b'For an AI system to be truly agentic, it must deeply un-Why Agentic AI May Herald Golden Age for Procurementderstand its environment. This means perceiving external conditions, processing contextual data, and dynamically determining the best courses of action in real time. Be-yond reactive responses, agentic AI actively sets and pur-Keith McFarlane is the chief technology officersues goals, ensuring its actions align with broader strate-at Palo Alto, California-based Globality, agic objectives. provider of artificial intelligence-enabledprocurement automation software. Views areMemory is also critical. An agent must retain past interac-the authors own.tions and experiences to refine its decision-making over time. By leveraging memory, it can continuously improve its performance, adapt to changing circumstances, and With the rise of agentic artificial intelligence, I believeenhance its ability to serve users effectively. were heading for a golden age in procurement. Beyond reasoning and decision-making, agentic AI inter-Research by HFS shows that AI-driven sourcing al- acts with the world through various input and output ready delivers 20% cost savings, making it one of themechanisms. This can include processing chat-based que-lowest-risk, highest-reward applications of AI for enter- ries, analyzing sensor data, responding to time-based trig-prises.gers, or accessing external services via application pro-gramming interfaces, or APIs. Agentic AI represents a giant leap forward, helping pro-curement intelligently select high-value suppliers, proac- The ability to take meaningful actions, whether engaging tively manage risk, and customize buying for differentin conversation, setting reminders, or initiating work-product categories and markets. This is particularly im- flows, is what distinguishes truly agentic AI from tradi-portant now, as business leaders focus on optimizing bothtional automation. short-term profitability and sustainable business value in a turbulent economy.Benefits for procurement Procurement is an ideal use case for agentic AI because Age of agents becoming realityits inherently complex, data-driven and prone to human While the age of agents concept is gaining traction, itserror. No procurement team has enough people to engage origins date back decades. Early AI research in the mid- every business stakeholder as quickly, easily and effec-1950s led to innovations like Shakey the Robot, a 1960stively as the CFO and CEO demand. The complexity is system demonstrating basic decision-making. Thesetoo great, and time is too limited.laid the foundation for intelligent agents in robotics and automation.Agentic AI can transform procurement in many ways, including supplier discovery, benchmarking, proposal Advancements in autonomous computing have since ena- analysis and negotiation. bled systems like the Mars Perseverance Rover and Al-phaGo. Now, generative agents, powered by large lan- Negotiation can be particularly tricky in that it requires guage models, are not just learning but actively executingbalancing price, trends, scope, timeline, talent and meth-complex business tasks with remarkable efficiency.odology. Some excel at it, while others avoid it. An agent can evaluate proposals holistically, provide tailored nego-Its important to first distinguish agentic AI from othertiation strategies, and suggest concrete actions based on forms of automation. True agentic systems possess sever- all relevant data. al defining characteristics: they are goal-oriented, autono-mous, adaptive, and interactive. This means they operate with a clear purpose, make independent decisions within defined constraints, adjust to evolving environments, andWhat sets agentic AI apart from previous technological communicate effectively with users and other systems.innovations that have impacted procurement is its ability to autonomously conduct multi-round negotiations, adher-Agents vs. traditional botsing to predefined rules. This ensures the best possible out-In contrast, traditional bots, narrowly focused AI applica- comes without sacrificing control or flexibility. tions, and early workflow automation tools like first-wave robotic process automation do not necessarily qualify as agentic AI. While they may perform useful tasks, they lack the dynamic decision-making, contextual awareness, and adaptability that defines true agency. 16'