AI labs are pouring money into drug development, selling a future of medical abundance. The pitch is simple: drug development takes intelligence, artificial intelligence is abundant, so drugs should be abundant, too. With enough liquid handlers, and enough inference, all the hard problems of medicine are solvable in a few years. There is no problem in mathematics, physics, biology, chemistry, or medicine, that won’t dissolve under enough funding and enough compute.
Drug development, though, has historically been stubborn, resisting technological advances and becoming more expensive, year over year. While large language models are new in widespread use, machine learning is not new. For decades technology has promised to reduce drug development to a big math problem, then solve it; to-date this has changed nothing. There is no amount of inference you can buy that will improve drug development, as long as you’re pointing it at the wrong thing.
Regulatory interactions define the success of a drug. Having the best drug, data, team, model, or budget is irrelevant if you cannot convince a regulator that the benefit is worth the risk. Expanding the top of your development funnel from 100 candidates to 1000 candidates to millions is just a science project, unless you have a plan for navigating approval.
Fortunately, the same technologies that are supercharging drug discovery can be also be deployed in program planning. Besides generating an ocean of binding data, AI can be turned towards institutional knowledge about what actually makes drug succeed. Regulatory intelligence is abundant, and medicine can be abundant too. The missing connection is not more simulation or 96-well plate data, it is automatically applying the overwhelming and disorganized regulatory intelligence available within every single step of a drug’s progression from plate to patient.
Enter Drug Development’s Planning Problem
We know a thing or two about submissions. Submission Builder is our flagship product adopted by pharma companies both large and small. And most problems we see at submission time take shape long before anyone begins assembling the dossier. Like clay that cracks in the kiln, early planning missteps often surface only once the heat of regulatory scrutiny has been applied.
This planning-related problem isn’t hypothetical. There are organizations we’ve interacted with in the past led by seasoned regulatory experts that ran several times more nonclinical studies for an IND than they needed to. What was supposed to take months actually took years, and they estimate the wasted cost in the tens of millions. Other plans may have gaps where evidence matters: a pivotal toxicology study run in the wrong species, or a shelf-life claim the stability data can’t support. Because planning happens one study at a time, across teams and years, no one sees the whole program until it’s too late to fix cheaply.
The problem wasn’t the science. It’s because the team wasn’t aligned early on what the program actually needed to show.
So why does this problem still exist?
Drug developers have never had more regulatory knowledge at their fingertips: a deep shelf of FDA guidance, warning letters showing where others have gone wrong, experts with decades of experience, and (at larger organizations) years of questions and answers from health authorities scattered across storage spaces.
Our view: Most drug development programs are planned with the molecule outward. Scientists ask (reasonably), what they can do with the asset in front of them:
- What studies are possible?
- What data would be interesting?
- What should the next experiment be?
These are good scientific questions, but they’re not the questions regulators ask. Regulators want to know what you’ve studied, whether those studies are sufficient, and whether you’re presenting the right evidence for a product in this population.
Other engineering disciplines have a leg up here. Aircrafts aren’t designed by asking what the materials can do and hoping certification works out; they start from the requirements and design backward. Drug development, for all its rigor, mostly hasn’t made that shift. When planning starts from what’s possible and not from what’s required, teams hit the speed bumps mentioned earlier. Add shifting priorities, leadership, and personnel, and programs easily lose their way.
The problem was never a lack of data. The challenge is turning that data into a living, realistic plan; one grounded in requirements and kept current as the program evolves.
That’s why we built Program Planner: to put institutional knowledge to work on the plan itself, before decisions harden into studies, budgets, and timelines.
Introducing Program Planner
Program Planner is built for the people who make program-level decisions: program leads, regulatory teams, and the executives who depend on them.
How Program Planner works:
The program lead uploads the planning material and other relevant internal documents they already have, and Program Planner turns it into a structured plan of studies and endpoints across clinical, nonclinical, and CMC.
Weave AI reviews that plan against a curated library of 300+ parsed guidance documents that are updated regularly from the FDA, CFR, ICH, EMA, and similar agencies, and flags what’s missing or misaligned, with every finding tied to its source.
Next, plans will connect to study data as it comes in, so they keep pace with the program instead of going stale.
Unlike a one-off chat with an AI tool, the plan persists: your team can edit and revisit it as the program evolves.
Interested in learning more? Contact our team to schedule a demo.