Vlad Kyle explains why AI will crush legacy ERP software
By Bram de Vries · Reporting from Amsterdam ·
On the a16z podcast, Vlad Kyle argued that AI agents will replace legacy databases, proving that stripping friction from procurement is the only way for modern businesses to survive.
A single missed email about a delayed part can cost an aerospace manufacturer hundreds of millions of dollars. On the a16z podcast, LEO co-founder Vlad Kyle, partner Sema Amble, and host Elena Burgerer explained this. They explained why enterprise purchasing is ripe for an AI revolution. Kyle argued that procurement is the unsexy engine of the global economy. He noted that a saving in procurement costs has the same P&L impact as an increase in sales. Yet, legacy software vendors have trapped this vital function in rigid databases.
The Expensive Illusion of the ERP
Established software giants like SAP and Oracle protect their turf by acting as systems of record. Amble argued that these legacy incumbents are limited by the very records they keep. They show the final price of aluminum, but they do not show the work. Kyle pointed out that procurement actually happens outside the database. It is buried in 500 emails, dozens of spreadsheets, and complex engineering analyses.
Incumbents are now trying to defend their monopolies by slapping simple chatbots on top of their old software. This is a cheap marketing trick. A retrieval bot cannot manage the real-world complexity of a supply chain. True efficiency requires autonomous agents that can execute end-to-end jobs. LEO’s multi-agent system does exactly this by coordinating tasks across legal, finance, and operations. It moves beyond simple data retrieval to apply actual business judgment.
The Folly of the Corporate DIY
The ablest defenders of the status quo argue that large enterprises should build these AI tools themselves. A Fortune 500 company has deep pockets and proprietary data. It seems logical to bypass startups, plug a foundation model into existing databases, and avoid hefty licensing fees. This approach promises total control over sensitive supply chain relationships.
This is a classic corporate delusion. Amble recalled a giant firm that tried to build its own cash collection tool. The project collapsed because the firm lacked clean data and integration expertise. History shows that internal corporate IT projects are where capital goes to die. Companies must focus on their core competencies. Building procurement software is not a core competency for a carmaker or a drone builder.
Trading Friction for Margin
The real value of AI lies in stripping friction from the physical world. Kyle described how LEO uses autonomous agents to run negotiations. Historically, enterprises ignored these smaller purchases because they lacked the staff to negotiate them. Startups exploited this by sending unnegotiated invoices. AI agents close this leak.
Some fear a future where buyer agents endlessly battle supplier agents in a zero-sum game. This fear is misplaced. While final prices are zero-sum, the tasks leading up to a contract are not. Both sides want faster transactions and shorter times to market. AI-to-AI coordination will speed up the entire supply chain.
Brussels bureaucrats write endless directives to regulate AI. American builders are using it to rebuild the plumbing of global commerce. European firms cannot afford to fall behind. If you do not automate your purchasing, your competitors will use AI to underprice you. Survival in global trade has always belonged to the efficient.