“

Do you have specific data nobody else has on a particular process? That’s my proxy for whether a project needs custom modeling versus off-the-shelf SaaS AI.

“

This isn’t scientist augmentation or one-to-one Copilot use—it’s rethinking how you do R&D. Specialized AI agents can generate millions of candidates, simulate them, filter them, analyze them, and loop back—all integrated with scientists.

“

On AI’s role in high-stakes financial decision-making: how do you measure your ‘return on attention’? As in, if you have a certain amount of time, where do you focus it.

“

Have an agentic strategy. Understand where your agents are going to be created, understand what you want to accomplish, and make sure they’re using protocols and standards that are widely available and interoperable. That’s how you avoid throwing away agents—how you avoid agents on islands.

“

Digital is really the automation of data collection from a business process—whether that’s patient data, a manufacturing line, or lab instruments.

“

Probably the biggest difference in that kind of culture of innovation is a comfort with failure.

“

If you don’t treat this as an iterative, ongoing, living process, the technology will move past your strategy.

“

The big change now is how you transfer from things being automated to things being intelligently controlled—and you need to be very confident these AI systems aren’t going to screw things up. When an LLM hallucinates, that’s just a bad answer; in a logistics or manufacturing operation, that’s lines down.

“

The difference from Excel spreadsheet modeling to AI is that you will have to hire dedicated data-science expertise, scoped to that specific problem—that’s the real line between deterministic modeling and machine learning.