Despite billions of dollars invested across the pharmaceutical industry, the organizations realizing measurable value from AI are rarely the ones with the most AI pilots. They are the ones that have invested in something far less exciting but far more powerful: a deliberate data strategy.
Over the past two years, I've sat in countless conversations where the question has shifted from "How do we use AI?" to "Why aren't we seeing more value from AI?" It's the right question.
The uncomfortable reality is that AI rarely creates value independently. AI amplifies the quality, accessibility, context, and governance of data that already exists within an organization. In other words, the limiting factor for AI in pharma is no longer the model, it's the quality of data and how the organization operates around it. Increasingly, the organizations creating a competitive advantage are recognizing that data is a business asset, one that directly influences the speed and quality of decision making across the R&D value chain.
The real objective is decision velocity
When organizations talk about becoming "data-driven" or "AI-enabled," the conversation often centers on technology, but technology is not the outcome. The outcome is better decisions, made faster. I increasingly think about this as decision velocity: the ability to move from data to insight, insight to decision, and decision to action faster and with greater confidence than before.
In R&D, decision velocity influences nearly every stage of value creation. It impacts how quickly organizations identify promising targets, prioritize assets, design clinical trials, recruit patients, generate evidence, and make portfolio investment decisions.
Every month saved in development has value. Every poor decision avoided has value. Every increase in confidence has value. This is where a data strategy creates impact, not by creating more data, but by improving the speed and quality of critical decisions.
Why pharma's productivity challenge is really a decision challenge
The pharmaceutical industry has spent decades trying to improve R&D productivity, yet bringing a therapy from discovery to market still takes more than a decade, costs billions of dollars, and carries significant risk. Drug development is inherently uncertain. Failures will always occur. Biology is complex, and there is no technology that can eliminate scientific risk.
AI has generated excitement because it promises to improve productivity. But if we look closely at organizations generating measurable outcomes, a different pattern emerges. The value is not coming from AI alone, it is coming from AI applied to connected, governed, context-rich data ecosystems. The organizations gaining the advantage are not asking, "Where can we deploy AI?" They are asking, "What decisions matter most, and how can we improve them?" That distinction matters because AI scales decisions. A data strategy determines how well decisions are made.
What real value looks like in pharma R&D
It's easy to point to the growing number of AI announcements across the pharmaceutical industry. New partnerships, new platforms, and new models are announced almost weekly. But the organizations creating the greatest value aren't simply adopting AI, they're fundamentally changing how scientific decisions are made. The common thread is the ability to connect trusted data, scientific expertise, and AI to improve decisions across the R&D lifecycle.
Improving target identification. Novo Nordisk's collaboration with Valo Health combines multimodal human datasets with computational modeling to better understand disease biology and prioritize targets with a higher likelihood of clinical success. Rather than replacing scientific expertise, these capabilities help researchers make earlier, higher-confidence decisions about where to invest time and resources. The real advantage comes from integrating trusted scientific data, computational insight, and human expertise to improve the quality of discovery decisions.
Better molecule and lead optimization. Takeda's expanded collaboration with Iambic Therapeutics reflects another important shift. Rather than positioning AI as a replacement for scientific expertise, the collaboration integrates AI directly into molecular design and lead optimization. The objective is straightforward: reduce experimental cycles, improve candidate quality, and help scientists make higher-confidence decisions earlier in discovery. AI isn't replacing scientists; it's improving the decisions scientists make.
Better protocol design and trial execution. Medidata's introduction of AI-enabled Protocol Optimization reflects another important shift. Rather than waiting until a trial is underway to identify design challenges, the platform combines historical clinical trial data and AI to help study teams evaluate protocol decisions before the first patient is enrolled. AI isn't replacing clinical expertise; it's helping clinical teams make better decisions earlier, when they have the greatest impact on study success.
Accelerated R&D productivity. Perhaps one of the clearest demonstrations of value comes from Insilico Medicine. Earlier this year, the company reported reducing its average drug discovery timeline from approximately 4.5 years to about 13 months, with a development candidate nominated in just nine months. Those results were enabled by integrating AI into an end-to-end discovery platform built on connected data, computational workflows, and continuous learning.
