Laying the foundations for AI transformation in life sciences

The life sciences industry’s agenda is certainly full. Between now and 2030, patents will expire on dozens of major drugs, placing up to $300bn of revenue at risk. Yet failure rates in drug development remain high and the risks involved in redesigning product portfolios via M&A and licensing deals are substantial. Meanwhile, the shift from traditional medications to biologics, cell-based and gene therapies is complicating innovation, manufacturing and supply chains. As healthcare systems attempt to reduce the cost of medication, tariffs have increased pricing pressures. In addition, elevated interest rates remain a persistent challenge and conflict has disrupted the industry’s intricate supply chain networks.

Only 18% of senior executives in the life sciences sector say they are very positive about growth prospects in 2026 according to our C-suite barometer, Forvis Mazars’ annual survey of business sentiment worldwide. By contrast, at the start of 2025, 37% felt this way. This finding is one of a number that point in a similar direction. For example, the number of executives in the sector who report that international expansion or new product launches are top priorities has declined significantly.

Under these circumstances, it is tempting to conclude that much of the pharmaceutical sector has either hit a complexity wall or is approaching one. The expression describes the sudden realisation that a company’s existing skills and processes are no longer sufficient to support growth in the face of multiple sources of disruption and increasing complexity. In many cases, AI-driven digital transformation is the only viable response that offers the potential to accelerate the innovation on which the industry depends.

Building a more efficient drug discovery pipeline

According to industry estimates, development of a new drug can take over a decade and cost over $1 billion. The likelihood that a new medication entering the R&D process will be approved for use on patients is just 14.3%. Key factors include the complexity of the science and technology involved, increasing compliance costs and the need to develop therapies for complex multifactorial conditions.

AI’s potential to enable a broad transition from experimental science to computational engineering is the only way out of Big Pharma’s R&D dilemma. Most pharmaceutical companies have already developed solutions based upon generative AI to scan and analyse vast quantities of scientific documentation in search of promising new drug discovery candidates. AI can work with noisy, heterogeneous or experimental data, detecting patterns and properties that human researchers might miss within datasets generated by failed clinical trials, for example. In addition, AI will increasingly supplement lab research with in silico (i.e. computerised) experimentation. AI systems also have the potential to predict how specific compounds will interact with the human body, improving the sector’s ability to identify therapies with the highest probability of success (POS). The ultimate goal involves transforming drug discovery into a highly automated learning process in which data drives a dynamic cycle of hypothesis and validation. Management teams should prioritise the following steps from the start of AI transformation:

• Building the technology infrastructure and governance policies required to make R&D data findable, accessible, interoperable and reusable. These design principles, known as FAIR, create the basis for responsible AI deployment by enabling traceability, explainability, bias mitigation and privacy compliance.
• Investing in the talent required for management and execution of data-driven R&D. In particular, AI transformation teams should adopt a hub-and-spoke organisational model that combines centralised resources for governance and infrastructure with local on-the-ground insights into use case scenarios and pain points.
• Resisting the temptation of innovation for its own sake, focusing instead on clear, measurable goals such as reducing timelines and increasing success rates.
•  Making decisions to scale, pivot or retire initiatives every quarter based upon updated potential to deliver measurable ROI.
• Building compliance into the design of transformation programmes by engaging regulators at the earliest possible stage (i.e. pilot projects) and implementing clear guardrails for explainability and auditability.

AI-driven transformations will not deliver broad-based success immediately in an industry as complex and regulated as life sciences. However, companies like Eli Lilly and many of its rivals are already investing heavily in proprietary data, supercomputing, robotics, automation and AI. Here and elsewhere, the future of life sciences is already being built upon foundations including high-quality infrastructure, superior data management and multimodal AI.

Balancing innovation with compliance

Multiple opportunities exist for AI-driven transformation of the pharmaceutical sector’s manufacturing base. Faster, more reliable manufacturing could result in double-digit revenue uplifts across the sector, accompanied by significant improvements in time to market. Collaboration between national regulatory authorities holds out the promise of global baseline standards. The multiple regulatory challenges include the need for AI to operate in an ‘explainable’ fashion, with high levels of validation and control.

Real-time data exchange will require improved data quality and standardisation, interoperable data platforms, localised high-speed connectivity and improved cybersecurity. None of this will occur overnight, but increasing regulatory clarity has the potential to act as the catalyst for scaled-up manufacturing transformation in four key areas:

• Optimised production: machine learning and digital twins offer the potential to optimise production planning and conduct multivariate, in-flight tuning of production lines. Flexible production relying upon modular equipment for rapid re-configuration will increasingly become the norm for small-batch, personalised medications.
•  Quality control and quality assurance: traditional QC and QA processes are labour intensive. By contrast, computer vision systems combined with deep learning algorithms enable continuous in-process output checks. In addition, AI can analyse deviation data to predict quality thresholds over time.
•  Predictive maintenance: AI-driven monitoring systems can substantially improve maintenance inspection routines, reduce downtime and extend equipment lifespans.
•  Compliance AI: the compliance burden in manufacturing plants will increasingly be streamlined by AI systems that exploit access to real-time data in order to generate compliance documentation.

Supply chains remain a target for upgraded technology

Four out of 10 executives in the sector expect the continuing supply chain disruption to act as a brake on growth this year, according to our global report: Strengthening supply chains. Significant investment in digital systems that improve operational agility is enabling better responses to geopolitical and climate change disruption (e.g. ‘what if’ scenario planning). In general, pharmaceutical companies are working to develop end-to-end visibility across their supply chains. However, incompatible datasets and legacy manufacturing systems used by suppliers frequently complicate the process.

