Gen AI has become a hot topic among leaders at most big companies, more than two-thirds of which intend to increase their investments in the technology over the next three years. McKinsey has estimated the opportunity for gen AI in biopharmaceutical operations at $4 billion to $7 billion annually through workload and cost reductions, productivity gains, improvements to equipment effectiveness, and quality enhancements. Many biopharma organizations now have sophisticated operations and are moving to use gen AI to take advantage of their data repositories. However, only a few organizations have started to realize value from it. This raises a question: Are organizations deploying gen AI in the places with the highest potential benefits?

Over the past five years, biopharma has adopted various digital and analytics solutions, including in silico models, process optimizers, and lab and manufacturing automation. This article explores how gen AI, which differs from traditional AI and other analytics, can help biopharma solve industry-specific challenges.

There are at least a dozen proven and potential gen AI use cases along the biopharma operations value chain. They fall into three main categories:

• entry-level use cases that biopharma organizations and vendors are already deploying using off-the-shelf products
• novel use cases that incorporate product, technology, organization, and domain-specific intelligence and require custom, in-house development
• frontier use cases that typically require fast processing of large amounts of real-time numerical data and affect elements of operations that require tight quality and regulatory oversight (Exhibit 1)

Vendors continually implement and improve entry-level use cases in their products. For instance, enterprise-resource-planning vendors are employing gen AI to improve such functions as demand forecasting, inventory management, and transportation planning. Many smaller companies are also developing gen-AI-supported point solutions for those entry-level use cases.

At times, companies need to customize the solution to fit their needs and workflows better, but they generally do so in partnership with the relevant vendors. So to take advantage of the existing technology, biopharma companies need only buy or upgrade software as it becomes available.

At the other end of the spectrum are frontier use cases, which are too risky today for most companies to want to explore. In this article, we focus on four novel use cases that require bespoke development but are within the current capabilities of gen AI. Agentic AI use cases, which incorporate a layer of autonomous execution, build upon other AI applications and fall into this category as well.

Novel use case examples

As the capabilities of gen AI evolve, the number of viable use cases is likely to increase. We discuss four ways that biopharma organizations can currently benefit from novel gen AI use cases.

Use case one: Boosting shop floor efficiency with gen-AI-assisted supervision

In our experience, supervisors spend up to 40 percent of their time manually producing reports, collating information, triaging equipment failures, and supporting maintenance. To communicate with their teams, they also spend time preparing and retrieving production data, creating visualizations, preparing for meetings, and crafting update emails.

A gen AI supervisor can support a human supervisor in all these activities (Exhibit 2):

• Provide technical assistance. The use of gen AI supervisors can save teams considerable time and effort by giving access to and synthesizing information from various sources, such as machine history, technical manuals, and production data. They can analyze machine data to diagnose equipment issues and identify potential solutions quickly.
• Automate shift preparation. Gen AI supervisors can provide immediate access to data such as batch records, equipment sensors, and in-process measurements. They can identify bottlenecks, aggregate performance data from previous shifts, and proactively communicate critical inputs for the upcoming shift. This safeguards continuity in production and improves overall efficiency.
• Enhance team leadership. By automating the creation of shift reports, presentations, and emails, gen AI supervisors can facilitate communication with team members and stakeholders. This allows supervisors more time for mentoring, coaching, developing skills, and fostering a more engaged and productive workforce.

After implementing a gen AI copilot tool for maintenance support, one biopharma manufacturing team achieved 5 percent reductions in breakdown time, speed losses, and minor stoppages and a 30 percent reduction in execution time. It also experienced a 40 to 50 percent workload reduction for corrective maintenance.

A gen AI technical assistant typically reduces the time spent identifying and synthesizing a technical solution by 20 to 40 percent. Gen AI production and huddle assistants can reduce the time spent aggregating shift metrics and preparing communications by 40 percent.

Use case two: Streamlining production maintenance with smart deviation management

The end-to-end process for managing both deviation and corrective and preventive actions (CAPAs) requires 4 to 6 percent of a manufacturing site’s resources and is fraught with challenges. Common pain points include delayed detection, manual tasks, a low rate of right-first-time solutions, low effectiveness, inconsistent documentation, and a reactive process.