Whether those timelines become the industry standard remains to be seen. What they demonstrate is the potential impact when AI is combined with high-quality data and redesigned scientific workflows. Viewed individually, these examples may look like AI initiatives or research collaborations. Viewed together, they reveal something more important: each organization is investing in its ability to make better scientific decisions with greater speed and confidence. AI accelerates those decisions, but a data strategy makes them possible.
That's what decision velocity means in practice: the ability to make the right scientific and business decisions at the right time, using trusted data with sufficient confidence to act.
A data strategy is really a value strategy
One of the biggest mistakes organizations make is treating a data strategy as a technology exercise. A true data strategy starts somewhere entirely different, and it starts with value.
Before discussing lakehouses, data fabrics, governance councils, master data, semantic layers, or generative AI, organizations should be able to answer three questions:
• What business outcomes matter most?
• What decisions drive those outcomes?
• What data is required to improve those decisions?
The answers vary across organizations. For one company, the priority may be increasing Probability of Technical and Regulatory Success (PTRS). For another, it may be reducing protocol amendments, accelerating patient recruitment, or improving portfolio prioritization. The common denominator is simple: data becomes valuable when it improves decisions that matter, not when it is merely collected.
Why AI readiness is really data readiness
Many organizations are currently pursuing AI readiness initiatives, but what does AI-ready data actually mean? In practice, it means data that is:
• Discoverable
• Trusted
• Governed
• Context-rich
• Connected across domains
• Accessible at the speed of decision making
The challenge is that pharmaceutical data rarely exists this way naturally. Clinical data, biomarker data, manufacturing data, regulatory data, real-world evidence, commercial data, and scientific literature often exist in separate systems managed by separate functions. Humans can work around fragmentation, but AI cannot. Organizations realizing value from AI are increasingly investing in semantic layers, knowledge graphs, metadata management, data products, and domain-oriented ownership models, because context, not algorithms, is becoming the scarce resource.
The operating model shift in R&D
The most significant implication of AI may not be technological. It may be organizational.
Historically, many pharmaceutical companies treated data as a centralized IT responsibility. Today, leading organizations are moving toward models where business domains own value while enterprise functions provide standards, governance, and enablement. Equally important, they recognize that while data has become increasingly easy to generate, scientific insights remain far more difficult to understand and share. Scientific decisions depend on developing a common understanding across disciplines, and AI has the potential to accelerate that process by helping teams synthesize evidence, communicate findings more effectively, and develop shared understanding more quickly.
Data governance becomes less about control and more about accountability. Technology teams become product teams. Analytics teams become decision enablement teams. Business functions become co-owners of data assets rather than consumers of reports. Organizations that modernize technology without modernizing their operating model often discover that adoption becomes a consistent bottleneck.
The new competitive advantage
For decades, the competitive advantage in R&D was defined by scientific expertise, intellectual property, and commercial scale. Those remain essential, but another differentiator is emerging: the ability to generate, connect, and learn from trusted data to make better decisions with greater confidence.
At first glance, that may sound like a business advantage, but it's a patient advantage. Better decisions lead to better science, more efficient development, and ultimately, better outcomes for patients. This is why a data strategy matters. Not because organizations need more data, or because AI is the latest technology trend, but because improving decision velocity helps bring life-changing therapies to patients sooner. That is the competitive advantage that matters most.
A practical path forward
For organizations wondering where to begin, the answer is often simpler than expected. Don't start with AI. Start with the decisions that matter most.
The most successful organizations aren't building data strategies around technologies or platforms. They're building them around the decisions and use cases that have the greatest impact on scientific outcomes, development timelines, patient outcomes, and enterprise value. That means asking a different set of questions:
• Which decisions have the greatest impact on R&D productivity?
• Which decisions most influence PTRS and development timelines?
• Where do teams spend the most time searching for, reconciling, or validating data?
• Which decisions would improve if trusted, contextualized data were available at the point of need?
Once those decisions are identified, the path forward becomes clearer. Data can be organized around value. Ownership can be aligned to business outcomes. Governance can focus on accountability rather than control. AI can be applied where it has the greatest potential to accelerate decisions and improve outcomes.
The organizations that do this well won't simply become AI-enabled. They'll become learning organizations, capable of continuously converting data into insight, insight into decisions, and decisions into measurable business value.
The companies that win the next decade of R&D innovation may not be the ones with the most AI. They may be the ones that learn fastest from their data.
Where is your organization spending the most time reconciling data instead of using it? I'd like to hear how other R&D and data leaders are thinking about this.