Ramji Vasudevan, Head of Business Units – Life Sciences, Altimetrik
The pharmaceutical industry is far beyond asking whether artificial intelligence (AI) can transform drug development. According to a survey by the Pistoia Alliance,1 68% of life science R&D professionals use AI and machine learning for their work. However, the same study revealed that 52% cited low data quality as the biggest barrier to AI implementation.
This paradox captures exactly where the pharma industry will find itself in 2025. Companies will carry out meaningful experiments, “What can AI do?” But most of it remains between the success of the proof of concept and the value of enterprise scale.
That hurdle is not a problem of technology. The AI model has improved dramatically, with new features emerging every few months as access is increasingly democratizing. The possibilities of AI are always there, and its basic promises have not changed.
However, McKinsey Research states that the generator AI model accounts for only about 15% of typical project efforts.2 The remaining 85% involves adapting the model to the company's internal knowledge base and use cases. This is where most pharmaceutical companies struggle.
The real barrier is about business transformation. Consider a common scenario for pharmaceutical manufacturing. Drug test results are shared via email attachments and PDFs between the contract manufacturer and the testing laboratory. These documents often contain handwritten notes that are scanned into PDFs. Then someone has to manually type in the handwritten notes so they can be entered into their system. It's slow and error prone.
Working with numerous pharmaceutical companies on AI implementation has revealed a clear framework for moving from successful experiments to corporate-scale value. This approach focuses on business outcomes rather than technology features.
Too many organizations start with technology and search for applications. Business outcomes should encourage the digitization and use of tools such as AI. A single IT function cannot effectively solve problems in all different business domains.
Instead, start by identifying specific measurable business issues. Consider generating a clinical research report: It can take weeks to months from the last visit of the patient in the clinical research report. AI can handle 60%-70% of documenting, much faster than humans, but can maintain human surveillance. Don't shoot 90%-95% at first. Start with a modest automation goal and iterate towards optimization.
Step 2: Create a joint accountability
One of the biggest barriers to AI scaling is the traditional gap between business and IT capabilities. A siloed approach cannot consistently provide enterprise value. According to McKinsey, 70% of the conversion failed. This is due to the disconnect between what the business wants and what the IT team offers.3
Gartner recently proposed to have the Chief Information Officer (CIO) and Chief Experience Officer (CXO) of Business Functions join together and hold them jointly responsible for the outcome.4 Each leader owns a domain, but shares ownership of the results. This creates a true partnership rather than a traditional handoff. Gartner says nearly three-quarters of CXOs, who co-lead targets from digital investments.
Step 3: Implement hub and spoke governance
Successful AI scaling requires a balance between centralized governance and federation execution. Central functions must maintain safety, regulatory compliance, and corporate-wide standards, particularly important in the regulated pharmaceutical sector.
However, use case development must be federated to business units. While there is a need to share some responsibility between CIOs and CXOs, safety and regulatory aspects remain centralized at the enterprise level. This allows business units to move quickly, while maintaining the necessary controls. Ultimately, businesses need to work towards democratized AI access that allows business users to build, test their own solutions, scale to production based on results, and discard them.
Step 4: Build a modular solution
Rather than attempting wholesale digital conversion, we focus on bolt-on solutions that do not require an overhaul of the entire system. For example, document processing examples do not require a fundamentally new system. It could be a bolt-on with the possibilities of what AI brings to the table.
Merck and McKinsey have co-developed a platform to generate clinical research reports. This reduced the time to generate the first draft from 180 to 80 hours, while reducing errors in half.5 It's a clear opportunity for AI augmentation without the need for a complete infrastructure replacement.
Step 5: Plans for democratization
The ultimate goal is to create an infrastructure that allows business users to access data, calculate power, and directly access AI models. That infrastructure will dramatically change the way an organization operates. Anyone with the right credentials can access data, calculations, storage, and AI models, connect analytics, and make informed business decisions in real time. The impact is almost instantaneous. Instead of a simple dashboard 3 month wait time, business users can create solutions in the afternoon. However, this democratization must be built on the basis of solid data to achieve a single source of truth, rather than maintaining data silos.
Competitive advantages in the future
While many organizations focus on experimenting with models and use cases, which is 15% of their efforts, the real value lies in solving the 85% challenges of business adaptation and integration. Companies that master this business-first approach to AI scaling create sustainable competitive advantages.
The goal is to allow organizations to operate at the speed of business decisions rather than at the speed of IT implementation. This will allow you to move from a multi-month process to real-time insights and behavior, bringing life-saving treatments to patients faster.
Success requires moving beyond the technology-first thinking that characterizes much of the industry's AI journey. Companies seeking to make business change efforts capture the possibilities at an enterprise scale.
Ramji Vasudevan is Head of Business Units – Life Sciences, Altimetrik
reference
1. What will the lab look like in 2030? Pista Alliance. https://marketing.pistoiaalliance.org/hubfs/lab%20of%20the%20 -future%20reports/lab%20of%20the%20future%20survey%20results%202024%20.pdf
2. Generated AI in the pharmaceutical industry: A transition from hype to reality. McKinsey & Company. January 9th, 2024. https://www.mckinsey.com/industries/life-sciences/our-insights/generative-ai-in-the-pharmaceutical-industry-moving-from-hype-to-reality
3. A perspective on conversion. McKinsey & Company. https://www.mckinsey.com/capabilities/transformation/our-insights/perspectives-on-transformation
4. Cultivating a CIO partnership with CXOS takes on a new urgency. Gartner. https://www.gartner.com/en/chief-information-officer/insights/cio-partnerships
5. GENAI, Merck, McKinsey Transform Clinical Authoring. McKinsey & Company. https://www.mckinsey.com/about-us/new-at-mckinsey-blog/with-gen-ai-merck-and-mckinsey-transform-clinical-authoring
