Archive | Upgrade to Premium | Sponsor

Hi {{FIRST_NAME|readers}},

"Can you grab the PDFs from my email about that Boston office and add them to the request?"

I was recently on a webinar watching a proof-of-concept demo. A category manager was talking on the phone to an AI agent during his commute about renewing a multi-million dollar facilities contract.

The agent pulls documents from his inbox, tags stakeholders, and builds out a complete sourcing request… All before he gets to the office.

It was pretty impressive.

This isn't live yet. And I know it'll take serious work to make it reality.

But I also saw the building blocks that will make this possible. And that's exciting!

90% of what you read about AI agents these days is either vaporware or ChatGPT prompts disguised as strategy.

This webinar was different. Real client examples. Production systems. Actual time savings.

But I asked the uncomfortable questions anyways:

The answers? More practical and less hyped than most vendors admit.

Essentially, the benefits don't happen magically. There's hard work ahead if you want to be the equivalent of an F1 team. But the tools are there. And they're performing at the level needed to get incredible results.

This week: what I learned from a 60 minute webinar on AI agents in sourcing.

No fluff. Just what's working, what's not, and where to start.

Let's dig in 👇

Onwards!

📰 In this week’s edition:

Note: Some of the content listed above is only available in the email version of this newsletter. Don’t miss out! Sign up for free to get the next edition.

👀 In Case You Missed It…
My Best Linkedin post this week:

Most procurement teams are thinking about AI agents completely the wrong way. Here's what actually works:

Because most of what you're reading about AI agents is either vaporware or someone's ChatGPT prompt disguised as strategy…

I just wrapped a webinar with the Fairmarkit team (Kevin Frechette, Shane Hetrick, and Tiago Melo) and we spent an hour cutting through the noise. Not theorizing about AGI. Not pitching the "future of work." Just showing what's actually working in production environments today.

Here's the key takeaway: Most companies are paralyzed by perfection. Waiting for clean data. Waiting for budget. Waiting for the "right time."

Meanwhile, a handful of companies are quietly pulling ahead.

Let's get something straight first. It’s my biggest pet peeve around AI…

AI isn't new. It's been around since the 1956 Darthmouth Conference. Deep Blue beat Kasparov at chess in 1997. IBM Watson crushed Jeopardy in 2011.

What changed recently? Three things:

That's it. No magic. No singularity. Just better tools becoming more accessible.

Kevin dropped a truth bomb on the call: "Every year someone says we've hit the tipping point. Then next year, we say it again."

He's right.

LLMs passed the Turing test recently and barely anyone noticed. The world didn't end. The robots didn't take over. Work just got... incrementally better yet again while passing a milestone.

The tech press loves declaring inflection points. But when you're actually in it, working with these tools every day, it doesn't feel like a revolution. It feels like steady improvement.

So here's a better question: What specific problem can AI solve for you this quarter?

Not "how do we prepare for AGI?" Not "what's our 5-year AI strategy?"

Just: what's broken today that these tools can help fix?

These examples aren't theory. This is what's working in production with real customers:

An end user types: "I need marketing content creation services."

That's it. That's all the detail they give.

The agent kicks in. It classifies the request, assigns the right buyer, determines the budget band, and starts asking clarifying questions, intelligently and based on YOUR DATA AND RULES. Not with a rigid form. With a conversation.

Then it taps into a second agent (the Statement of Work agent) that helps the requester build out a complete, structured scope document through an interview-style process.

What used to take 20+ hours of back-and-forth between stakeholders and procurement now happens in minutes.

Complex RFP with qualitative responses, references, certificates, capability statements buried in PDFs?

The agent reads everything. Creates bespoke summaries. Analyzes risk factors. Lets you chat with it to explore different angles: "How should I look at this from a risk standpoint?"

It's not replacing the human decision. It's giving the human superpowers…

Because let’s face it, half the selection committee isn’t reading all the documents anyways 😂

Ok… I lied… We did dip into the “future of work” a tiny bit…

(but only after building solid foundations on the actual current state)

Couldn’t help ourselves 😅

Kevin demoed a concept where a facilities category manager gets a Teams message:

An AI agent calls him 15 minutes later, during his commute.

