FDE: Who Makes AI Work for the Customer?

Pavel Golubev, Principal AI FDE Manager at Microsoft, describes a shift from customer engineering funded through Azure consumption to paid engagements with business commitments. Managers enter discussions earlier, turn broad ambitions into use cases, estimate resources, and connect sales with staffing. Pavel emphasizes that the new operating model is still developing.

The central contradiction is promising an outcome before investigating the data. Pavel proposes a separately funded feasibility stage; Alexander observes that prohibiting a small initial investment can increase the risk of a large project. They discuss two inflexible engagement packages, possible revenue sharing, and contractual checkpoints that allow a project to change or stop.

In an oil company’s document workflow, complex permissions made custom agents too expensive for the smaller engagement. Copilot Studio offered a way to use existing user permissions, but internal IT opposed business self-service. The compromise kept solution creation with IT. The example shows why authority, departmental interests, and willingness to change processes must be considered alongside technical architecture.

Pavel explains engineering participation in team-based sales and the FDE matrix: engineers, data scientists, and technical program managers divide work collaboratively. Customer engagements include sharing engineering practices. The closing career discussion stresses rapid learning, negotiation, and comfort with uncertainty. An appropriate solution may use ordinary rules even when the initial request was for an AI agent.