AI alone won’t deliver always-on supply chains: Rewire or stay reactive

This HFS Point of View is for chief supply chain officers, COOs, and supply chain transformation leaders rewiring their operating model to close the replanning gap and build an always-on network.

Tariff regimes shift in days and demand spikes in hours, but the average enterprise still takes roughly six weeks to systemize a tariff response and two to three months to validate a demand trend. This brutal arithmetic emerged at the recent HFS–Genpact Always-On Supply Chain roundtable in New York, attended by senior supply chain leaders from pharma, medical devices, life sciences tools, specialty chemicals, animal health, beauty and consumer products, food retail, consumer appliances, luxury retail, and leading academic medical centers.

Exhibit 1: Senior supply chain leaders from life sciences, healthcare, consumer products, specialty chemicals, and luxury retail debating the always-on agenda at the HFS–Genpact roundtable in New York

Source: HFS Research, 2026

The consensus in the room was that the bottleneck in an always-on supply chain is fragmented functions, glued together with spreadsheets outside the core ERP and governed by approval cycles built for a slower world. Until enterprises rebuild the supply chain as an interoperable network with shared accountability across the enterprise and the service provider, agentic AI will keep optimizing inside silos and the always-on ambition will remain unmet.

Five insights from the room that reset the always-on agenda

The roundtable surfaced a consistent pattern across industries. Each insight pointed back to the same structural failure: the limiting supply chain operating model.

1. The replanning gap is the crisis. Executives reported tariff response cycles of around six weeks, demand validation of two to three months, and supplier requalification stretching even longer for non-trivial cases. The root cause was not AI maturity. It was a process flow stitched together across ERP systems, spreadsheets, and partner systems, with no enterprise-grade orchestration.
2. Autonomy lives in slivers, not networks. Almost no leader reported a fully autonomous function. Narrow wins such as ERP three-way matching, optimal carrier selection in eCommerce, blockchain traceability, and predictive maintenance on logistics assets exist. But each one of them is a pocket. The network around them remains manual, sequential, and approval heavy.
3. The next-up wish list is unanimous and revealing. The items here include planning and purchasing tied to live forecast and inventory, autonomous purchase order (PO) determination and dispatch, par-level replenishment in clinical settings, and last-mile carrier allocation on real-time demand. Every one of them requires the same prerequisite: a unified data spine and shared decision rights across functions and partners. None needs more advanced AI. The technology is already commercially available.
4. Leadership behavior is the under-discussed barrier. Multiple executives flagged that, even when data and tools exist, they still want individual approvals for routine decisions. The result is that agentic systems get installed as recommenders, not decision actors. This is the third stage of the HFS AI Trust Curve at work: “behavioral trust,” where data is credible and models are confident. However, humans have not yet ceded decision rights.
5. Service provider engagements are misaligned with how the work actually needs to be delivered. Most enterprises still buy supply chain transformation as multi-year programs measured in milestones. CSCOs need quarterly impact and continuous improvement on replanning latency, fill rates, and supplier variance. The commercial model is set up for projects. The operational need is set up for products.

Mount Sinai raised the bar by delivering a keynote on always-on patient care

Ed Robinson, SVP and Chief Resource Officer at Mount Sinai Health System, delivered a cameo keynote that recalibrated the conversation. In a healthcare supply chain, an out-of-stock isn’t a fill-rate metric. It’s a clinical event. Par-level replenishment failures, traceability gaps, and supplier reliability issues carry weight that industrial supply chains rarely confront with the same urgency.

The healthcare lens did two things for the room. It reframed “always-on” from an efficiency ambition into an operational floor, a baseline that other industries should aspire to. It exposed how much of the regulatory, safety, and dual-use complexity that healthcare and life sciences operate under daily is now arriving in mainstream consumer and industrial supply chains via tariff regimes, ESG mandates, and geopolitical fragmentation. The healthcare playbook is no longer adjacent to the manufacturing or retail playbook.

