There is a popular vision of where AI is taking science: models will become so capable that they will make the key decisions for us. They will identify the right target, choose the best molecule, predict what will work and tell us what to do next.

Maybe someday. But not yet.

The constraint is not simply AI capability. It is that we still know remarkably little about biology.

Biology is extraordinarily complex and context dependent. Change the cell type, disease state, concentration, timing, patient population or one of hundreds of other variables, and the result can change. AI can learn from the observations we have generated, but in much of biology we simply have not generated enough of the right observations to reliably predict what nature will do.

That is why we think we may be asking the wrong question.

Instead of asking, ‘When will AI predict biology and discover drugs on its own?’ we should be asking ‘How much faster can AI make the scientific method?’

That is a very different question, and one we can answer today.

The scientific method isn't going anywhere

Science advances through a simple, powerful loop: idea → design → experiment → analysis → decision → repeat.

AI does not have to eliminate that loop to transform science. It just has to make the loop faster.

A huge amount of time around an experiment is spent finding the right technology or supplier, writing specifications, getting quotes, navigating compliance, coordinating samples, locating prior work, dealing with payments, interpreting results and deciding what to do next. Much of that still happens through emails, spreadsheets, searches, forms and waiting.

AI can remove that friction now. It can help scientists design better requests, identify the right providers, surface relevant knowledge, embed compliance directly into workflows, automate transactions, summarize results and frame the next experiment.

None of this requires AI to solve biology. It requires AI to help science move faster.

Increasing scientific velocity

The metric we think our industry should focus on is scientific velocity: how quickly can an idea become reliable experimental data, and how quickly can that data lead to the next useful experiment?

A five-minute improvement once is trivial. Repeated thousands of times, it matters. A one-day delay removed from one experiment is nice. Remove that same delay across dozens or hundreds of experiments and you can take months out of a program.

This is where the gains compound. Faster science means more experiments. More experiments mean more data. More data mean better models, and better models help us choose better experiments. Each cycle improves the next one, creating what we think of as compound innovation: a virtuous cycle in which faster experimentation generates better knowledge, better knowledge improves AI, and better AI helps science move faster still. Ironically, accelerating experimentation may be the fastest path to the predictive future everyone is waiting for.

Scientists remain at the center

The scientist does not disappear in this model. The scientist becomes more powerful.

AI can increasingly handle the searching, organizing, coordinating, checking and administrative work surrounding experimentation. That gives scientists more time to do what scientists do best: ask interesting questions, recognize unexpected results, connect seemingly unrelated observations and decide which questions are worth pursuing.

The scientific method has endured because it is a way of dealing with what we do not know. And biology still contains an extraordinary amount that we do not know. Eventually, AI may predict biological systems with an accuracy that seems unimaginable today. We hope it does. But we do not need to wait for that future.

The near-term promise of AI in life sciences is not that we stop doing experiments. It is that we do better experiments, make better decisions between them, and move from one experiment to the next dramatically faster.

We are not replacing the scientific method. We are accelerating it. And ultimately, time saved in R&D is not merely money saved. It means getting to a result sooner, making a decision sooner and, when we get it right, getting a treatment to a patient sooner.

‍