Why most enterprise AI supply chain bets still aren’t paying off
Nearly nine in 10 operations leaders say their technology investments haven’t delivered the value they promised — even as agentic AI reshapes what’s technically possible on the factory floor and in the freight yard.
Supply chain leaders have spent the better part of three years living inside a crisis simulator. The COVID-19 shock exposed how brittle single-source sourcing strategies really were. The Russia-Ukraine war rattled energy markets. Most recently, the Middle East crisis has choked one of the world’s most critical trade arteries: the Red Sea and Gulf of Persia corridor that carries a quarter of the world’s oil trade through the Strait of Hormuz, according to a new report from PwC India, Demystifying AI in the ever-evolving supply chain landscape, published this month.
The numbers behind that disruption are stark. The Suez Canal, which normally handles 12% to 15% of global trade, is now seeing transit volumes drop below 70% of 2023 levels as shippers reroute around the Cape of Good Hope, the report notes, citing UNCTAD data. Add in a Middle East region that supplies 30% of global oil production and 17% of natural gas, and it’s little wonder that supply chain executives have stopped treating volatility as an occasional disruption and started treating it as the operating environment itself.
That’s the backdrop against which agentic AI — AI systems that don’t just recommend a decision but reason through it and act on it — is being pitched as the next operating model for supply chains. PwC’s report, based on a survey of more than 150 senior executives across five industries, paints a picture of an industry that’s genuinely excited about the technology’s potential but still very much stuck in first gear on execution.
From dashboards to decision-makers
The pitch for agentic AI is that it moves supply chain technology beyond the passive monitoring that has defined enterprise resource planning systems for decades. Rather than simply tracking procurement, planning, and logistics after the fact, agentic systems are designed to identify disruptions across interconnected data, reason about likely outcomes, and then plan and execute a response — largely without waiting for a human to sign off, the report explains.
In practice, PwC lays out four areas where this is already playing out:
Planning — AI agents continuously analyze customer behavior, supplier performance, and risk signals to spot disruption patterns before they hit, then stress-test inventory policies and replenishment strategies against simulated scenarios rather than relying purely on historical demand data.
Manufacturing — Agents ingest real-time shop-floor data from IoT sensors to monitor equipment health, flag maintenance needs before a breakdown occurs, and adjust production schedules on the fly as demand shifts.
Logistics — Control-tower agents reoptimize transportation routes and network design in response to capacity constraints, regional disruptions, and demand swings, improving visibility into lead times and inventory replenishment.
Procurement — Perhaps the most autonomous use case in the report, agents are increasingly handling end-to-end sourcing: consolidating purchase requests, running sourcing events, negotiating with suppliers against a set price baseline, and even drafting contracts and monitoring compliance obligations.
Taken together, PwC describes this as a shift from AI that supports better decisions to AI that increasingly makes and executes them — a distinction the report treats as the real dividing line between “AI” and “agentic AI” in a supply chain context.
The adoption numbers tell a more cautious story
For all the enthusiasm around what agentic AI can theoretically do, PwC’s survey data suggests most organizations are still circling the runway. Forty percent of respondents say they’re still in early discussions and trying to understand GenAI’s relevant applications, and 36% report no immediate plans to adopt it at all. About a third have a deployment roadmap in place, and 44% have gotten as far as pilots or proofs of concept — but only a small fraction have use cases actually delivering tangible value in production.
Where organizations are choosing to point their early AI investment matters, too. Supply chain risk monitoring and automated supplier and customer communication top the list of priority sub-functions, followed by warehousing and logistics, procurement, and planning. Seventy-seven percent of respondents said AI and machine learning specifically help improve transparency across the supply chain, and 74% pointed to data integration and availability as an even more critical lever.
The report is candid about why so many initiatives stall or get paused altogether. Budgetary constraints and data security concerns top the list, cited by the large majority of organizations that have hit pause on GenAI projects specifically, with insufficient data quality and availability close behind. It’s a familiar pattern for anyone who has watched enterprise technology cycles before: the ambition is real, but the underlying data and governance infrastructure hasn’t caught up.
The bigger gap: ambition versus realized value
The most pointed finding in the report doesn’t come from the supply chain survey at all — it comes from PwC’s companion 2026 Digital Trends in Operations Survey, which the supply chain report leans on to make its central argument. Eighty-five percent of leaders in that survey believe they’re ahead of competitors on digital transformation. Yet 89% admit their technology investments haven’t delivered the value they expected. Only 27% say they’ve fully embedded an AI-driven strategy across business units, and just 37% say they’re comfortable letting AI agents run an end-to-end process without a human checkpoint.
PwC’s read on that gap is blunt: most organizations remain trapped in pilot mode, and the industry’s yardstick for success needs to shift from “has AI been deployed” to “what measurable business impact has it delivered.”
What CXOs are being told to do about it
Rather than treating this as a technology gap alone, PwC frames the fix as an organizational readiness problem, laying out five prerequisites for organizations hoping to scale AI beyond isolated pilots: sustained leadership commitment, a genuinely connected digital ecosystem rather than siloed point solutions, tight alignment between business and technology teams on use-case definition, formal governance frameworks, and — often underweighted — AI literacy among the people who are meant to act on an agent’s recommendations.
Data maturity underpins nearly all of it. In the companion Digital Trends survey, 87% of operations leaders said poor data quality has directly hampered their ability to extract value from digital initiatives — a finding that echoes across nearly every AI maturity study published in the past two years, but one PwC argues supply chain leaders still haven’t fully internalized.
The report’s recommendations to CXOs boil down to three priorities:
Fix the digital foundation first. Fragmented ERPs, siloed data lakes, and inconsistent master data are named as the primary structural barriers to scaling AI. PwC’s advice is to treat digital foundation work as a multi-year capital commitment rather than a one-off IT project, and to build cloud-native, API-driven infrastructure specifically so AI agents can interoperate across functions.
Anchor every AI investment to a business metric. The report pushes CXOs to ask, before funding any AI initiative, which metric will move, what the baseline is, and whether the organization’s data and processes are mature enough to support scaling. It’s a discipline aimed squarely at preventing the kind of pilot-purgatory the survey data suggests is already widespread.
Treat governance as a value driver, not a compliance checkbox. As decision authority shifts from human-reviewed recommendations to autonomous execution, the report argues that the margin for error narrows sharply — an agent’s flawed forecast can trigger a bad procurement order before anyone notices. PwC cites its own 2025 Responsible AI survey, in which 60% of executives said responsible AI practices actually boost ROI and efficiency, and 55% linked them to improved customer experience.
For Indian enterprises specifically, the report points to a growing regulatory patchwork to navigate, including the Digital Personal Data Protection Act 2023, MeitY advisories, and sector rules from bodies like the Directorate General of Foreign Trade — alongside international frameworks such as the EU AI Act and the U.S. NIST AI Risk Management Framework for globally operating firms.
The payoff for getting there first
PwC’s closing argument is less about the risk of inaction than the size of the prize for moving early. Citing its separate Reinventing Supply Chains 2030 research, the report projects that organizations pursuing enterprise-wide supply chain transformation — rather than isolated function-by-function pilots — are positioned to unlock cost reductions approaching 19% and revenue gains of roughly 16%.
The technology itself, the report’s authors argue, is no longer the bottleneck. Cloud infrastructure, IoT-driven real-time data capture, and large language models have made AI meaningfully more accessible than it was even two years ago. What’s holding organizations back, according to PwC, is a much older and more familiar set of problems: fragmented data, thin governance, and leadership teams that haven’t yet made the multi-year commitment agentic AI actually requires.