Why Your Procurement Process Needs Agents, Not Just Automation
Your Guide to Getting Started with AI Agents in Procurement
Why Your Procurement Process Needs Agents, Not Just Automation
In 2018, I stepped into a new role responsible for a supply chain valued at £30-35 million. At least, that’s what the paperwork said. The problem was I couldn’t trace how they’d arrived at that figure. No visibility into the timeline, the process, nothing.
Eventually, I found the answer in a room roughly 20 feet by 20 feet. The entire supply chain process had been mapped across all three walls, extending even above the windows. A few people had created this massive visualisation, and apparently, this was considered the normal way to approach things.
Looking at each component was mind-boggling. They’d created an absurdly complex system attempting to map every possible condition, creating a monstrous process that could never achieve its goals. Different stages for each supplier, grouped around later activities like user acceptance testing or onboarding. Yet it missed fundamental elements like financial checks.
The process treated every supplier identically. Some were simply providing standard, commercially available off-the-shelf goods. Others were developing, building, and machining entirely new solutions. Same treatment. Equal distribution of attention to unequal levels of risk.
I’d love to say this was an anomaly. It wasn’t.
The Process Problem Hasn’t Gone Away
I still see organisations managing procurement through spreadsheets and email chains. Sending spreadsheets, receiving data, evaluating it, storing it in SharePoint. No supplier tiering. No systematic risk assessment. Limited resources spread thin across every vendor regardless of strategic importance.
The fundamental issue isn’t a lack of technology. It’s a lack of process thinking.
Modern procurement applies a risk lens to every supplier. Tier A suppliers (high spend, strategic importance, access to sensitive data) get intensive review. They’re involved in your direct procurement, the projects you’re working on and products you’ll sell. These suppliers determine your contract clauses, due diligence approach, performance management strategy.
COTS commodity vendors? Streamlined checks. Appropriate attention for their actual risk profile.
This tiering determines everything else. Without it, you’re just adding technology to chaos.
Automation Isn’t Enough
Here’s where I see teams go wrong: they hear “AI” and think it’s the solution to their procurement problems. It’s not. At least, not directly.
Even basic robotic process automation with simple workflow triggers does much of the heavy lifting for well-designed processes. The technology isn’t the constraint. The process is.
The last thing I want is to deploy an agent across poor workflows and unnecessarily complex processes that don’t need to be as convoluted as people often make them. You’ll just accelerate bad decisions.
Before adding AI layers, especially agents, we must optimise how we work with suppliers. Get the foundation right first.
Where Agents Actually Change the Game
With that caveat firmly established, let me tell you where agents genuinely transform procurement operations. Because I’ve been working with them firsthand, and the results have been exciting.
When we examine procurement through a job-to-be-done lens, we discover remarkable opportunities to deploy agents that can think, review, reason, and make decisions, not for us, but before us. This distinction matters.
Due Diligence at Scale
Consider a typical supplier onboarding process. For strategic suppliers, we might ask between 50 and 250 questions. That’s substantial content to review carefully. Suppliers provide various policy documents: information security protocols, risk management strategies, ISO certificates, SOC certificates, business continuity plans, disaster recovery procedures.
This review process takes hours per supplier. When onboarding multiple suppliers simultaneously, the workload becomes immense. Meanwhile, stakeholders are eager to finalise contracts and get moving.
Now, an agent can review all these responses and documents in about 30 seconds. The output is a comprehensive assessment identifying strengths, weaknesses, and areas of concern, along with recommendations to accept, reject, or request additional information.
We have customers implementing this technology right now. They’re reviewing vast amounts of data in seconds and receiving approval or rejection recommendations that would have taken a procurement manager hours to produce. Frankly, the agent’s review is more thorough than most human reviewers can deliver under time pressure.
🎬 See it in action: AI-Powered Supplier Onboarding Demo
Pre-Approval Validation
Before requests even reach human approvers, agents validate whether they meet minimum criteria. Budget confirmed? Business case complete? Required information present?
The agent runs a “what” check, a “why” check, a coherence check. This isn’t replacing human judgment. It’s ensuring humans spend their judgment on decisions that actually require it, not on sending requests back for missing fields.
🎬 See it in action: Due Diligence & Risk Assessment Demo
Financial and Compliance Checks
Bank details validation prevents payment fraud by checking vendor banking information against documentation. Finance reviewers validate billing frequencies and GL coding against approved budget codes. Cybersecurity reviewers check security clauses against internal InfoSec standards.
Each review takes about 12 seconds. Each saves a human from repetitive verification work that’s critical but mind-numbing.
The Numbers
At Gatekeeper, we now have over 50 agents covering contract and vendor workflows. The math is straightforward:
Human review time for our processes totaled 223 hours. Agents, working on workflows, generating summaries, extracting contract data, save 220 of those hours. We’re confident enough in this approach that we guarantee customers will achieve at least 3x their Gatekeeper spend in FTE savings from agents alone.
What’s most remarkable is that the current benchmark for these agents represents their worst performance level. They’ll only improve from here.
Human-in-the-Loop Is Non-Negotiable
I want to be clear about philosophy here. We value agents highly at Gatekeeper, but we don’t want to offload our thinking entirely.
Agents handle the preliminary analysis. They consume and analyse large amounts of data to support business-critical decisions. They surface recommendations. Then humans validate, adjust, and make the final call.
This isn’t autonomous decision-making without oversight. It’s augmented decision-making with better information, faster.
The agent reviews the 200-question due diligence pack and tells me: here are the concerns, here’s what looks good, here’s my recommendation.
I still decide. But now I’m deciding with comprehensive analysis in hand rather than skimming a document at 10pm because I’m behind on onboarding three suppliers simultaneously.
What This Means for Your Team
When agents handle the administrative burden, procurement teams escape firefighting mode. More time for strategic projects that deliver competitive advantage. Faster business velocity: suppliers onboarded in days, not weeks. Better coverage: every supplier reviewed to standard, not just the ones you had bandwidth for.
But none of this works if your underlying process is the three-walls-of-chaos approach I inherited in 2018.
So here’s my challenge: Before you evaluate any AI or agent technology, map your current procurement process. Not the aspirational version. The real one.
Ask yourself:
Do we tier suppliers by risk?
Do we allocate review effort proportionally to risk?
Are we collaborating with risk, InfoSec, legal, and finance, or operating in a silo?
Is the process documented, or is it tribal knowledge?
Fix that first. Then deploy agents against a solid foundation.
The wall-mapped nightmare I walked into seven years ago didn’t need better technology. It needed better thinking. The agents we deploy today are powerful precisely because they operate on processes worth automating.
What does your procurement process look like on paper versus reality? I’d love to hear the war stories. Reply and let me know.
Yes, this aligns 100% how we are approaching automation and AI agent deployments.