Episode publish date
October 15, 2025 6:00 AM (UTC)
Last edit date
Oct 21, 2025 7:39 AM
Last snip date
October 19, 2025 8:06 PM (GMT+1)
Last sync date
October 21, 2025 8:15 AM (GMT+1)
Show
The AI in Business Podcast
Show notes link
Snips
10
Warning
⚠️ Any content within the episode information, snip blocks might be updated or overwritten by Snipd in a future sync. Add your edits or additional notes outside these blocks to keep them safe.
‣
Your snips
‣
[04:20] AI Value Requires Organizational Alignment
‣
[04:55] Census Drop Shows Early AI Disillusionment
‣
[10:24] Focus On Few Transformative Use Cases
‣
[10:33] Chain Use Cases Into A 'String Of Pearls'
‣
[12:06] Hire AI-Native Translators, Not Just Foundational Scientists
‣
[13:00] Make Change Management A Core Investment
‣
[20:23] Quality In Research Multiplies Development Value
‣
[23:05] Real-World Testing Limits In-Silico Gains
‣
[26:45] Useful Hallucinations In Discovery
‣
[28:01] Different Functions Demand Different AI Strategies
What I Learned Today
Key Concepts
- AI value in life sciences varies by function: Research focuses on quality improvement of drug candidates, development emphasizes process acceleration, and commercial targets revenue uplift and patient experience.
- Quality multiplies value in drug R&D: Because development costs far exceed research costs, improving the quality of molecules entering the pipeline creates exponential value—better research outputs justify the massive downstream investment.
- In-silico vs. real-world gap: AI can generate novel molecular designs from a virtually infinite chemical space (10^60 possible molecules), but physical synthesis, testing capabilities, and cost-effective manufacturing must industrialize to keep pace.
- Productive hallucinations exist: In drug discovery contexts, AI "hallucinations" can be valuable when domain experts evaluate and iterate on novel outputs—not all hallucinations are harmful.
Practical Applications
- Tailor AI adoption to organizational culture: Different functions in life sciences have different time horizons and success metrics—creative research (years), process-driven development (months-years), and quarterly-focused commercial operations need distinct change management approaches.
- Test AI-generated molecules systematically: Drug discovery still requires in vitro validation; the process isn't purely in-silico until human trials, so testing infrastructure is critical.
- Focus on asset quality in research: Even modest improvements in candidate quality justify AI investment because they multiply value throughout the expensive development cycle.
Questions for Further Exploration
- How can organizations accelerate real-world synthesis and testing to match AI's in-silico generation speed?
- What metrics best capture quality improvements in research assets entering development?
- How do successful life sciences organizations balance AI experimentation across their different functional cultures?
Personal Reflections
What resonated most: The idea that different parts of the same organization need completely different AI strategies based on their culture and time horizons—not a one-size-fits-all approach.
How this connects to my work:
Action items: