Lilly opened access to AI models trained on $1 billion in research.

On September 9, 2025, Lilly announced TuneLab, giving selected biotech partners access to drug-discovery models trained on proprietary research data that Lilly said cost more than $1 billion to obtain.

That is a meaningful change in access to research capability.

It is also a more precise story than a company giving its data away. Lilly described federated learning that lets partners use the models without directly exposing either party’s proprietary data. Selected partners contribute training data in return for access.

The strategic question it raises for me is how a company can help others build on its expertise while protecting the knowledge that makes the capability useful.

The cost of learning

AI can make parts of research easier to explore: searching material, comparing explanations, identifying questions and preparing for expert conversations. That can widen the set of opportunities a team is able to investigate.

It does not remove the need to check evidence, understand a market or validate a scientific result. Access to a model and proof that a product will work are different milestones.

Revisit the boundaries of your strategy

Some adjacent opportunities were set aside because learning enough to evaluate them looked too expensive. Those decisions are worth revisiting as research tools change.

The useful move is to test whether you can learn enough to make the next decision at a lower cost. Some opportunities will still be outside your capabilities. Others may now be practical to explore.

The same question applies to competitors. What might they now be able to understand about your market that was previously difficult to investigate?

The new strategic metric: initiative velocity

Here's the KPI that should emerge on your exec dashboard:

Initiative velocity: how many new frontiers can your organization meaningfully explore at the same time?

Meaningful meaning: research conducted, experts consulted, feasibility assessed, directional decisions made. Not a dashboard full of zombie pilots. Actual exploration.

Research capacity is one constraint. Management attention, specialist judgment, access to customers and the ability to act on findings are others. A wider exploration program still needs owners and clear decisions.

The compounding advantage

Exploring multiple domains at once doesn't just add, it compounds.

Exploring several domains can reveal connections you would miss by studying each in isolation.

Cross-domain insights generate moves that single-lane competitors can't see. New adjacencies. New customer segments. New business model patterns lifted from industries you weren't previously watching. The value isn't the ten individual explorations. It's the intersections between them.

What comes next

Start with human-directed research: a defined question, a source standard and a decision the findings need to support.

As the workflow becomes repeatable, automate the steps that can be checked reliably. Keep uncertain evidence and consequential decisions visible to the people responsible for the work.