“When you’re a hammer, everything looks like a nail.” Maslow’s famous line succinctly captures where many life sciences organizations find themselves with AI today. The hammer, in this case, is generative AI, an astonishingly powerful tool that seems to promise everything from accelerated molecule discovery to automated lab reporting.
But as with any new instrument, the value depends entirely on how, where, and when it’s used.
A July 2025 MIT Project NANDA report, The GenAI Divide: State of AI in Business 2025 made waves for finding that despite an estimated $30–40 billion in enterprise investment into generative AI, 95 percent of organizations are getting zero measurable return. The gulf between pilot-project enthusiasm and true enterprise transformation is jarring. Yet “zero return” does not mean zero progress. Rather, it signals how much potential remains locked up inside early pilots that have not yet scaled or connected to broader R&D workflows.
Leading pharma and biotech companies are already generating real results, particularly in high-friction workflows like molecule review and formulation development. But these wins too often remain islands of innovation, promising signals that struggle to propagate across the enterprise. And for life sciences CIOs, it raises some fundamental questions: Is your organization architected to scale AI beyond pilots, to change how work actually gets done? Can your data, workflows, and culture sustain AI at scale, or are they still built for experimentation rather than integration?
Few industries stand to gain – or potentially lose – more from AI than life sciences. The technology’s promise to compress R&D cycle times, improve clinical predictability, and enable faster, safer therapies is immense. Yet, as Shahram Ebadollahi, principal at Nav.AI and former Chief Data & AI Officer at Novartis, noted in a recent panel discussion: “Eighty percent of success and getting value from AI doesn’t have anything to do with technology.”
In other words, modern platforms and ML tooling are essential, but they form only the foundation. True transformation begins when organizations redesign processes, rewire workflows, and embed intelligence into day-to-day decisions. Without that structural shift, even the best use cases struggle to scale beyond their original context.
In the early 20th century when electricity first arrived in factories, most businesses simply swapped steam engines for electric motors without redesigning production lines. Productivity did not soar until the process itself was reinvented. AI in R&D follows a similar trajectory where success hinges less on what AI can do and more on how organizations reshape their ways of working to use it.
Shahram calls these “AI-able processes” or workflows where AI can play a natural role in the decision and action loop. And the most mature organizations are shifting toward a future where AI becomes practically invisible, seamlessly embedded into the fabric of R&D, quietly accelerating science rather than calling attention to itself.
Why AI Momentum in Life Sciences Stalls
Life sciences R&D operates within one of the most complex and distributed information ecosystems on earth. Every discovery, experiment, and regulatory decision generates mountains of structured and unstructured data, from genomic sequences and imaging files to clinical notes, spreadsheets, and more. Yet much of this data remains siloed, fragmented, and inconsistently governed.
Three persistent challenges tend to hinder R&D teams from successfully scaling their AI projects:
- Volume and variety of data: AI systems thrive on integration, not isolation. But R&D data is often locked away in function-specific systems or legacy LIMS databases, making multimodal learning and analysis nearly impossible.
- Explainability and trust: In a highly regulated domain, it is not enough for AI to be right; it must show how it reached its conclusions.
- Organizational inertia: Siloed accountability and incentive structures discourage collaboration across IT, data science, and domain teams.
Despite these constraints, early adopters are already proving out high-value patterns.
In the molecule review process, one of the most labor-intensive decision points in pharmaceutical R&D, AI systems now automatically ingest data from across functional domains, model competitive and clinical scenarios, and surface decision-ready insights. The result: portfolio decisions made in days instead of months, with greater transparency, consistency, and scientific rigor.
Similarly, in formulation development, AI retrieves and interprets multimodal data such as text, images, and lab reports, and recommends next steps or flags stability risks. Generative design models are now even proposing novel molecular structures and formulation strategies, providing chemists with a scientifically grounded starting point rather than having to start from scratch.
Without shared infrastructure, common data standards, and clear ownership, organizations risk creating a patchwork of AI experiments that, while useful in isolation, are incapable of driving durable competitive advantage.
3 Questions Every CIO Should Ask
In order to move from isolated AI proof of concept pilots to repeatable business value, life sciences CIOs must address these three critical questions:
- Is AI championed by the business, or confined to IT? Transformation may begin with the CIO, but it only scales when business leaders, from R&D and clinical operations to regulatory, treat AI as a strategic enabler. When AI outcomes connect directly to established corporate KPIs such as cycle-time reduction, submission quality, and patient safety, the work becomes durable. As Shahram emphasizes, this is less about the technology itself and more about how the business can better align it with scientific and operational outcomes.
- Are incentives and accountability shared across teams? AI rarely fails for lack of algorithms. Rather, it fails for lack of alignment. Incentives must connect data science, biostatistics, clinical, and regulatory teams. In practice, that might mean tying performance goals to joint outcomes, ensuring that both technical accuracy and scientific impact are rewarded appropriately. Forward-looking organizations that structure incentives this way find that collaboration becomes self-reinforcing, allowing AI adoption to spread organically.
-
Chris McSpiritt
Do your systems enable integration across all R&D data types? AI will only be as strong as the data that supports it. Life sciences organizations generate terabytes of structured and unstructured information every day, but without a unified infrastructure, shared standards, and interoperability between modalities, every new model inherits the same limitations. A scalable AI ecosystem requires a common language for data, not a collection of disconnected silos.
Life sciences organizations are at an inflection point. Generative models, multimodal data fusion, and AI-driven agents are already reshaping discovery, development, and clinical execution. Yet technology alone will not bridge the gap between innovation and impact. When AI is no longer viewed as a blunt force instrument to be wielded at isolated problems, but rather woven into an organization’s operational and clinical fabric, you will know your R&D organization is truly ready to scale AI.
Chris McSpiritt is VP of Life Sciences Strategy, Domino Data Lab
