What does "compounding intelligence" actually look like for Pharma BD & strategy teams? Well, it doesn't start with AI.

It starts with how the best operators on these teams already work: define the decision, identify the few questions that would prove or disprove the thesis and then work backwards to the evidence required.

Imagine evaluating a Phase 2 oral asset in a market dominated by injectables:

• Does its efficacy hold after normalizing across trials?
• Can oral convenience offset a modest efficacy gap?
• Is there a patient segment where that trade-off wins?

The historical bottleneck has been turning those questions into reliable data without spending weeks manually aggregating and cleaning it, but that's beginning to fall (this is where Sleuth can slot in).

With the right data, AI can become much more useful. It can quickly process this data and then identify the next question to dig into, supporting the natural diligence process. For example, a landscape reveals a crowded mechanism, which prompts a differentiation analysis, which uncovers an overlooked patient subgroup, which ultimately might change the deal thesis.

Each of these analyses leaves behind tangible artifacts (a dataset, a deck, a memo). Just saving that in SharePoint or OneDrive is not enough though - to become reusable institutional knowledge, the artifact also needs to preserve the decision trail behind it: what was asked, what evidence was available at the time, and how the team reached its conclusion.

That turns an output into a decision-grade artifact: something another team can understand, trust, update, and reuse.

With that feedback loop in place (question → data → analysis → artifact → next question) the remaining principles become clearer:

• Every conclusion should create a monitoring trigger. As the underlying evidence changes, the relevant datasets and analyses should update. Over time, teams move from reactive to predictive intelligence.
• You want your best people's judgement to be captured without exposing sensitive internal information. This requires thoughtful decisions on which layers can be externally fed in vs. must be owned & managed internally.
• All knowledge needs the right access controls - some should be centralized across the org, while some should remain within a team.

So compounding intelligence is not just using AI to make outputs, it's actually building a system in which every piece of work improves the next decision.

We built Sleuth around this loop. Every analysis in the platform keeps its evidence trail, updates as the market moves, and stays reusable by the next team that needs it. If you want to see what that looks like on a question your team actually has, book a demo.