Why agentic procurement orchestration matters now
By 2030, 90% of procurement reviews will be done by AI. Here’s what that means.
Five months ago, at the first Zip Exec Summit in New York, we started a conversation about the future of procurement. Since then, the world has changed economically, geopolitically, and most notably, technologically.
In these few months, the pace of AI innovation is forcing a profound shift in how we think about work in general, and by extension, the work we do in procurement.
Most teams aren’t doing the work that matters
When I talk to procurement teams, here’s a question I always ask: How did you spend your time yesterday?
Was it meaningful work, like strategic alignment, stakeholder engagement, supplier innovation? Or was it chasing approvals, entering data, and reviewing 50-page contracts for the tenth time?
For too many, it’s the latter. In fact, one procurement leader at a major bank told me: “Sourcing doesn’t source.”
That quote stuck with me. It speaks to a deeper misalignment between what procurement is capable of and what it’s actually doing day-to-day.
But we know what top-performing teams should be doing. According to Deloitte’s 2024 CPO study, the highest-performing procurement teams, what they call “Orchestrators of Value”, focus on:
- Aligning procurement with stakeholder goals and business metrics
- Designing more strategic operating models
- Building high-performing teams and supplier relationships
- Collaborating upstream on innovation and risk mitigation
This is what procurement wants to be doing. This is the work that creates value.
But most teams aren’t spending their days like this.
This disconnect between the work procurement should be doing and the work it gets caught actually doing is frustrating. It’s expensive. It delays innovation, drains morale, and limits the functions' influence.
Agentic AI can help.
AI is finally good at the work holding procurement back
Over the past two years, AI has reached a new threshold of capability, especially when it comes to the tasks that bog procurement teams down. We’re entering what AI researchers call the jagged frontier: when AI can do some tasks extremely well, and others not at all.
To me, exploring this ‘jagged frontier’ is a welcome opportunity. It allows us to ask the question more deeply: what do we want the core work of procurement to be?
Because now is the right time to be defining that for the next five years.
The reason is because AI agents are now an accessible and reliable part of the procurement process. An AI agent is a system that can apply reasoning to complete skills and tasks on your behalf, and do so reliably, effectively, and convincingly.
Let’s be clear: AI can’t replace human taste, strategic thinking, or complex negotiations. But there are some things it can absolutely do, today:
- Compare documents for discrepancies
- Validate intake forms and match them to submitted files
- Conduct vendor risk assessments using structured and unstructured data
- Generate content for policies and RFPs
- Retrieve and synthesize relevant regulatory requirements (like GDPR or DORA)
These are actual, tested and validated use cases that are in the market now, well beyond what was possible just two years ago.
But AI alone is not enough: the answer is orchestration
There’s a popular (but misleading) idea that AI agents will function like virtual assistants, ie. one super-agent you talk to that can do everything. In reality, the most successful implementations rely on specialized agents, each designed for a specific, high-leverage task.
At Zip, we’ve launched 50+ of these agents, purpose-built for tasks like intake validation, regulatory review, legal redlining, and more. But to enable AI to take action instead of merely deliver insight, we need a key ingredient: orchestration.
To complete a task, not just analyze it, an AI agent needs to be embedded in an orchestrated process. It needs access to:
- Contextual data from integrated systems
- Workflow triggers to initiate tasks
- Guardrails for human review and approval
- Decision history to improve over time
Orchestration is what makes AI useful. It’s the difference between an assistant with good ideas and a partner that gets things done.
A shift from human-initiated to machine-initiated workflows
Today, every step of the procurement process starts with a human. A stakeholder submits a request. A buyer validates it. A legal team redlines the contract.
But in an agentic future, every step can begin with an agent.
Not in a black-box way, but in a collaborative, human-centered way. Agents pre-populate information, flag discrepancies, and surface risks, so humans can make faster, better-informed decisions.
Imagine a workflow like this:
- An AI agent collects and pre-validates intake data
- Another summarizes key legal risks in a contract
- A third assembles a compliance packet before the human reviewer even opens the file
Now, instead of doing the work, humans are reviewing, deciding, and optimizing.
We call this model human-run, machine-initiated. And it fundamentally changes the way procurement approaches work.
What the future looks like (starting now)
At Zip, we believe this is the future. By 2030, we expect to process over a billion procurement reviews annually. Of those, 90% will be either fully completed or significantly advanced by AI.
This is an incredibly positive thing. This level of automation won’t diminish the role of procurement, but elevate it.
It will free teams to move upstream, to advise on make-vs-buy decisions, to influence forecasting and budgeting, shape risk management strategies, and build supplier ecosystems. In other words, it will shift procurement from managing the inputs of a process to influencing the outcomes of that process.
And that, finally, is work that moves the needle.
Curious what this could look like in your own workflows? Book a demo, and I’d be happy to talk you through how we’re bringing agentic procurement orchestration to life.
