The Biggest Mistake You're Making With Your AI Experiments

n our latest Proc n Roll jam session, Conrad sits down with Ajay to explore the massive shift from basic generative AI to true agentic procurement. If you are still forcing your business users to hack their way through complicated self-service catalogs, this episode is a wake-up call.

Here is a sneak peek at what we unpack in this episode:

The Return of the "Travel Agent" UX

‍Why the future of procurement looks less like a modern self-service portal and more like the 1990s white-glove corporate travel agent—but powered entirely by AI.

Mandatory Data Guardrails

‍The hidden dangers of giving agents unrestricted access and why protecting your confidential pricing and contract data is absolutely critical.

Boosting Your AIQ (Agentic IQ)

‍Practical, zero-fear steps to start building your own AI agents, including how to set up a "gem" to automate personal tasks like managing receipts.

Relationships Over Results

‍A candid discussion on leadership, and why the most effective careers eventually shift focus from demanding pure results to fostering intentional, relationship-based team building.

Transcript: Proc-N-Roll | The Biggest Mistake You're Making With Your AI Experiments

Conrad: Hello and welcome back to Proc and Roll. Today we have a special guest, Ajay Khosla, who is in the middle of procurement transformation and automation at Google.  

Ajay: Thanks, Conrad, I’ve been looking forward to jamming with you on this session.  

Conrad: A year ago, everyone was talking about generative workflows and using tools like Zapier, but tools like Open Claw have completely changed how I think about AI and enterprise applications. Did I miss something?   

Ajay: It has been a rapid evolution. Hallucinations are down significantly, making these models highly capable of augmenting human work, though we still need humans in the loop.  

Conrad: The more access you give an agent, the more powerful it is, but it also becomes dangerous if you don't stand it up in a controlled way. The shift is incredible; I am currently using an agent to build a highly complex Excel spreadsheet better than I ever could. I believe the future user experience for enterprise applications will just be a simple chat interface.  

Ajay: Procurement organizations are strapped for resources. Agents can lessen that burden by taking on tasks like market analysis and kicking off RFPs. However, you absolutely need strict guardrails, or highly confidential pricing and contract terms could leak out into the wild.  

Conrad: Exactly, we've already seen salespeople unintentionally train public models on their proposals because they didn't protect their pricing data. With your broad scope at Google, what gets you the most excited about transforming our domain?   

Ajay: We are on a mission to standardize procurement across a 25-year-old company. I am most excited about moving away from standard, sequential forms to an interactive experience where an agent asks the relevant questions, scopes out what you need, and identifies the right supplier through a simple dialogue.  

Conrad: In the 90s, we had corporate travel agents who provided a white-glove experience, but that was replaced by complex self-service portals. What you're describing sounds like giving every business user a "CPO-level brain" sitting right next to them to handle that complexity for them.  

Ajay: Yes, end users shouldn't have to care about sequential steps or specific portals as long as they follow policy. The agent can guide them, enforce buying thresholds, and escalate to a human only when necessary.  

Conrad: Right now, highly skilled people are bogged down spending up to 30% of their time on administrative tasks like supplier onboarding. Will these agents take our jobs?   

Ajay: In the future, 80% of low-dollar, low-complexity transactions will not need a human in the loop. This allows procurement teams to pivot to more complex and enjoyable advisory roles, like advising the business on data center sites. I don't think people will lose their jobs; instead, product launch cycles will shorten, and we will get much more done in a year. People just need to be flexible and learn new AI skills quickly.  

Conrad: How do we move beyond proof-of-concepts to delivering actual value? What are your top tips for procurement leaders to raise their "AIQ" (Agentic IQ)?   

Ajay: First, you must get your data right. If you have data gaps, the AI will try to fill them for you, causing hallucinations. AI is now getting very good at scraping metadata from complex contracts and varied invoices to build actionable insights. To boost your AIQ, get a $20 personal subscription, keep your company data out of it, and build a mini-agent ("gem") to automate your personal tasks to overcome the fear. Also, find a digital-native mentor in their 20s and start learning together in a team environment.  

Conrad: I love that. I recently used an agent to recover over 100,000 photos from my wife's frozen hard drive in 15 minutes, which I never could have done before. You won't be replaced by an agent; you will be replaced by a human who knows how to use an agent. With the ground moving beneath our feet, how should procurement leaders approach their tech stack decisions right now?   

Ajay: Do not wait. Move away from relying entirely on legacy enterprise tools and build a nimble, modular stack focused strictly on the outcomes you want to drive. Traditional CLMs will likely evolve into basic repositories, and you will layer AI capabilities on top of them to extract the value.  

Conrad: As we wrap up, what is the most impactful leadership principle you would share with your younger self?   

Ajay: Always do the right thing. And remember that while delivering project results is important, the true value of your career lies in the "brothers and sisters in arms" relationships you build with your team.  

Conrad: I completely agree, intentionally investing in those relationships yields the best results. This has been an absolute pleasure. Thanks for being with us, Ajay, and to everyone listening, please subscribe, like, comment, and share!   

Ajay: Thank you so much, I had a blast.  

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