Over the next two decades, precision oncology will be reshaped by artificial intelligence. AI systems capable of modeling tumor evolution, predicting drug responses, and identifying resistance pathways are advancing rapidly. As multimodal datasets expand, from genomic sequences and proteomics to spatial pathology and real-world clinical outcomes, the idea of designing cancer therapies largely in silico becomes increasingly plausible.
Yet even in this future, cancer remains a profoundly complex, adaptive disease. Tumors evolve under therapeutic pressure, acquire resistant phenotypes, and rely on intricate interactions with immune cells, stromal components, and metabolic cues. While AI can learn correlations and simulate biological scenarios, it still cannot fully recreate the emergent behavior of living tumor ecosystems without an empirical reference point. The inherent variability and contextuality of cancer ensure that some aspects of tumor behavior will always require direct biological measurement.
This is why in-vitro testing will not disappear; instead, its center of gravity will shift decisively toward platforms that possess genuine predictive capacity, functional assay systems that can reproduce in vivo–like drug responses in vitro. Regulators will continue to demand biological confirmation for cancer therapies, not out of skepticism toward AI, but out of necessity: when patient survival is at stake, models must be grounded in evidence that mirrors real human biology.
The in-vitro models that will dominate this future are the ones validated against clinical outcomes. Patient-derived organoids that maintain the architecture, heterogeneity, and behavior of original tumors; co-culture platforms that reconstruct immune–tumor and stromal–tumor interactions; and functional precision oncology systems that have demonstrated correlation between assay results and patient responses will all gain strategic importance. These platforms matter because they do not merely model tumors, they behave like them. They can capture the same drug sensitivities, resistance trajectories, and phenotypic shifts observed in the clinic.
As AI increasingly guides early discovery, narrowing down targets and drug candidates with high computational confidence, these clinically validated functional platforms become the critical biological filter. They provide the essential “ground truth” that AI cannot generate: a physiologically relevant context where drug behavior can be observed, compared, and validated. They reveal which AI predictions hold up in a living system, and which fall apart under the complexity of real tumor biology.
Importantly, these functional systems do more than confirm predictions, they enhance them. By capturing in vivo–like responses, they generate high-quality biological signals that refine AI models, helping them learn nuances of tumor behavior that no dataset alone could ever capture. This creates a powerful feedback loop: AI proposes, biology verifies and enriches, and the combined system becomes increasingly predictive.
This dynamic opens new opportunities. Companies providing clinically validated functional cancer models will become foundational partners in an AI-driven oncology ecosystem. They will supply the biologically credible platforms that drug developers need to de-risk candidates early. They will generate the datasets that strengthen next-generation predictive models. And they will enable precision oncology approaches where treatment decisions are informed by functional responses that meaningfully align with patient outcomes.
Far from being eclipsed by AI, these platforms become indispensable because they do what AI cannot: they reproduce human tumor biology in a controlled, testable format. In a future where computational predictions are abundant, the systems that can faithfully replicate in vivo drug response in vitro become the ultimate arbiters of therapeutic relevance.
The future of cancer therapy development will therefore be hybrid, computationally accelerated, but biologically anchored. And the most valuable assets in this landscape will be the functional in-vitro cancer models whose predictive power has been proven in patients.