The always-on operating system rests on three interoperability blocks and a shared accountability model

HFS anchored the discussion alongside Genpact’s supply chain leadership, including Gopal VK, Senior Vice President and Global Head for Supply Chains. HFS framed the always-on supply chain not as a tech upgrade but as an operating system shift. The foundation is three interoperability blocks that turn a chain into a network.

1. End-to-end visibility across material, money, and information flows: Current location of inventory, status of working capital, condition of assets, all readily accessible across suppliers, production, logistics, and reverse flows.
2. Autonomous, self-adjusting behavior: One change in the system automatically triggers the required changes across the network with minimal human effort.
3. Connected ecosystem: A single source of shared data intelligence feeding internal departments and ecosystem partners, with network performance optimization as a shared responsibility and a single source of truth across supply chain, marketing, finance, and procurement, not as a data lake project but as a governed contract between functions.

On top of this foundation sit the 4Ps of agentic supply chains, i.e., productivity, prediction, personalization, and performance, which together shift the supply chain from cost center to value engine. Productivity automates the work. Prediction turns reactive into preemptive. Personalization tailors the network to each persona, including the CSCO, the plant manager, and the logistics lead. Performance ties every action to revenue protection, perfect order rate, and working capital velocity.

But the operating system only works when the accountability model is rebuilt. SLAs alone are obsolete in a world where service providers run model accuracy and self-healing rates while enterprises own perfect order rate and working capital days. The room aligned on a three-tier shared accountability model. Service-Level Agreements (SLAs) remain provider-owned and cover system uptime, model accuracy, and self-healing rate. Experience-Level Agreements (XLAs) are co-owned and cover forecast accuracy, order cycle time, supplier lead-time variance, and planning response time. Value-Level Agreements (VLAs) are enterprise-owned and cover OTIF (on-time in-full), working capital days, customer fill rate, and perfect order rate.

The connective thread is a Supply Chain Guardian, an orchestrator role that owns the end-to-end value stream, monitors process flow across the 15-plus systems most enterprises run, and ensures no demand signal is lost from forecast to delivery. The core idea is that the network gets a single point of orchestration and a single point of truth.

A three-move playbook to close the replanning gap

The roundtable conversations, the Mount Sinai cameo, Genpact’s client stories, and HFS field research converged on a playbook the leaders in the room could act on within four quarters.

1. Make the network the unit of design.
   Most transformation programs still treat planning, procurement, manufacturing, and logistics as separate domains, each with its own roadmap, budget, and copilot. Always-on requires the network as the design unit, with shared data, shared KPIs, and a single orchestrator. Until that shift happens, every new agentic capability creates faster silos, not faster networks.
   
2. Move from SLAs to VLAs through a Supply Chain Guardian (orchestrator).
   Define what the enterprise owns (VLAs), what the provider owns (SLAs), and what they co-own (XLAs). Make a single Supply Chain Guardian responsible for the connective thread. This is the structural move that turns shared accountability from a slide into an operating reality, and it is the move services providers are least prepared to deliver under existing project-based commercial models.
3. Pick one workflow and measure replanning latency instead of project milestones.
   Choose one high-value flow, which could be clinical par-level replenishment, last-mile carrier allocation, supplier requalification, or autonomous PO dispatch, and target a step-change in awareness-to-action time. Replanning latency is the single metric that exposes whether your supply chain is genuinely always-on or merely well-instrumented.

The Bottom Line: AI is an enabler to accomplish always-on supply chains. However, the operating model needs to be rejigged to incorporate the 4P success framework, process guardian accountability model, and organizational restructuring.

The replanning gap is closing for the few who are rebuilding the supply chain as a shared, interoperable, outcome-accountable system. For everyone else, every new agent just means a faster route to the same six-week answer.

The room left New York with a simpler test for every supply chain investment. Does it shorten replanning latency, broaden shared accountability, and move the enterprise up the AI Trust Curve?