The Contract Mixtape: Sampling AI for Smarter Procurement

From Prompt to Powerhouse: How AI Agents Are Transforming Procurement Workflows

In the latest episode of Proc-N-Roll, Conrad Smith and Zachary Bachir go beyond theory and get practical with AI in procurement.

Zach shares a live demo of the AI-powered contract analysis tool he built, using Zapier, OpenAI’s GPT-4, and Google Workspace, all with just 5-10 hours of learning.

His workflow automatically:
✅ Detects new contract uploads
✅ Extracts and analyzes 15 key clauses
✅ Checks compliance against company policies
✅ Outputs structured data and reports

If there’s one message from this episode, it’s this: you don’t need to be a developer to automate complex procurement tasks.

With just a weekend of effort, you can build tools that save your organization hundreds of hours and dramatically increase your own impact.

Watch now or read the transcript below.

Transcript: Proc-N-Roll | The Contract Mixtape: Sampling AI for Smarter Procurement

Conrad: Welcome to Proc-N-Roll, your guide to practical procurement, where we make procurement rock and roll every day. Today, Zach and I are diving into AI, specifically how to get started building AI agents for practical procurement use. Zach, why don’t you tell us how you got started and walk us through the process?

Zach: Thanks, Conrad. A lot of people are talking about agents. It’s a big buzzword, and the space is moving fast. Companies like ServiceNow and Zip are rolling out their own agents. The evolution is interesting: we went from basic LLM prompts to orchestrated workflows, and now to autonomous agents that can make tool calls and operate more independently.

One key development is the Model Context Protocol (MCP) from Anthropic, which standardizes how agents connect with applications. Today, I’ll walk through building an orchestrated workflow with some agentic elements.

Conrad: When I first heard about agentic AI, I thought it was just another confusing buzzword. But it’s essentially about giving AI clear, specific instructions so it can perform specialized tasks, like extracting contract metadata. You can string these agents together to create more complex workflows.

Zach: Exactly. And today, I’ll show how I built a workflow using Zapier and OpenAI to automate contract analysis. This is accessible to anyone, you don’t need to be an AI expert. Let’s dive in.

Conrad: Let’s rock and roll.

Zach: I used Zapier to create an automated workflow. These platforms have exploded because you can drop AI into workflow tools to do specific tasks.

• Step 1: Trigger. I connected Zapier to Google Drive so that when I drop a new contract in a folder, it triggers the workflow.
• Step 2: Processing PDFs. Zapier can’t read PDFs directly, so I used a workaround: upload the PDF to a “processed” folder and convert it to text that the LLM can read.
• Step 3: JavaScript. I added a few lines of JavaScript to convert the PDF to readable text. No coding experience needed—just copy-pasted from a tutorial.
• Step 4: AI Analysis. I connected OpenAI’s GPT-4 through Zapier. My prompt told it to act as a contract analyst, extract 15 specific clauses, and return them as structured data.

Conrad: And just to clarify, someone could copy that prompt and use it directly in ChatGPT, right?

Zach: Yes, but using Zapier lets me set output fields to reuse the extracted data in other steps, which makes the workflow more powerful.

• Step 5: Google Sheets. The extracted clauses are mapped to columns in a Google Sheet. This creates a structured database of contract metadata.

Conrad: That’s huge, automating something that would take legal teams hours or days.

Zach: Exactly.

• Step 6: Compliance Analysis. I built another prompt for GPT-4 to evaluate each clause for compliance based on internal policies, like minimum termination periods or payment terms.

Conrad: So it’s not just extracting data. it’s analyzing it against your rules.

Zach: Yes. For example, I told it that termination for convenience must be at least 90 days. If a contract says 60 days, the AI flags it as non-compliant.

• Step 7: Formatting. I wanted a clean report, so I asked GPT-4 to output the analysis in Markdown, then converted that to HTML via Zapier.
• Step 8: Reporting. The formatted report is saved as a Google Doc in a designated folder. I also set up email notifications for completed analyses.

Conrad: That’s so cool. I’m thinking about all the times procurement or legal asked for contract summaries, this automates all that.

Zach: Exactly. And by breaking contracts into clauses, we reduce hallucinations and errors. Plus, we can scale this to process hundreds of contracts.

Conrad: And you could easily add another agent to fix non-compliant clauses, right?

Zach: Yes, I’m working on a “contract remediation agent” to propose compliant alternatives automatically.

Conrad: I’m blown away. This is the kind of automation that transforms procurement, doing in minutes what used to take weeks.

Zach: And the barrier to entry is low. Anyone can build something like this in 5-10 hours using online resources and AI assistance.

Conrad: The takeaway for everyone listening: get off the sidelines. If you can describe a task, you can probably automate it.

Zach: Absolutely. Even if you’re not technical, AI can guide you through building these workflows. The new generation will expect tools like this to be available.

Conrad: Thanks, Zach. This is a game-changer for procurement. We’d love to hear from listeners experimenting with AI workflows.

Zach: Definitely. If you’ve built something cool, reach out.

Conrad: Thanks, everyone, for listening. We’ll see you next time on Proc-N-Roll.

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