The AI in Business Podcast Scaling Regulated Data Workflows Without Lock‑In - with Juan Orlandini of Insight
15 snips
Apr 17, 2026 
Juan Orlandini, CTO of North America at Insight, is a finance-focused AI and data engineering leader. He warns that generative AI is statistical, not mathematically precise for compliance. He discusses how automation can increase verification work, parsing unstructured PDFs, using vendor controls first, building an AI-ready data layer, avoiding vendor lock-in, and investing in people.
AI Snips
Chapters
Transcript
Episode notes
Generative AI Is Statistical Not Mathematical
- Generative AI models produce statistical answers, not guaranteed mathematical results.
- Juan Orlandini warns finance must constrain these models with existing financial controls to avoid hallucinated, convincing but incorrect outputs.
Delta PDF Map Showed AI Power And Gaps
- Juan parsed a Delta PDF with a generative tool to draw travel maps and it worked beautifully on first pass.
- The tool missed a few flights, demonstrating visually impressive results can still contain errors requiring human verification.
Use Mature SaaS Vendors Before Custom Builds
- Start with mature SaaS vendors rather than building custom AI tools in-house.
- Juan advises using vendors like Workday or SAP because they embed institutional controls and reduce early operational risk.