Supply chain AI transforms from cautious pilot to strategic driver amid emerging security risks

Two years ago, many procurement leaders approached artificial intelligence cautiously. Today some are moving from scepticism to selective trust, deploying systems that augment decision-making while they contend with unresolved risks in the wider supply chain ecosystem.

At Hexion, the chemicals group that last December completed the acquisition of Smartech, the transformation is tangible. According to the company announcement, the deal was designed to weave AI-driven autonomous ...

manufacturing capabilities into Hexion’s operations and customer offerings in order to lift efficiency, cut waste and improve product quality. The shift in mindset is reflected in comments from Hexion’s chief procurement officer, Gael De Martelaere, who warned that waiting for flawless AI could leave companies behind: “somebody else is way ahead of you.”

Practical wins are emerging across sectors. Manufacturers, logistics firms and even mattress makers report measurable benefits when AI is focused on narrow, well-defined tasks. Essentia Organic Mattress’s founder Jack Dell’Accio built bespoke tools that plug into his open-source Odoo ERP so sales, purchase orders and freight fluctuations feed procurement forecasts in near real time. The result, he says, is tighter stock control and freed-up cash from abandoning spreadsheet-driven planning. “Like a new employee, eventually trust [in AI] is built,” he added.

Healthcare logistics providers describe a similar pattern: limit scope, verify sources and rebuild models from current, authoritative documents to avoid dangerous drift. Mercury’s CEO Josh Medow described how reliability improved only after isolating the chatbot from broader datasets and retraining it on verified, up-to-date material. “We’ve really limited it,” he said, explaining that tighter constraints produced more dependable answers for logistics coordinators. Even so, Medow stopped short of full autonomy for complex, multi-party decisions: “I can’t see that in the near future.”

Beyond task automation, proponents argue the most valuable role for AI in logistics is early warning. Shane Curtis, chief technology officer at defence intelligence firm Artorias, said systems that ingest public feeds, social media, traffic and shipping reports, alongside internal data can flag weak points before disruptions cascade. “It’s a manpower multiplier more than anything,” he said. Blue Yonder, a supply chain software provider, echoes this emphasis on threat awareness and has been working on published vulnerability frameworks for large language models while red teaming its agents to uncover possible supply-chain exploits.

Yet those defensive efforts respond to an accelerating security challenge. Analysis from specialist platforms warns that AI introduces new classes of supply-chain vulnerabilities: model poisoning, compromised shared prompts and third-party toolchains that fall outside conventional cyber defences. According to TraxTech, the pace of AI adoption is outstripping protective measures, creating blind spots across supplier networks. That risk is compounded when organisations bolt generative or agentic systems onto legacy infrastructure without embedding them into core workflows or oversight processes, a weakness highlighted by TechRadarPro.

The gulf between potential and peril helps explain why adoption is uneven. A survey cited by IT Pro shows only around 32% of UK firms had integrated AI into supply-chain operations, reflecting barriers such as legacy systems, data quality issues and executive caution. Industry consultants warn that mistakes have direct financial consequences: missed deliveries can trigger penalties that harm top and bottom lines and damage customer trust.

For forward-leaning companies the route is disciplined and incremental. Hexion’s roadmap includes a mix of vendor tools and in-house projects; the firm is planning multiple procurement use cases over the next 18 months and is recruiting product leadership to embed AI into procurement and logistics workflows, according to its careers posting. Other firms are prioritising limited, high-adoption pilots over broad but shallow rollouts. As De Martelaere put it, deployment count matters less than use and uptake: he prefers a small number of tools that achieve full adoption over many that sit unused.

The immediate imperative for supply-chain executives is therefore twofold: harvest the tangible benefits of narrow, verified AI applications while investing in governance, security testing and integration so more ambitious agentic capabilities do not introduce unacceptable risk. As the technology matures, businesses that balance cautious implementation with proactive defence and workflow integration appear best placed to convert AI’s promise into resilient operational advantage.

Source: Noah Wire Services