News / An 'across-the-spectrum rethink' needed as firms eye use of AI in procurement

The disparity between the relentless hype and its present limitations has left most people sceptical of the promise of artificial intelligence.

Still, experts predict an imminent revolution as it takes a role in procurement and supply chain management, with some firms ‘wading into new ponds’.

Last week, Amazon announced the deployment of AI in 1,000 of its electric delivery vehicles after early tests with vision-assisted package retrieval technology, which identifies packages to be delivered at the next stop, and produced average time savings of 30 minutes per route.

Today this type of application – used within a narrow scope in operations and for administrative tasks – characterises practical uses of AI, rather than the grandiose visions of intelligent systems that do all your work and serve a coffee to boot. However, some experts predict a quantum leap within the next 18 months that will transform procurement, with far-reaching implications.

“Procurement will look very different,” declared Remko van Hoek, professor of supply chain management at the University of Arkansas, at the Digital Procurement World (DPW) conference this month.

“It requires an across-the-spectrum rethink.”

DPW’s 10X Procurement study, based on input from over 200 international procurement leaders, indicates that respondents are planning “a transformative leap forward in digitisation of strategic sourcing and supplier relationship management” that aims to move “beyond the traditional focus of peer-to-peer automation”, said Prof van Hoek.

The study predicts adoption of AI is set to grow 187% in 2025, up from about 20% this year.

At the same event, Kris Timmermans, lead of Accenture’s supply chain & operations, stressed the need for advanced AI capabilities in procurement and supply chain management. He said: “The growth strategy of most companies is increasing the complexity of supply chains,” which, he added, was impossible to mange efficiently without AI support.

“It’s humanly impossible to optimise this,” he said. “We need continuous supply chain planning and execution.”

This points to autonomous sourcing on the back of AI-driven platforms that engage with system users, review documents and produce RFPs, to name just some anticipated functions.

Procurement consultancy and tech provider GEP predicts that AI agents will handle functions like order placement and supplier communication autonomously, and not only anticipate potential disruption through predictive analytics, but also recommend proactive measures, suggesting alternative suppliers, autonomously rerouting shipments and renegotiating contract terms. This extends to functions like monitoring commodity prices and suggesting adjustments to procurement strategies.

Some companies are moving in this direction: retailer JC Penney is implementing machine learning in its pricing and assortment planning and has begun to leverage AI in logistics to determine whether to dispatch orders from a store, a fulfilment or a distribution centre. And management has the full supply chain in its sights.

Overall, AI adoption is still very low, as companies wait for more evidence of tangible advances that would close the gap between anticipated functionality and current limitations of applications. Observers have likened the rise of AI to the emergence of the internet and the dot-com bubble, which saw a vast amount of hype that imploded, taking down about 95% of the early players, but real progress subsequently building from the remaining 5%.

Boardrooms are sceptical about the ROI of early ventures into AI, so far there has been just a trickle of results from early adopters. Fidelity Investments reported that the implementation of autonomous sourcing capability had cut contract negotiation time by 50% and resulted in an average overall cost reduction of 20%. The software has been able to spot instances of “contract padding”, leading to price adjustments.

There also appear to be benefits for suppliers, such as being included in RFPs they may have missed, and more transparency of the bidding process, allowing them to better assess their chances of winning a contract.

Besides scepticism about AI’s actual capabilities and ROI, there are question marks over its fit into existing IT infrastructures. According to one report, only one in five AI projects in the manufacturing sector has been successful. Legacy technology – chiefly ERP, material requirements planning and manufacturing execution systems – poses a huge obstacle that complicates, and sometimes frustrates, attempts at AI implementation.

GEP advises companies to prepare for AI implementation projects by ensuring data quality, establishing and embracing standards and ensuring employee buy-in.

But it also cautions that firms should move deliberately through a phased implementation, rather than try to do too much in one go. It is best to start with small pilot programmes where AI agents can deliver immediate value, it said. This can provide valuable insights into how AI agents will fit into the broader organisation.