Done. The agent pulls documents, tags stakeholders, sends a Teams reminder to upload one more file later.

The Category Manager logs in later and sees a fully contextualized request with market insights, category strategies, and recommended approaches (with pros and cons for each option)…

All before finishing his morning coffee at the office.

Based on what I’m seeing in the market, this isn’t 10 years away… For organizations who get their act together and start experimenting with the right partners, this is around the corner!

Someone asked on the call: "Don't I need perfectly organized contracts, vendor data, and sourcing documents before AI can help?"

Kevin's answer was blunt: "No."

Better data = better outputs. Obviously.

I agree.

But here's what matters more: Build processes with AI that help you clean your data incrementally while solving real problems in “business as usual”.

Fairmarkit uses natural language processing to read messy data. Anomaly detection to spot issues. Machine learning to improve over time.

Perfect data isn't the starting line. It's what you build toward while getting value along the way.

Forget the hype about Artificial General Intelligence (AGI) and "tipping points."

Break your sourcing process into chunks:

Pick ONE. Find the biggest pain point. Apply AI there.

Then move to the next one.

The highest value use case will depend on your company’s spend profile. If most of your spend is in professional services, the answer will be different than in a manufacturing business.

Snowflake and Goodyear aren't running autonomous sourcing processes because they waited for perfect conditions. They started small, learned fast, and scaled systematically.

Once you’ve identified your highest value use cases, you’ll need to determine what kind of AI usage makes the most sense in context…

Not all tasks should be automated the same way.

Think about two dimensions:

Low-risk, deterministic tasks? Let AI run. This is soul-crushing work nobody wants anyway.

High-risk, creative tasks? Keep humans in the loop. But give them an AI co-pilot trained on your best negotiators, your historical data, your market insights.

One of Fairmarkit’s customers said their biggest use case for agentic AI was their $100M+ strategic contracts (not because AI does the negotiation, but because it coaches the human through every step with institutional knowledge that would otherwise be scattered across people's heads).

That's lights-out value.

Fairmarkit showed a graph that clarifies where most companies actually are:

Here's what matters: Some companies are already at "AI brings work to me" maturity. Most are still at "AI as part of the process" or "asking AI to perform a task."

Kevin's observation: "This gap is only going to keep getting bigger."

But it's not about panic. It's about being honest about where you are and deciding where you want to be.

Boeing is focused on making processes faster with embedded AI. Coca-Cola is experimenting with autonomous agents. Snowflake and Goodyear are pushing hard on fully autonomous sourcing.

Other companies that are less “sourcing-heavy” are doing this same exercise but with other parts of the procurement process and with other technologies/partners.

Different maturity levels. Different approaches.

But they all have one thing in common: all making progress by investing in *practical* digital literacy.

Three takeaways:

Procurement is entering an era where one person can manage processes that used to require entire teams.

The question isn't whether AI will change sourcing. It already is, in production, with real customers.

The question is: What are you going to do about it?

Not in a panic. Not because of FOMO. But because you see an opportunity to make your team more effective.

Here's what separates the companies making real progress from everyone else:

They invested in platforms and co-development, not rigid point solutions.

Not another tool that requires IT tickets every time you want to test something. Not another system where every experiment needs a business case and a CapEx meeting. Not another "enterprise solution" where you're stuck waiting on someone else's roadmap.

A platform gives you a toolbox to iterate independently within a confined scope. To run experiments. To learn fast. To scale what works and kill what doesn't.

Then, you implement the governance model to support constant iteration.

Because here's the real barrier: It's not the technology. It's organizational velocity.

The gap between companies isn't about who has better AI. It's about who can move, learn and iterate faster.

Snowflake isn't running autonomous sourcing because they have better tech. They have a platform that lets them experiment without asking permission to the rest of the organization....

So when you're building your business case, don't ask for "AI tools."

Ask for a platform and a partner that provides fit-for-purpose AI tools that gives procurement the freedom to move at the speed of the business.

That's the real differentiator.

Want to see these AI agents in action? The Fairmarkit team is shipping production-ready systems with real customers. Watch the webinar replay if you want to see the above in action: