Better, Faster, Stronger: The Power of Agentic Workflows

Can you build a fully functional AI procurement agent without writing a single line of code? In this special “Show and Tell” episode of Proc & Roll, host Zachary Bachir proves it’s possible. He rolls up his sleeves to demonstrate a custom AI intake agent he built from scratch using N8n, a low-code automation platform, in just about 50 hours of work.

The episode moves beyond the theoretical hype of AI to show exactly how agentic workflows operate under the hood. Zach demonstrates a live purchase requisition where an agent parses a stakeholder’s email, checks it against procurement policy using RAG (Retrieval Augmented Generation), and drafts a perfect, human-like response—all without human intervention.

Topics Discussed in This Episode:

• The AI “Show and Tell”: Zach gives a live demo of his orchestrator agent. He sends a simple email (“I want to buy software for 200k”), and the agent automatically parses the intent, identifies it as a purchase requisition, and consults sub-agents (“Commercial Analyst” and “Compliance Analyst”) to validate the request against policy .
• Demystifying RAG: The episode provides a practical explanation of Retrieval Augmented Generation (RAG). Zach shows how he built a pipeline that takes unstructured data—like PDF contracts and policy documents—and “chunks” them into a vector database, allowing the AI to actually “read” and cite internal documents .
• The Governance Debate: The demo sparks a heated discussion about the risks of “Shadow AI.” While Natasha is “superbly impressed” by the technology, she raises concerns about the potential chaos of every procurement professional building their own un-governed workflows .
• The 50-Hour Investment: Zach reveals that building the entire system took only about 50 hours of learning and work. The hosts discuss how this relatively small investment can pay huge dividends, not just in automation, but in technical literacy.
• A New Negotiation Superpower: Natasha points out a key takeaway: if procurement professionals learn how to build these tools themselves, they become far better negotiators. Understanding the true effort required to build AI agents means you no longer have to “blindly trust” software vendors when they claim a feature is too labor-intensive or expensive to build .

Watch now or read the transcript below.

Transcript: Proc-N-Roll | Better, Faster, Stronger: The Power of Agentic Workflows

Conrad: Hello everybody, welcome to Proc and Roll. Today we have a special show and tell with Zach. It seems like AI continues to be a challenge and an opportunity with a lot of hype, which is why we wanted to get practical . Zach, what is the goal of this AI tool, and how did you decide to jump into it?.

Zach: I built an intake and orchestration agent using a tool called N8n, which is a low-code platform . I wanted to pierce the opaque bubble surrounding agents. Before showing the agent, I need to explain the RAG pipeline. RAG stands for Retrieval Augmented Generation. It simply means taking your unstructured data—like policy documents and contracts—and “chunking” them into a database so the AI can read them .

Conrad: So it’s translating documents into little snippets that are in the language of an AI?.

Zach: Exactly. It turns text into strings of numbers (vectors) and compares how similar they are to a user’s query . Let me show you the demo. I’m sending an email to my agent as a stakeholder saying, “Hi, I want to buy some software for 200k. I have budget approval.” .

Zach: The workflow executes: the “Orchestrator Agent” parses the intent, recognizes it as a new purchase requisition, and then consults two sub-agents: a Commercial Analyst (checking preferred suppliers) and a Compliance Analyst (checking policy) . It then automatically replies to the stakeholder. The email says: “Threshold for sourcing is usually 100,000… To move forward quickly, please provide currency, cost center, and business justification.” It did all that without me touching it .

Natasha: I am superbly impressed by your ability to build it, but I am very skeptical . Risk is the first thing I think about. Do we think that every procurement professional will be expected to build such flows? If they do, aren’t we creating chaos where everyone builds their own parallel flows outside of policy? .

Conrad: That’s an interesting point. Who is going to create these? To me, an agent that delivers capability across the organization should be built by a Center of Excellence . However, as a worker, I’m asked to do “trivia stuff” constantly, like answering “what’s the status of my PO?” If I can build a simple agent to automate that 20% of my time, that’s valuable .

Zach: I agree. Not everyone needs to do this. Procurement functions should invest in teams that build these processes so they can automate strategic benefits . But big platforms like Zip and Aoro will eventually offer these tools out of the box .

Conrad: How long did this take you?.

Zach: About 50 hours of work.

Natasha: That is very affordable. And here is my favorite soapbox: If buyers understand what it takes to build this technology, it makes us more knowledgeable negotiators. We no longer have to blindly trust a supplier’s message about how “labor intensive” a feature is to build because we know what it actually takes .

Conrad: My advice to leaders is to roll up your sleeves. Every month you don’t try this is another month you fall behind. You don’t need to start by putting your entire enterprise at risk, but you need to get comfortable with the tools.

This transcript has been edited for clarity while maintaining all substantive content