Before the Algorithm, There Was the Argument
In the spring of 2011, a University of Michigan debate team beat Harvard at the University of Georgia Tournament. Their debate director called it one of the most significant wins in a decade. One half of that team was Edmund Zagorin - a fifth-year senior whose research style a teammate once described, affectionately, as "wide-ranging and a little less organized." The organized half was Maria Liu. Together they won 20 debates - a feat fewer than 20 teams had managed since 1946.
That background is not a quirky footnote. It is the through-line. Zagorin spent his undergraduate years at Michigan studying philosophy, international affairs, and public policy - and spending an unusual number of hours refining arguments about things that actually mattered: policy tradeoffs, second-order consequences, what happens when you press a claim all the way to its logical end. That training, it turns out, is excellent preparation for what he does now.
Arkestro - the AI-powered Predictive Procurement Orchestration platform he co-founded in San Francisco in 2017 - is, at its core, an argument about the future made executable. The argument goes: most procurement decisions happen too slowly, with too little information, and are guided by instincts that can be modeled and beaten. The AI disagrees with your pivot table before you build it. It runs the simulation before you run the RFP.
"AI-driven solutions can transform procurement into a proactive function by automating or eliminating manual steps in cycles ranging from supplier selection to benchmarking - allowing teams to act on quotes and secure optimal pricing in days rather than weeks."
Edmund ZagorinThe founding story has the texture of a good anecdote because it actually happened that way. Zagorin was working as a procurement consultant - advising healthcare providers, contract manufacturers, and multi-campus retail brands on data-driven supplier negotiations. He reconnected with his childhood friend Ben Leiken, who had built his career in engineering and product at SurveyMonkey. They were in Potrero Hill. Zagorin was describing, not for the first time, how much time sourcing teams wasted building Excel pivot tables in software that felt like it was designed to resist the very task it was sold to do.
Leiken said: what if we fixed that? On April 1, 2017, they founded BidOps - which would later become Arkestro. The name change was deliberate: Bid Ops described a feature. Arkestro described an orchestra. The AI doesn't just automate a task; it coordinates the full sourcing cycle - supplier selection, negotiation, benchmarking, outcome prediction - the way a conductor pulls together instruments that have never met.
Arkestro's approach rests on three pillars Zagorin calls the Three Sciences: Negotiation Science (game theory applied to pricing strategy), Supplier Science (AI-driven supplier selection and engagement), and Process Science (workflow automation that compresses months into days). The platform pre-populates validated supplier data, simulates a procurement cycle before it begins, and sends AI-generated quote recommendations to suppliers - all before a category manager has opened a single spreadsheet.
The results are specific enough to be persuasive: an average of 18.8% cost savings on every million dollars of spend, and a 60-90% reduction in sourcing cycle time. These aren't ballpark estimates pulled from a marketing brief. They're outputs from a system designed by someone who spent years watching real sourcing teams struggle and built the tool they actually needed.