AI IN PROCUREMENT
“ RPA is incompatible with messy data and ambiguity. It needs highly structured inputs and a predictable path to operate properly, or it runs into issues,” says Rakesh.
Similarly, RPA has zero understanding of business context, so any slight anomaly or process deviation breaks the workflow and requires manual intervention.“ RPA is a functional sticking plaster and more of an interim solution to spur immediate efficiency until a more robust, scalable architecture is in place. But you still need human eyes – or smarter, cognitive AI – to catch the warning signs when someone is trying to cheat the system,” adds Rakesh.“ This is the point where we advise transitioning from traditional bots to Agentic AI. We say this because Agentic AI can interpret unstructured data, manage exceptions autonomously and execute complex, reasoning-based decisions.”
Agreeing with Rakesh, Joel explains that the process becomes too complex for standard RPA when it stops being a clear set of rules and starts requiring judgment. Procurement is rarely one uniform process. A bot can move information quickly, check fields, route documents and apply rules, but it cannot fully determine the right acquisition strategy just because the data is available.“ The clearest sign that RPA has reached its limit is when exceptions become the normal workflow. If the bot constantly needs human intervention, then the process needs professional judgment, better process design or a more advanced technology layer,” says Joel.
procurementmag. com 107