PROCUREMENT MAGAZINE WHITEPAPER
Multi-agent orchestration in practi
Effective orchestration does not depend on a single, monolithic AI. It relies on networks of specialised agents, each responsible for a defined domain of expertise. One agent may interpret intake requests and classify intent. Another may apply global and local policy controls. A third may evaluate supplier risk. A fourth may reconcile purchase orders against contracts and invoices.
These agents do not operate in isolation. They share structured data and validate each other’ s outputs. The orchestration layer coordinates their activities, ensuring that processes unfold logically and that compliance rules are applied consistently.
At Bayer, this architecture underpins what the company describes as its procurement“ front door.” Operating across pharmaceuticals, consumer health and crop science with more than 88,000 suppliers and hundreds of spend categories, Bayer faced the challenge of balancing strict governance with ease of use.
By partnering with ORO, Bayer reimagined procurement as a coordinated network of human and AI agents. Employees submit requests through a single entry point. Behind the scenes, agents
An AI agent must be understood as a job, not a feature. It has responsibilities, data, outputs and performance measurement, just as a human role would. This distinction ensures accountability. It also prevents the common mistake of rebranding simple automation as‘ agency’
Lalitha Rajagopalan, Co-Founder, ORO Labs
110 March 2026