AI Agents: Beyond the Hype - What Really Works [in Procurement?]

Here I am again with another article about AI, and you might be thinking - "Not another piece about AI agents!" 😊 But this time we're going to cut through the hype and look at what's really happening in the world of AI agents, especially in procurement.

Over the last few days, I've read a few articles on the subject of agents from Anthropic and Hugging Face. I've noticed that we often don't use the term "agent" correctly - I'm one of those people myself. You may remember how often I wrote about AI Agents in Procurement. But now I probably need to revise my words.

Another reason for this article is my personal experience in building an AI app as an agentic setup. I used a very well-known Agentic framework for this. First of all - I failed and I don't see this as a defeat but as a very important learning. The point is: I wanted to prove to myself that agentic setup is the best. But it wasn't.

The Great "Agent" Confusion

First, let's clear up something important. Not everything that uses AI is an "agent." I've seen many solutions labeled as "AI agents" that are really just well-structured workflows or process automation. And you know what? That's perfectly fine! Sometimes, that's exactly what we need. And if we look at the procurement world, it's primarily AI automated workflows.

Let me break this down based on what I've learned from both implementing AI in procurement and studying the latest developments:

1. What We Call "Agents" Today

   • Simple automation tools with AI capabilities

   • Workflow systems with predefined paths

   • Basic chatbots with some decision-making abilities

   • Actual autonomous agents (rare!)

2. What Real Agents Should Be

   • Systems that can plan independently

   • Solutions that can adapt to new situations

   • Tools that can learn from their interactions

   • Applications that can truly make decisions

What We Actually Need in Procurement

Let's be honest - how many of our procurement processes actually need a fully autonomous AI agent? In my experience working with procurement teams, what we often need is much simpler:

1. Document Automation

   • Contract creation and analysis

   • RFP response evaluation

   • RFP creation

2. Data Analysis

   • Offers analysis

   • Spend analytics

   • Market intelligence gathering

3. Process Automation

   • Purchase order creation & processing

   • Invoice processing

   • Report generation

And here's the interesting part - most of these tasks work better with straightforward workflows rather than complex agents. It's like using a sledgehammer to crack a nut!

The Right Tool for the Right Job - my assessment

Based on my experience implementing AI in procurement, here's what really works:

Simple Workflows (80% of Cases)

• Predefined processes with clear steps

• Regular, repetitive tasks

• Standard document processing

• Basic data analysis

Advanced Workflows (15% of Cases)

• Multi-step approval processes

• Complex document analysis

• Supplier evaluation systems

• Market intelligence gathering

True Agents (5% of Cases)

• Complex negotiation support with a large number of variables

• Strategic sourcing decisions with complex decision matrixes

• Risk management scenarios

What Should We Do?

1. Start Simple Don't jump straight to complex agent solutions. Begin with basic workflow automation and build up from there. I've seen too many projects fail because they tried to implement sophisticated agents when simple automation would have worked better.

2. Focus on Value Ask yourself: What problems are we really trying to solve? Often, the answer isn't "we need an AI agent" but rather "we need to automate this specific process."

3. Build Gradually Start with one process, perfect it, then move to the next. This approach has worked well in my projects - it allows teams to learn and adapt while delivering immediate value.

Looking Ahead

The future of procurement isn't about having AI agents everywhere - it's about having the right mix of tools, workflows, and, yes, sometimes agents, to make our work more effective. It's about augmenting human capabilities rather than replacing them.

Think about it this way: Would you use a Ferrari to do your grocery shopping? Probably not. Similarly, not every procurement task needs a sophisticated AI agent.

Remember - sometimes the simplest solution is the best solution! 😊

Looking forward to your comments and experiences.