Enterprise AI Innovators

The episode contains several substantive, non-obvious insights about AI deployment in pharma - particularly the framing of 'clean sheet' thinking, the cultural reframing from science company to data company, and the specific mechanics of the AI accelerator (6-8 people, 12 weeks, bi-weekly checkpoints to prevent distraction). However, roughly 30% of the conversation consists of soft motivational content, lightning-round platitudes, and restated frameworks that dilute the density of actionable insight.

“let go of all of our biases about why there needs to be a spreadsheet to do this and just ask, what is the spreadsheet doing and why does it exist in the first place”

“making them too small to fail in a public or embarrassing way was really important”

Greg offers a few genuinely fresh angles - particularly the idea that constraints are 'old assumptions that AI quietly made obsolete' and the reframing of negative experimental results as valuable training data rather than waste. The AI accelerator structure is concrete and non-obvious. However, the broader framework (hypothesis-driven testing, top-down + bottom-up prioritization, psychology beats technology) recycles familiar executive playbooks, and the closing advice leans heavily on Charlie Munger quotes and well-worn change management wisdom.

“what you'll find when you really ask that question, you start to find that there are certain reasons for it that are baked in constraints that when you really look at it from an AI perspective, are no longer constraints”

“negative results matter as much as positive results”

Greg Myers as CDTO of a top-10 pharma company ($45B revenue, 30k employees) managing several thousand people is highly relevant and operating at genuine scale. He has concrete authority over AI deployment across R&D, clinical, and commercial functions. His background spanning life sciences, Motorola, and agriculture gives him cross-industry perspective. This is a practitioner with real P&L responsibility and decision-making authority, not a consultant or evangelist.

“Bristol Myers Squibb is one of the top 10 pharmaceutical companies”

“I run an organization of, uh, several thousand people”

The episode excels in concrete metrics and named examples: the lung cancer CT scan study (14 sites, one year, 2x increase in diagnoses, 116 new cases identified, five-year survival window). The AI accelerator structure is specified precisely (6-8 engineers, 12 weeks, two-week sprints over six cycles). The organizational numbers ($45B revenue, 30k employees) and BMS's claim of 'four commercially scalable AI products' are provided. However, most other claims lack specifics - the 2.5% drug success rate improvement claim is vague on methodology and timeline, and many strategic initiatives are described at 10,000 feet without concrete numbers.

“using AI to read those CT scans, this uh, is about 14 sites, different clinical sites, over about a year long study that we, we did, it led to a two times increase in the number of newly diagnosed patients in non small cell lung cancer. Now if you figure that that is that flagged about 2,600 patients in that one study alone and, and identified 116 new lung cancer patients”

“we give them six sprints, so 12 weeks. And over those 12 weeks every two weeks they come in and present what they've achieved”

The hosts ask solid setup questions that pull concrete examples (lung cancer study, AI accelerator structure). However, follow-ups are often shallow or miss opportunities to pressure-test claims. When Greg mentions four AI products, neither host asks which ones, what the failure rate was, or comparative efficacy data. The 80% drug failure rate claim goes unchallenged. Sam's question about 'how do you square velocity with regulation' is reasonable but doesn't push into specifics. The lightning round becomes softball territory. Notable absence of genuine pushback or skepticism.

“What are some of the most exciting use cases of AI that are currently being deployed at bms?”

“how do you square that with like AI iteration, AI velocity?”