The fourth episode in a series of Silicon & Steel Founder’s Talk conversations with the founders building what’s next. Today’s guest is Arkestro’s Edmund Zagorin.

Say your company buys 40,000 parts a year. Every one of those has a price someone had to agree on. Today that mostly happens over email. The supplier sends a quote, your team pushes back, and three weeks later you land on a number nobody is sure was a good one.

My guest today built a company that does it the other way round. Arkestro is built on a simple idea: if your team already knows roughly what the price should be, stop waiting and make the offer yourself. The software predicts the number, and the buyer goes first.

Edmund Zagorin is the founder and Chief Strategy Officer. He started the company as Bid Ops in 2017, and it is now backed by New Enterprise Associates and Aramco Ventures. Before this he studied philosophy at the University of Michigan and was one of the best college debaters in the country.

The full conversation is on YouTube, linked below. What follows is the short version.

Let’s get into it.

I asked Zagorin to explain Arkestro without jargon, not to a VC but to his mother. He said he had already had that conversation and gave me the same answer he gave her.

Underneath that is a spectrum. On one end, rare and complex deals meant to last years. On the other, simpler purchases repeated constantly. What that spectrum really measures, he argues, is not complexity or frequency by themselves but how predictable the outcome intrinsically is. A good category manager can already tell you how most deals will land before they begin, which is exactly why they can tell you whether you got a good one. Arkestro’s stated aim is to replicate parts of that instinct.

i) Flip who makes the first offer

In a normal negotiation, the supplier quotes and the buyer responds. Zagorin’s team asked a blunt question: if a skilled procurement person already knows what they want, why are they waiting to be told a price? Arkestro lets the buyer make the offer first, with the reasoning attached.

A lot of the cycle time, he says, gets consumed by the supplier trying to work out what the customer wants before pricing it.

ii) Zero-sum thinking creates a two-sided incentive to stall

This was the part that stayed with me longest. Most of us were taught negotiation is win-lose. Zagorin’s point is that this belief has a cost you can measure in weeks. Procurement withholds the award to extract concessions. The salesperson withholds agreement for the same reason.

Every negotiation has a zone of possible agreement, the range of outcomes you could reasonably expect. His argument is that when procurement is the one making the offer, the small incentives inside the process change, and both sides can move quickly to a final number instead of circling it.

iii) The predictive model is not the new part

I asked what happens when the supplier runs a model too, and both sides show up with software instead of a person. Does that make it faster and fairer, or just an arms race decided by whoever has the better model? His first answer was that even when the two models disagree, referencing an outside benchmark beats the alternative, which is collecting quote after quote by hand. He compares it to Kelley Blue Book: price discovery is labor-intensive, and a shared reference point removes a lot of the stop-start.

He also pointed out that models are already everywhere, from search to airline pricing to the should-cost models procurement teams build by hand in Excel. Cost modeling, he notes, originates to some degree with Josiah Wedgwood’s pottery business in the eighteenth century. The question is not whether predictive models arrive in procurement. It is what form they take inside negotiations.

iv) Someone still has to sign

He does not think humans get subtracted from this.

Partly that is enterprise risk aversion, someone on payroll has to be the signer. Partly it is something more interesting: as supply chains speed up, markets get more volatile, and more information becomes AI-generated, person-to-person trust gets more valuable. Having your supplier’s cell phone number on speed dial when an emergency hits is the example he used.

v) Automation is the small version of the idea

Zagorin draws a line between what he calls the automation enterprise and the predictive enterprise, the latter also being the title of his book out this October. The automation view asks which units of work a machine can replace, and measures success as work replacement. He thinks that framing misses the prize, and he uses history to argue it. He points to Stanford economist Paul David’s essay on the Solow paradox, named for Solow’s line that you could see computers showing up everywhere except in the productivity statistics. David’s argument is that the same lag had already happened with electricity. Factories had been built vertically around one large steam turbine. Small electric motors let production run in parallel instead of in sequence, but it took roughly twenty years for anyone to redesign around that.

The redesign he is pointing at: what happens to a supply chain when demand can be communicated much earlier, and committed to with enough confidence to act on.

There is a second set of experience sitting behind this company. Arkestro’s CEO is Rob DeSantis, who co-founded Ariba. Chief customer officer Greg Cruz was the first Ariba implementation consultant and brought its first customers live. SVP of marketing Stuart Gold created Ariba Live. Zagorin called working with them a humbling experience, and he talks about that generation of procurement software with real affection rather than as something to be replaced.

The lesson he takes from it is mostly not technical. Ariba invested heavily in category creation, building the idea of spend management and then developing the measurement frameworks together with customers rather than for them. Those frameworks spread and became best practice. Arkestro tries to work the same way, staying close to how customers define excellence, and when someone brings a new metric the team treats it as a signal worth understanding.

The metric Zagorin wants to move is different from the old one: beyond purchase order compliance, toward procurement as a strategic contributor to margin. He also mentioned rethinking the supplier directory itself, since most are still organized alphabetically by brand name, which he notes is not that useful in a crisis.

The bet, in one line: value in procurement moves from processing the transaction to predicting it, and changing who makes the first offer is the lever that speeds the whole thing up.

The honest counterweight came when I asked for an unpopular opinion, and he aimed it at his own industry.

He is careful about what he is and is not saying. There may well be transactions with agents on both sides. What he struggles to picture is someone building an oil rig or standing up a new car factory and letting agents handle the key decisions. A hybrid, where agents support tasks inside those processes, is much easier for him to imagine than wholesale replacement of large human teams. He added that he could be wrong, and that you never bet against technology.

His comparison is Waymo. A rider still hails the car. Remote staff still supervise it. The humans did not actually leave the system.

I end these conversations the same way. His advice to anyone building here was narrower than I expected, and better for it: find a pain point specific to one industry vertical first. Industries name things differently, measure differently, and hold their own norms about what is acceptable in a negotiation. The solutions he sees gain traction fastest start with a small group who all share the same problem and describe it the same way. One of them finds a fix and tells the others, and that is where fast adoption comes from. Get it into as many hands as quickly as possible, he says, and work out the economics as a second step.

He is also unusually direct about the moment.

Supply chain was never the profession anyone pointed a spotlight at. He thinks that is changing fast, that AI is a big part of why, and that there are more than enough problems left for everyone who wants to come and help solve them. Then he closed with this:

Which is as good a reason as any to keep watching this space, and this series.

Subscribe if Silicon & Steel is your kind of thing.
More conversations are on the way, and my thanks to Ben and Drew for being on this talk show.

Gaurav Singh Chaudhary
Silicon & Steel
Silicon & Steel, intelligence for chips, batteries & trade

Advertisement

Ads Ends