Coupa replaced its chief executive in late August. Leagh Turner stepped down. Michael Lipps, a Thoma Bravo operating partner who has sat on Coupa’s board since 2023, took over the following Monday.
Two weeks on, the personnel story is finished. The argument it set off is not, and most of it has very little to do with Coupa. Analysts, chief procurement officers, board directors and vendor founders filed several hundred public comments in a week. Read together, they amount to the most candid public audit of procurement software the market has produced this year.
The background is short. Thoma Bravo took Coupa private in 2023. Rob Bernshteyn, who ran the company for 14 years, left the same year. Turner arrived from Ceridian in November 2023 and led through a take-private, four acquisitions and an AI cycle that aged every legacy platform. Lipps is the fourth name at the top in roughly 40 months. Two of those appointments came from the sponsor.
Orchestration finally got cross-examined
Intake and orchestration has been the loudest category story in procurement for two years. The threads split on whether it is a product category or a symptom.
Jason Busch of Spend Matters argued the wave came from something real. Suites promised a clean process and delivered something else when the work broke down under actual conditions. Buying starts messy and cross-functional and almost never inside a purchase-to-pay workflow. Challengers built their entire go-to-market on that gap.
Koray Köse of KŌSE Advisory called orchestration a symptom of unmodernized procurement organizations and restless capital, not a destination. He brought numbers. MIT’s NANDA initiative found that 95% of enterprise generative AI pilots deliver no measurable profit-and-loss impact, against $30bn to $40bn chasing the promise. Procurement technology specifically reports implementation failure rates of 80% to 95%.
Practitioners were harsher than either analyst. Tony Marino, a CPO and board director across private equity portfolio companies, wrote “Orchestrator tools are kinda B.S. for me” and described them as a way for CPOs to look like they are enabling the business.
Rob Turner, a chief procurement officer, gave the most useful diagnosis in either thread. He sees implementations fail because orchestration layers sit on infrastructure that cannot carry them. His test question for any vendor is “where is the data foundation maintained”. He argued that procurement teams now need people with genuine source-to-pay architecture expertise on staff. Andrew Daley of Edbury Daley agreed that most CPO teams lack it and called operating without it a false economy.
That is the practical takeaway from the entire debate. If nobody on your team can answer the data foundation question, you are buying an interface and calling it a strategy.
Data is not an advantage until it can act
The strongest consensus had nothing to do with leadership.
Matt Zimmerman of Prime Digital Solutions made the point across both threads. Intelligence sitting behind dashboards is not a moat. The contest is “who can turn intelligence into governed execution at scale”. Knowing what should happen, knowing what you are authorized to do, and actually getting it done are three separate problems. Most vendors have solved the first one.
Georg Wall supplied the missing half. Agents need structured, real-time access to what can actually be bought, including prices, availability, alternatives and commercial rules, plus the authority to transact. That is a different problem from catalogs and punchouts. Busch endorsed it as a sixth priority to his own list.
Derek White argued the data only pays off if it becomes a procurement-specific language model with a knowledge graph for governance. He also warned that switching-cost protection may already be running out for some accounts.
Marino delivered the customer verdict in one line. Getting value out of the data takes too much effort, and it should be instant. That is what a paying enterprise customer thinks about a decade of accumulated spend intelligence. Every vendor claiming a data moat should read it twice.
Somebody pays when a pilot fails
Köse raised a point nobody else did, and it applies across the sector.
When an orchestration or agentic pilot fails at the rates already documented, the customer rarely blames the vendor in public. The team quietly returns to the incumbent platform to patch the data, reconcile the mess and restore the workflow the failed tool broke. Internally, that gets narrated as loyalty. Köse calls it a hidden tax. The incumbent’s support organization absorbs the operational cost of somebody else’s failure, usually for free, and never gets paid for the risk it just underwrote.
Buyers should invert that. If your team has quietly rebuilt what a failed pilot broke, the cost belongs in your business case and your vendor scorecard, not buried in support tickets. Failed procurement pilots are almost never accounted for honestly, which is exactly why the failure rate stays high.
Busch’s version of the same exposure runs the other way. He believes customers are already developing alternatives their incumbent has not detected, and put the window at roughly 18 months before quiet renewal conversations turn into real suite losses.
Leadership churn belongs in your risk register
Marcel Van Wonderen framed why any of this matters beyond one vendor. Many companies have made themselves materially dependent on a single platform. The direction that vendor picks sets the direction of their procurement function. That is concentration exposure, and it deserves the same treatment as any Tier 1 supplier.
Jack Miles of MainSpring Advisors asked the governance question directly. A board member stepping into the seat raises whether the move was sudden or long planned. He put the correction at six to 18 months and warned that a further leadership change would end the reset. Patrick Scully, a non-executive director, read the appointment plainly: “this is a signal that Coupa will be shopped soon”.
The dissent was blunt. Michael Lamoureux argued the platform is now a single interface rather than a coherent suite and said “Coupa’s glory days ended with Rob Bernshteyn’s departure”. Markus Hornburg’s judgment on the years since drew the widest agreement in its thread.
Others pushed back. Javier Carnevali and Michelle T. defended Turner’s record on customer focus and culture. Köse opened his own piece with a tribute to her stewardship through a take-private and a brutal AI cycle.
Two lines summarized the risk better than any analysis. Jan-Henner Theissen wrote, “Sometimes assets can become a burden”. Van Wonderen pressed the same nerve: “Sometimes doubling down is just refusing to turn around”. Both apply to any incumbent vendor sitting on a large installed base, and to any procurement team that built its process around one.
How to read vendor proof
Two exchanges are worth more to buyers than the strategy talk.
Tricia Miller, a former Accenture procurement leader now at Coupa, cited a 2026 Gartner Customers’ Choice distinction for source-to-pay with a 99% willingness-to-recommend rate. Busch challenged the methodology, arguing that crowdsourced peer models are a different instrument from analyst research, and said a board should watch net promoter score over time as the leading indicator of net revenue retention. Whatever you conclude, the exchange is a reminder that badges in a vendor deck do not all measure the same thing, and few buyers check which is which.
Köse also asked Nico Bac, founder of Digital Procurement Now, to declare his commercial relationship with a competing vendor before endorsing a position. Procurement technology commentary runs heavily on relationships that are never stated. Somebody finally said so in public.
Takeaways
Treat vendor leadership change as supply risk, not industry news. Run the same review you would run on any critical supplier under new ownership. Document your dependency, your switching cost and your realistic alternative before the renewal conversation, not during it.
Set a diligence standard for every agentic vendor you evaluate, incumbent included. Know which model runs underneath, who controls the weights, whose data pipeline it uses, and where the legal exposure sits the day it fails. Demand a validated, monetizable use case with measurable KPIs before signature. A demo is not evidence.
Audit your own data foundation before you buy another layer on top of it. Ask who maintains it, who owns governance, and whether anyone on your team can defend the architecture. That question decides more implementations than the vendor selection does.
What would you change in your vendor review process after reading this?
Continue the discussion on Chain.NET, the global supply chain community, at https://www.chain.net/c/ask. Ask questions, join events and access exclusive resources.
Membership benefits are listed at https://gscc.co/benefits, and the full event calendar is at https://www.chain.net/c/events/.
