Procurement leaders are under pressure to show where AI can deliver more than individual productivity gains. Drafting an RFP faster or summarizing a contract can save time, but those improvements rarely change how procurement operates.
That is where agentic AI raises the stakes.
Technology providers are increasingly introducing multiagent systems that make it possible for AI agents to pass off tasks and move procurement work across systems, whether guiding intake requests or supporting supplier onboarding and data handoffs to other systems. These capabilities are promising, but they require reliable data and clearly defined processes. Those are two areas where many procurement organizations continue to struggle.
This is a factor in why Gartner predicts that through 2027, only 20% of procurement organizations will have sufficient data and process maturity to use multiagent systems. Organizations that are ready for this next step will be positioned to gain an advantage because rather than single task-based AI agents, multiagent systems can help teams operate with greater speed, while also providing strategic value to the business.
The barrier today is not a shortage of AI ambition. It is the condition of the data and processes that AI depend on.
Preparing to scale Agentic AI
Many procurement teams are already seeing value from narrow AI use cases. A contract team may use AI to take a first pass review of a contract and identify deviations from the contract standards. These applications are often successful because the work is contained within a specific task and relies on a narrow set of data.
Scaling agentic AI is different. The moment an agent is expected to move across workflows, the weaknesses in procurement data become much harder to hide.
Take supplier risk monitoring. On paper, a supplier may appear low risk because it passed onboarding two years ago and still has an active profile in the sourcing system. Yet that view may be incomplete. The contract repository might show that the supplier's insurance certificate has expired or the quality system may reveal an increase in defects. If those and other signals are not connected, an AI agent may evaluate the supplier based on only a portion of the available information. This is why procurement data maturity matters.
The productivity problem is also a process problem
Data is only part of the challenge. Procurement also needs to rethink how work moves through the organization.
A Gartner survey found that only 36% of chief procurement officers are very confident in their ability to redesign the procurement function for AI. At the same time, organizations are seeing productivity gains from GenAI in areas such as time savings, output and quality. The challenge is that those improvements are not consistently translating into broader business results.
For example, a category manager may create a first draft of a strategy faster, but if the surrounding workflow does not change, the organization may simply move the bottleneck somewhere else.
To address the challenge, procurement leaders should start by identifying where decisions stall and what steps need to be changed. Understanding those gaps can reveal opportunities where AI agents can support the broader workflow rather than accelerating only one task within it.
Improving procurement intelligence
Procurement does not need perfect data before using AI. Waiting for perfection would stop progress altogether. Leaders do, however, need a realistic view of where AI is ready to operate and where foundational gaps remain.
For a process such as supplier onboarding, procurement needs to understand what information an agent would require, where that information resides, who maintains it and what actions the agent is authorized to take. This work is less exciting than launching another AI pilot, but it determines whether pilots evolve into sustainable operating capabilities.
The human role will also change. Buyers and category managers will need to understand why an agent flagged a supplier and what evidence influenced the recommendation. For example, an agent might identify a supplier as a potential risk based on late deliveries or recent financial signals. Procurement professionals will need to evaluate whether those signals tell the full story or whether important context is missing.
That requires more than learning how to write prompts. It requires strong data literacy and the confidence to challenge AI-generated recommendations when necessary.
Agentic AI offers significant promise for procurement. Yet the organizations that realize the greatest value will not be the ones that launch the most pilots. They will be the ones that improve data quality, connect critical workflows and establish clear governance around how decisions are made.
Before scaling AI agents, procurement leaders should ask a blunt question: Would they trust today's data and workflows enough to let AI act on them?
For many organizations, that answer is still no. And that is exactly where the work needs to begin.
