Agentic AI transforms procurement from reactive to proactive decision-making

Artificial intelligence in procurement has travelled a long way from the era of broad promises and speculative pilots. In category intelligence, where teams must track suppliers, prices, risks and shifting market conditions, the technology is now beginning to show practical value.

According to Madhukar Irvathraya, co-founder and managing partner at Oraczen, the real change is not simply better reporting, but a move towards systems that can keep watch continuously. He argues tha...

t the future of category intelligence will be defined less by how much information teams can collect than by how quickly they can translate changing market signals into decisions.

That shift matters because traditional approaches are under strain. Procurement teams have long relied on periodic analysis, manual research and static dashboards, but markets now move too quickly for that cadence to remain effective. A sourcing plan that looked sound a month ago can be overtaken by events such as shipping disruption, supplier failure or geopolitical tension. The Red Sea crisis, for example, forced many organisations to revisit transport routes, inventory assumptions and supplier choices with little warning.

The problem is not a lack of expertise. It is capacity. Category managers are expected to oversee more areas than ever, yet much of the data available to them is never fully used. McKinsey has said less than 20% of procurement data informs decisions, a reminder that volume alone does not guarantee insight.

Another long-standing weakness is that a great deal of procurement knowledge remains hidden in people’s heads, emails and spreadsheets. Supplier histories, negotiation lessons and category-specific know-how are often scattered across systems or never formally recorded. When experienced staff leave, retire or move on, organisations can lose years of commercial memory.

This is where agentic AI is beginning to make a difference. Rather than generating one-off analyses, these systems are designed to monitor external signals, combine them with internal knowledge and surface relevant changes as they happen. PwC has argued that agentic AI can automate a large share of procurement activity, while Zoho has pointed to estimates that efficiency gains could reach 25% to 40% as routine work is increasingly handled by AI agents. IBM has similarly described AI agents as tools that can manage vendor, contract and order processes in real time.

Vendors are now pushing the idea further. Lio says it has built a multi-agent procurement system in which specialist agents handle tasks such as vendor research, negotiation, approvals and delivery tracking in parallel. Proacure, meanwhile, markets an AI-led platform that uses spend data to build sourcing workflows and identify savings opportunities. SpendConsole is applying a similar logic to accounts payable, with an AI layer that reads documents, answers questions and supports finance and procurement workflows.

What links these offerings is a broader shift away from occasional intelligence and towards continuous, reusable insight. The appeal is not simply speed, but consistency: more categories can be monitored at once, knowledge can be shared across the organisation, and procurement teams can spend less time assembling information and more time acting on it.

Just as importantly, the technology is becoming easier to adopt. Subscription-based platforms are lowering the barrier to entry, allowing firms to use agentic AI without embarking on lengthy development projects. That is helping turn procurement AI from an abstract ambition into something closer to an operational tool.

For Irvathraya, that is the point at which the hype starts to become real. The value of AI in category intelligence will increasingly be judged not by the amount of data it generates, but by whether it helps organisations respond to market change in time to make a difference.

Source: Noah Wire Services