath & Body Works is reworking its own origin rules at a portfolio level. Restages of hero forms such as moisturizing body wash and a new flat-back spray sanitizer imply new formulations, packaging suppliers and tooling. To avoid stranded stock and obsolete packaging, supply chain, product and merchandising teams are now working side by side from early stages. That integration is what lets the business chase outperforming innovations such as the new hand soap, adjust buys mid-season and still keep inventory five percent below prior-year levels.
Routing and flow decisions are also being recoded. On’s channel mix, with D2C at nearly 42 percent and 67 owned stores whose latest openings are around 40 percent larger than the prior estate, demands differentiated fulfilment rules. High-productivity flagships in Tokyo, Paris or Shenzhen need dense, apparel-heavy assortments and frequent top-ups, while wholesale partners in the Americas and EMEA rely on reliably full pallets and stronger pre-season order books. The strong Fall/Winter 2026 order book gives On forward visibility that then feeds into how much LightSpray and foam capacity is allocated per region and channel.
Bath & Body Works is using Amazon as a distinct wholesale leg, with a 50-SKU curated range, while its own digital channel adopts a lower free-shipping threshold. That split changes order economics and network flow: Amazon orders move as wholesale cases to an external node, while the brand’s own e-commerce must now handle more, smaller baskets without destroying margin. The exit from a third-party fulfilment centre and new investment in logistics and fulfilment capability suggest that orchestration rules for which orders flow from which node are being rewritten, with AI likely allocating orders between stores, DCs and drop-ship based on cost and service.
TJX’s mechanics are anchored at the other end of the spectrum: opportunistic sourcing and agile allocation. With inventory up 14 percent on the balance sheet and 10 percent per store, but described as ‘great’ given exceptional merchandise availability, the company is clearly holding more stock to exploit supply. AI may not be front and centre in its language, but the underlying need is similar: systems must ingest offers from around 21,000 vendors, score them by margin, trend and location, and then allocate to over 5,000 stores and online outlets in a way that keeps the value gap to competitors while raising average retail through a better-good-best mix.
In practice, this kind of orchestration requires a small set of structural capabilities:
- A planning spine that integrates demand signals from stores, digital, marketplaces and wholesale into a single view of required buys and inventory deployment.
- A data model that tags inventory by attributes that matter for allocation and pricing decisions, from tariff exposure and freight mode to margin band and trend life.
- Execution systems that can translate those decisions into factory orders, DC wave plans, store allocations and price or promo rules without handcrafting every step.
Across these companies, there are clear signs of that spine being built. On’s talk of an AI blueprint from design into supply chain, Bath & Body Works’ sequencing of marketing events, collaborations and inventory, and TJX’s ability to move quickly into under-served categories or geographies all depend on more than local spreadsheets.
Risk, Constraints and Trade-offs
AI-led orchestration does not remove hard trade-offs; it makes them visible faster. Working capital versus continuity is the first. On is holding inventory of roughly 419.8 million Swiss francs with net working capital at 18.9 percent of net sales, aligning volume growth with a guided 23 percent sales increase. That is disciplined for a high-growth, innovation-heavy portfolio, but it still ties up cash. TJX is deliberately expanding inventory per store to exploit vendor over-supply, enabled by 6.2 billion dollars of cash on hand. Bath & Body Works, by contrast, is taking inventory down and living with lower top-line guidance while it rebuilds its product and channel model; gross margin guidance of about 42.4 percent bakes in both investment and tariffs.
Tariff and policy risk is another shared constraint. On has already absorbed higher United States import tariffs and still expanded gross margin, helped by FX tailwinds and mix. It also notes the two to three month lag between tariff regime changes and P&L impact because inventory is already cleared through customs. Bath & Body Works expects a 150 basis point gross margin headwind in the first quarter from tariffs that were not present a year earlier, even as it plans to offset tariffs at the earnings line over the full year through cost and mix. TJX assumes in its guidance that it can neutralise tariff pressure by shifting buys and leaning into confused markets where its buyers can secure better terms. All three are effectively using orchestration logic to price tariff risk into origin decisions, inventory positions and margin expectations.
Channel and capacity balance is a subtler tension. On is clear that D2C will outgrow wholesale and that apparel will outgrow the total, pushing more volume into smaller, more complex orders and returns regimes. That raises last-mile and handling costs even as it supports price and margin. Bath & Body Works is layering Amazon wholesale on top of a still-prominent store fleet, where about 60 percent of locations are off-mall, and a digital channel whose economics have been altered by a lower free-shipping threshold. TJX continues to add 146 net new stores in a year and remodel around 540, with e-commerce still a minority but growing part of the mix. In all cases, orchestration choices about which node serves which order shape the cost base as much as they shape service.
There is also a constraint around organisational and data maturity. On has invested in an R&D team that has grown tenfold in five years, with over 400 experts across disciplines and a strong operations and supply chain team automating distribution. Bath & Body Works is mid-way through a Consumer First Formula that explicitly brings supply chain, product and brand into the same decision frame, backed by a 250 million dollar cost and reinvestment programme. TJX’s structures are more decentralised and vendor-driven, relying on 1,400 buyers and long-practised processes. AI-led orchestration amplifies strengths in such environments but can also surface historical fragmentation: if pricing, allocation and marketing are not aligned, AI will optimise for one at the expense of the others.
Operational Self-check
AI-led inventory orchestration is less about owning algorithms and more about whether flows and decisions behave as if the network is one system. A few questions expose whether that pattern is present or absent:
- Whether forward events such as collaborations, tariff changes or big marketing pushes are reflected in buys, origin rules and node capacity three to six months ahead, not three weeks.
- Whether over-performing products or categories can be chased into within the same season without forcing blunt, margin-killing promotions on slower lines.
- Whether inventory attributes such as tariff status, channel eligibility and margin band are structured enough that AI or rules engines can meaningfully route and price stock, rather than relying on manual exception layers.
What This Pattern Signals
Taken together, the behaviour at On, Bath & Body Works and TJX points to a durable shift: inventory is no longer treated as a static buffer between volatile demand and rigid supply, but as a set of assets that can be dynamically steered by AI-informed decisions across a multi-node network. That shift changes what counts as core infrastructure. A LightSpray plant in Busan, a domestic agile fulfilment backbone, or a 5,000-store off-price network become programmable assets, activated differently as data shifts.
If this pattern persists, network and sourcing design will be framed less around fixed lead times and static tiers and more around decision rights: which elements of volume, origin, mix and allocation are locked in, which are deliberately left open for late binding, and which are governed continuously by software. As tariffs, channels and consumer behaviour continue to move, the operators that have already embedded AI-led orchestration into their origin choices, inventory positions and routing rules will find it easier to flex without tearing up their networks each time.
This is not a temporary response to a noisy couple of years. It is a structural re-weighting of where intelligence sits in the supply chain, away from periodic planning cycles and towards always-on decision layers that are already visible in how these companies buy, make, move and price.
This article is based on recent earnings reports and public disclosures from the companies referenced.
