At Adobe, the finance department isn't just balancing books—it's the primary engineering environment for deploying agentic AI. Dan Durn, leading finance, technology, and security, has architected a system where pilots bypass organizational friction and move directly to live operations. His team uses autonomous agents for forecasting, contract analysis, and communication, creating a measurable internal benchmark.

The technical execution focuses on structured data and governance to maintain accuracy while increasing deployment speed. Three applications demonstrate the system's design. First, an AI assistant ingests thousands of PDF documents—transcripts, reports—into Adobe's own PDF Spaces, extracting themes in minutes. Second, contract review agents parse key provisions across revenue and procurement documents, cutting analysis time by half and enabling semantic search across the entire repository. Third, email management agents automatically handle routing and responses for common inquiries, managing 300,000 messages last year.

Durn's method treats these projects as process redesigns, not just model deployments. The engineering trade-off is clear: minimal new spend on infrastructure, but significant investment in change management and integrating with existing Adobe platforms. This approach prevents the finance organization from becoming a bottleneck to the company's overall product velocity. The result is a production-grade pipeline that other departments can replicate, proving value through scaled automation rather than theoretical pilots.

Source: Webpronews