What is Defining the Future of AI in Supply Chains?

Industry leaders from Baringa, Amazon and Infosys took to the stage at Supply Chain LIVE: The London Summit, hosted at the QEII Centre in Westminster, London, on the 8th and 9th of September to discuss the future of AI in supply chain.

Sameer Mohammed, Partner and Supply Chain Practice Lead at Baringa, Umesh Rangasamy, Leader of Private Brands Supply Chain Technology at Amazon, and Ambeshwar Nath, Executive Vice President and Industry Head of Consumer Goods, Retail and Logistics at Infosys, examined where human oversight remains necessary despite automation advances.

The discussion centred on bridging the gap between AI potential and current operational realities in supply chain management.

Practical AI application in supply chain operations differs from theoretical discussions of autonomous systems. Current deployments provide value through the system of intelligence layer, which processes disparate near-real-time data to inform decisions.

According to Sameer, expectations around fully autonomous operations exceed current capabilities. "Where I think the hype is maybe too much at this moment in time, is this idea of lights-out and the idea that you can basically run the entire thing without human beings, or at least with a human in the loop to make sure that the AI is effectively doing the right things," he says.

Three barriers prevent full automation in supply chain operations. Legal frameworks require human accountability for decisions, the technology still needs context for accurate application, and data quality in many organisations remains insufficient.

Umesh described ongoing experimentation at Amazon in autonomous supply chain responses. "We keep talking about autonomous supply chain where it's just reacting to situations. Supply chain is one of those challenging and also interesting areas where AI can work and make a big difference because things keep changing on a daily basis. So, we're looking at where AI can react to that fast" he says.

According to Ambeshwar, AI implementation at Infosys reduced client supply chain planner requirements by 30% to 40%. The technology could shift planner workload allocation from 70% reporting and 30% execution to the reverse ratio.

Amazon achieves demand prediction accuracy between 80% and 90% across its retail network, according to Umesh. This accuracy range reflects current capabilities in large-scale supply chain forecasting.

AI-first engineering has changed software development timelines in supply chain systems. Legacy system updates now require less time and lower costs, according to Umesh.

However, complete elimination of human oversight in supply chain operations remains impractical. "It's only when you trust the intelligence and say, 'Look, take action independently.' That's where system engagement comes in," Ambeshwar notes.

Enterprise organisations often attempt to scale pilot programmes without addressing underlying friction in supply chain operations. The panel identified this approach as a common failure point.

Successful AI integration requires focus on business outcomes rather than technology strategies. Strong data foundations, cross-departmental alignment and executive sponsorship support effective deployment in supply chain environments.

"The question has moved now from, 'What is my AI strategy? What is my AI supposed to do?' to, 'This is the business outcome I want. How can AI help me realise those goals?' The token economics are becoming clear, the cost efficiencies don't seem as apparent once the technology scales up. So, all those parameters are having clients really look at where can you apply it, and do one thing, but do it deep," Ambeshwar says.

Umesh agrees that defining specific business outcomes prevents unfocused AI deployment. "Identifying the business outcome or KPI is key because then that will bring focus," he says. "Otherwise, what happens is AI is going to go all over the place. It's going to try to come up with its own metric."

Out of 330,000 employees at Infosys, 250,000 have received formal AI training, according to Ambeshwar. This training programme could show the scale of workforce preparation required for AI integration in supply chain operations.

According to Umesh, AI models bring solutions to 80% completion quickly, but the final 20% requires substantial human effort. This pattern applies across supply chain technology deployments.

New AI systems in supply chain operations should launch with initial accuracy rates between 80% and 85%, according to Ambeshwar. Expecting immediate perfection creates unrealistic deployment parameters.

Strategic workforce planning for AI integration requires organisations to look five to ten years ahead, according to Sameer. This timeline addresses upcoming skill gaps in supply chain management.

AI automation of routine reporting and data aggregation changes supply chain professional responsibilities. The shift moves workers from task execution to strategic oversight functions.

Human roles in AI-enabled supply chain operations focus on relationship building and cross-functional collaboration. Capacity freed by automation allows for strategic problem solving rather than eliminating positions.

The panel indicated that supply chain professionals need different skills as AI handles data processing tasks. Long-term planning and contextual decision making remain human responsibilities in automated environments.

Source: Supply Chain Digital