Mike Ryan, BPN CEO, is a former Goldman Sachs analyst who later ran its global equity business and managed Harvard’s $18 billion endowment, and we spoke about why powerful AI still fails investors when its answers cannot be trusted. After repeatedly receiving polished but incorrect information from generic tools, he decided to develop a more reliable approach. As he puts it, “AI wouldn’t pass a first-round job interview at most firms because it’s not trustworthy.”
Ryan explains that AI has a “big stomach, but a very small mouth”: it can process enormous volumes, yet each answer depends on the limited information selected for that prompt. His method maps every question to the most reliable and relevant sources, uses trusted spreadsheets for calculations, preserves citations and source controls, and keeps one person directing the process through an “AI plus 1” model. Purpose-built agents can screen opportunities, identify the one or two highest-value priorities, and support complex decisions as new evidence arrives. The result he describes is decision-grade memos, models, and presentations produced in 80% less time, with templates or first drafts often completed within one or two days.
For listeners, this is a practical blueprint for reducing processing work while preserving human judgment, accountability, and confidence in consequential decisions.
Key takeaways
- Map every AI prompt to the most reliable, relevant sources.Keep one human responsible for supervision, interpretation, and final judgment.Use trusted spreadsheets for calculations, then visualize results for faster review.Let AI screen opportunities before committing time to deep analysis.Update complex decisions iteratively as new evidence arrives.Use saved time for company visits, customer calls, debate, and judgment.