What Are Your Agents Standing On?

Whether they’re asking me or their teams, procurement leaders keep coming back to the same question: why buy a platform when you could point a few AI agents at your ERP and build your own?

As the CPO of one of those platforms, I certainly have a horse in this race. But that doesn’t mean the question should be handicapped in my favor; it deserves a straightforward answer.

If the product were AI and AI alone, that perspective would not be without merit. Generic AI is a commodity anyone with a credit card can leverage. But AI alone was never our moat. So, what is? There are a few ways to look at it:

1. Elegant, patented technology

For many organizations, procurement runs today the way it always has: send the RFQ, collect quotes, and react to whatever comes back. Arkestro flips that.

Before a supplier quotes, the platform predicts and then recommends a competitive price so procurement can make the first offer. The supplier can accept it, counter it, or beat it. When they counter, it counters back. And, if you’ve ever bought a car, you know: the first number frames the negotiation.

None of this is possible with general purpose AI. This combines behavioral science and game theory to understand supplier behavior, track what’s moving prices, and help procurement design better negotiations. The ideas fit in a paragraph. Applying them and meticulously engineering them to work on real spend took much longer. The result is patented technology, built on proprietary algorithms and data – both very hard things to replicate.

2. A proprietary dataset, built up over time

A prediction is only as good as the data behind it. Arkestro has been learning from live procurement activity since 2017.

Billions in spend, millions of line items, tens of thousands of sourcing events, and every quote, counterbid, behavioral signal, and award in between.

That history makes capabilities like Suggested Pricing (Arkestro’s “anchoring at scale” capability) possible. Knowing what a good price looks like before you go to market and letting other parts of our platform (like Arkestro Intelligence) continuously monitor, flagging opportunities proactively, instead of waiting for a project. This data flywheel is always spinning so every event, quote, action, or input makes it a little bit sharper.

Every new entrant or internal build starts with a blank slate, a “data zero.” You can choose to start with nothing and hope for the best, or you can leverage nearly a decade of data foundations to predict winning outcomes from day one. In procurement, starting from zero puts you at a real disadvantage. The better your data, the stronger your position.

3. Supercharging the value of the stack you already own

Arkestro works alongside your ERP, CLM, and P2P systems rather than replacing them. It sits atop these systems as an intelligence layer for sourcing and supplier decisions, answering questions those systems can’t: Am I getting the right price? Is this the right supplier? Would my team make the same call across all plants and categories?

The challenge is that most procurement knowledge doesn’t live in the ERP at all. It lives in emails, quote attachments, supplier conversations, spreadsheets, and in the heads of your best buyers. An agent can only work with what it can see.

Arkestro’s platform captures that context, structures it, and stores it in the organization’s “procurement memory.” Over time, that memory becomes a decision engine that both your team and your agents can harness.

4. Elevating the supplier experience

The supplier experience is where most internal builds falter. If the experience feels like another e-auction — or worse, a bot built to squeeze — suppliers won’t show up.

We spent years building the supplier side of the platform: a fair suggested offer, a fast process, and visibility on where they stand. This provides a real reason to participate. Suppliers who trust the process engage more, bringing optimal outcomes without harming the relationships your business depends on.

An agent can’t negotiate with someone who won’t answer.

5. Manufacturers make things. They don’t make software.

Most of our customers are manufacturers, energy companies, and industrial operators. Their best engineers are focused on products, plants, and core operations, which is exactly where they should be.

“Build it with a Copilot subscription” sounds easy. But, in practice you’re signing up to staff engineering, data science, and product, then test, secure, integrate, support, and maintain the thing forever. The demo is the easy 10%. Everything below the waterline (performance, quality, security, 2 a.m. support) is what makes enterprise software dependable, and that burden never goes away.

Realistically, we’re talking about a multi-year build before anyone trusts the output, with no guarantee buyers will adopt it or suppliers will engage with it. AI made code cheaper to write. It didn’t make software cheaper to own. That math hasn’t changed.

This is not an argument against agents. We’re entering the agentic era, and it’s changing how work gets done. The nature of work itself — particularly knowledge work like procurement and coding — is changing as we all welcome new agentic “coworkers” into our ranks to work alongside us on our teams.

But one thing remains certain: the best equipped teams will win. At Arkestro, we’re continuously improving a platform that helps agents and professionals predict and act on procurement and supply chain decisions. That’s the question I’d put to anyone evaluating AI in procurement: what are your agents standing on?