“After Covid-19, risk management changed. The ideal supply chain in pharma became one that was ready to address any challenge, anywhere.”

Jacques Lambert, Senior Advisor, Life Sciences, Forvis Mazars Group

Clinical trials form a specialised part of the industry’s supply chain. Trials are costly, persisting for an average of eight years in the US alone, and can account for up to 70% of the cost of bringing new drugs to market. They require small-batch therapy production, specialised labelling, frequent use of cold-chain logistics and return pathways for unused drugs. Increasingly, AI is being used to select patients and run synthetic control groups. Patient attendance at a medical facility is no longer a necessity in all cases. Increasingly, home-based trials leverage remote sensors and telemedicine to reduce the need for travel. Depending on the circumstances, decentralised approaches like these may create a need for significant edge computing capacity, including reliable connectivity and the ability to secure patient data by maximising the amount of data processing conducted on site.

We recommend the following approaches to supply chain management system upgrades:

• Begin the drive to establish a single source of truth for supply chains by unifying isolated data silos across multiple systems. Digitise remaining paper-based workflows, clean datasets where necessary and impose data governance.
• Encourage adoption of cloud-based platforms that offer scalability, interoperability and real-time collaboration internally and externally with suppliers. Examine the potential of supply chain ‘control towers’ that enhance visibility and allow faster responses to supply chain disruption.
• Examine AI’s potential to improve demand planning by combining predictive analysis of epidemiology, sales data, demography and other external market trends.
• Shift to digital and machine-readable data for clinical trials. Examine AI’s potential to accelerate patient selection. Investigate the potential of remote monitoring, home care and telemedicine for decentralised clinical trials.

Managing real-time governance and regulatory lags

Deployed within autonomous AI agents, generative AI has the potential to transform compliance. Drawing on knowledge bases containing previous regulatory submissions, for example, AI has the potential to reduce re-submission requirements by identifying unsupported claims, tightening up language and anticipating regulators’ follow-up questions. It also seems likely that agentic AI will supplement the skills of compliance project managers, assessing timelines and the resource requirements imposed by impending submission deadlines.

Equally, AI is creating new risks that need to be regulated. Both the technology and accompanying regulations remain in flux, but the latter needs to crystallise before life sciences companies can invest significantly in solutions that promise major efficiency gains. In a highly globalised sector, the risks of regulatory divergence between regional economies looms large. The fear of retroactive compliance burdens is another significant concern. Questions also exist about the extent to which life sciences companies can legally protect intellectual property developed by AI. In addition, fears persist that a shortage of AI-literate talent within regulatory bodies may slow down approvals.

We recommend the following initial approaches to deploying AI to accelerate compliance workloads:

• Focus on the potential of agentic AI to automate documentation-heavy processes, including regulatory drafting, submission, compliance monitoring and audit readiness.
• Conduct detailed analysis of the data sources that will be accessed by AI systems. Ensure the integrity of data feeds. All associated infrastructure must be secure and compliant.
• Define when human oversight of AI systems is necessary and which human executives have responsibility for machine-made decision-making. Set limits to agentic AI’s access rights and decision-making powers. Map AI governance against evolving regulatory guidelines.
• Examine how human roles in the compliance team will change as agentic AI is introduced. The shift from manually based compliance to a situation in which human employees supervise agentic AI systems will require significant change management.

Preparing for what’s next: the road ahead for AI in life sciences

According to our C-suite barometer, relatively few senior executives working in the sector – just 22% – describe digital transformation as a top priority for their organisation over the next three to five years. Just one in three senior leaders in life sciences organisations with a digital transformation strategy believe that AI will have a major bearing on the success of that strategy. In other vertical sectors, more C-suite executives describe transformation as a top priority and highlight AI’s contribution to success.

In some cases, this apparent conservatism is the result of organisational complexity generated by a history of M&A-fuelled growth. As a result, life sciences companies are frequently complex and often rely upon siloed, legacy technologies that complicate transformation. Adding emerging technologies to this equation only increases perceptions of risk. Regulatory uncertainty relating to AI is another component in the complexity wall confronting the industry. For example, a recent research report by the Pistoia Alliance, a not-for-profit forum designed to encourage cross-sectoral innovation, notes weaknesses in data quality and governance, AI validation techniques, organisational models and value definitions.

Challenges like these are not unique to life sciences. As a minimum baseline, therefore, we recommend the following strategic initiatives that will enable AI transformation in the medium term:

• Investigate the options for improved data governance and modernising technology infrastructure that can cope with the demands of AI-ready data, computation, security and data privacy.
• Focus on technology upgrades required in manufacturing, including IT/OT integration and real-time data flows.
• Build a culture of data- and AI-literacy, with cross-functional teams working within a hub-and-spoke operating model that combines centralised resources with decentralised knowledge of business processes.
• Maintain engagement with regulatory authorities responsible for AI, helping to inform their policy choices and signalling challenges that require further clarity.

Failure to prepare for AI transformation today will only increase the size of the complexity wall confronting the sector tomorrow. Surviving uncertainty is no longer sufficient. Even in the presence of continuing disruptions, the industry’s goal must be to create the conditions to leverage transformation at scale in the near future.