Re-Thinking the Pipeline: Development in the Age of AI

Artificial intelligence (AI) is reshaping drug discovery. With rapid in silico analysis of vast datasets and predictive modeling, previously “undruggable” biologic targets are now feasible, enabling the design of custom-built molecules with highly optimized target docking and improved safety. All of this is done at AI speed and moves screening from physical labs to virtual simulations, reducing the cost and time from target to candidate selection.  What once required years of iterative screening can now be simulated in hours.   

However, many AI-designed molecules that exhibit promising in silico druggability carry developability challenges, such as poor solubility, stability limitations, and poor manufacturability. These issues carry real costs. They can drive material loss, extend timelines, and force repeated scale-up work, particularly when processes behave differently across development phases.

As AI accelerates the front end of the pipeline, success increasingly depends on getting development right early. Nimble, scalable formulation strategies can preserve timelines, protect asset value, and prevent avoidable rework downstream.

The Growing Gap Between Discovery Speed and Manufacturing Reality

Discovery teams are no longer constrained by the historical limits of medicinal chemistry. AI can generate binding hypotheses, rank candidate molecules, and optimize virtual scaffolds at scale. Industry forecasts suggest that AI-enabled discovery platforms are already reducing early-stage discovery timelines by 20 to 50 percent and cutting early development costs by up to 50 percent. This presents a significant opportunity in an industry where traditional development still takes more than a decade and averages $2.6 billion per approved therapy.

But the expanding chemical space has consequences. 

Many AI-generated molecules push far beyond the boundaries that traditional formulation technologies and manufacturing approaches were designed to handle. It is well publicized that 40% of approved drugs and nearly 90% of drug candidates are poorly water-soluble, and that low solubility directly reduces their developability. In practice, this means many of the molecules AI identifies as most promising are also the ones most likely to encounter serious delivery and scale-up challenges. Many AI-designed molecules have high melting points and exhibit poor solvent solubility, narrowing the available formulation strategies or translating to risky and costly manufacturing processes.  

Where Promising Molecules Encounter Their Real Limits

AI is helping the industry push past undruggable biology by expanding what can be designed in silico. The more difficult challenge is what follows. Turning those designs into bioavailable, stable, scalable, patient-ready products still depends on formulation science and process know-how, not algorithms. That is where CDMOs become essential.

An agile development partner can intervene, helping to overcome developability challenges by offering advanced delivery and formulation strategies. In doing so, they preserve options, reduce rework, and prevent promising assets from being abandoned prematurely.

Why Development Strategy Matters More Than Ever

For AI-designed molecules to reach the clinic, developers must answer three questions very early:

1. Can it be formulated to achieve therapeutic levels in the body?
2. Can it be formulated to achieve acceptable product stability?
3. Can it be manufactured reproducibly and in compliance with regulations?

These are problems that AI has not yet solved because they depend on physical behavior, rather than theoretical optimization. This is where development expertise becomes strategic.

Early-phase development partners can rapidly assess deliverability through preformulation, rapid small-scale prototyping, in vitro and in vivo performance studies, and thorough accelerated stability studies. When done well, this work does more than determine whether a molecule can advance. It helps teams choose development paths that will hold as programs scale. Agile development groups increasingly rely on a set of complementary approaches:

• Fusion-based amorphous dispersion technologies that bypass solvent and heat constraints while enabling a broader range of polymers and stabilizers than traditional binary or heat-extrusion systems can support. These modern fusion methods expand formulation flexibility for molecules that fall outside conventional processing windows.
• Tailored excipient systems capable of stabilizing unstable APIs through multi-component combinations of polymers, wetting agents, and functional excipients, far beyond the limited options available in older formulation toolkits. This allows formulators to match excipient functionality more precisely to the unique challenges posed by AI-designed molecules.
• In silico PK modeling married with real-world in vitro and in vivo testing to guide case-by-case optimization of solubility and absorption, particularly for molecules that stretch the boundaries of modern chemical space. These integrated models help predict which formulation pathways remain viable as molecules become more complex.
• Case-by-case formulation rescue that leverages expanded design space, mixing polymers, stabilizers, and surface-active agents, to avoid discarding promising molecules simply because they do not fit the constraints of older delivery systems. This flexibility allows developers to retain their best molecules without compromising potency or specificity.

Together, these approaches allow for more precise alignment between a molecule and its optimal delivery strategy. That precision improves performance, reduces downstream rework, and can strengthen a program’s intellectual property position. Complex, differentiated delivery systems are often more difficult to replicate, providing an additional layer of protection alongside traditional patents. 

The Advantage Belongs to Teams That Build for Reality

AI will continue to accelerate discovery and generate molecules no human chemist could design. What it cannot do is bypass the physical constraints of solubility, stability, manufacturability, or patient use. As chemical space expands, these constraints are no longer secondary considerations. They are defining factors in whether innovation succeeds.

In response, more developers are turning to hybrid development models that bring advanced formulation tools in-house while relying on CDMO expertise for strategy, scale-up, and manufacturability. This approach improves speed, reduces API consumption, and ensures delivery is evaluated as quickly as AI generates new candidates. Leading CDMOs are also applying data-driven insights from prior programs to anticipate which formulation strategies are most likely to scale, reducing avoidable rework downstream. The next generation of successful therapies will combine sophisticated discovery with equally sophisticated development strategy. Teams that engage formulation and manufacturing expertise early will be better positioned to preserve timelines, protect asset value, and move with confidence from concept to clinic.

In modern drug development, designing the molecule is only the beginning. Designing how it will be made, scaled, and used is what turns innovation into medicine.