How Fortune 500s Use Procurement to Manage Vendor Lock-In in AI Models

Episode notes

In this episode of Enterprise Tech with Fexingo, Lucas and Luna dive into how Fortune 500 procurement teams are tackling a new kind of vendor lock-in: dependency on a single AI model provider. They unpack the case of a global bank that spent $8 million integrating GPT-4 into its customer service workflow, only to find it couldn't easily switch when a competitor offered better latency for half the cost. Lucas explains how procurement is now inserting 'model portability' clauses, requiring APIs that support multiple model providers, and even building internal benchmarking teams. Luna highlights the rise of model routers—software that dynamically chooses the cheapest or fastest model per query. The hosts discuss the tension between performance and flexibility, and why procurement's role in AI is becoming as strategic as it was for cloud infrastructure. A must-listen for anyone involved in enterprise AI purchasing or vendor risk management.