Our conversations with pharma companies highlight a unique value proposition for such a tool. We found that 65 percent of drug shortages are caused by issues related to deviation management and that 15 to 20 percent of deviations recur because of ineffective remediation. A gen AI tool can help an investigator manage deviations, providing proactive support and insights throughout the process (Exhibits 3 and 4):

• Identify similar deviations. Gen AI can analyze historical deviations with similar characteristics and provide context and potential solutions.
• Accelerate root cause analysis. By automatically summarizing potential root causes based on similar deviations, gen AI can allow investigators to focus their efforts and quickly identify the source of the problem.
• Suggest effective CAPAs. Gen AI can leverage historical records to recommend proven CAPAs tailored to the specific deviation. This can improve the effectiveness of remediation and preventive efforts.
• Automate documentation. By autopopulating reports with relevant information, gen AI can streamline documentation and provide consistency and compliance with quality standards.

A gen AI tool for one life sciences manufacturing company could synthesize 70 percent of deviations and connect them to similar events. This allowed for easy investigation and hypothesis generation. The same tool also generated a first draft of CAPAs for more than 80 percent of cases.

This approach typically results in 30 to 40 percent fewer deviations through improved prevention, with greater reductions in recurring and critical deviations. Case studies also show a 40 percent reduction in deviation closure time and 10 to 30 percent fewer quality- and expiry-related write-offs through reduced mechanical degradation.

Use case three: Accelerating product development with product process intelligence

The race to bring new therapies to market depends on the speed and efficiency of process design, process development, and technology transfers. The data required to take a product through these steps are often fragmented among various systems and functions within an organization. Gathering and distilling input can take time and delay decisions on unit-operation design, parameter optimization, and scale-up. Such delays can also lead to quality issues and high costs of goods and development.

A gen-AI-powered tool can act as a centralized hub for product and process knowledge. It can seamlessly integrate and analyze historical data from R&D labs, pilot plants, and commercial manufacturing sites, providing scientists and engineers with valuable insights to accelerate development:

• Leverage prior designs. By identifying successful unit operation designs and configurations for different molecules, gen AI can aid in raw material selection and early-stage parameter optimization.
• Optimize parameters. Gen AI can refine critical process parameters such as temperature, pH, and raw material variability to safeguard robust quality and cost-effective performance at scale. It can accelerate trial design and execution by generating draft protocols from historical trial data and automating document creation.
• Streamline experiment design. By capturing and organizing knowledge from experienced process engineers, gen AI can make it readily available to others. It can automatically generate process documentation, including flow diagrams, operating procedures, and batch records, to save time and reduce errors.
• Guide technology transfer. Gen AI can facilitate smooth and efficient technology transfers among facilities by identifying potential risks, generating training materials for new technology transfer staff, and answering questions from researchers and technology transfer professionals via chatbots.

Such a gen AI tool can reduce the integrated cost of development by reducing lab space, the number of experiments, and the quantity of materials needed to design and optimize processes. It can also improve the cost of goods and the robustness of commercial processes by optimizing parameters and enabling prompt troubleshooting. Last, it can permit quicker development and launch cycles.

We have seen similar but more traditional AI and machine learning use cases reduce the timeline to investigational new drugs by nearly one-third. Additionally, we have seen them increase development efficiency by 40 percent.

Use case four: Optimizing supply chain performance with a copilot

The data required to make accurate supply chain decisions are often fragmented among supplier databases and production systems. This can lead to limited visibility of stock levels, lead times, demand forecasts, and delivery performance. Such limitations can result in stockouts, production delays, overstocking, and operational disruptions, all of which affect the availability of products and the cost of goods.

An integrated gen AI tool can consolidate supply chain data, demand data, performance targets, and production data from multiple sources into a unified platform. This can provide a comprehensive view of inventory and performance, from raw materials to finished products. The tool can therefore facilitate informed decisions about stock levels, procurement, logistics, and production (Exhibit 5):

• Improve decision-making. By providing planners with insights and what-if-scenario analyses, gen AI can empower them to make informed decisions that improve efficiency and responsiveness.
• Optimize inventory management. Gen AI can help reduce stockouts and excess inventory by accurately forecasting demand, predicting potential bottlenecks, creating data views, and analyzing scenarios to help optimize inventory levels.
• Increase productivity. Using gen AI to automate routine tasks and provide easy access to critical information can free planners to focus on strategic initiatives and higher-value activities.
• Mitigate risk. Gen AI can analyze data from the supply chain and on markets, the weather, and geopolitical events to identify potential disruptions and allow for timely mitigation.

By consolidating, analyzing, and providing insights on fragmented data, such a gen AI tool can help biopharma companies double the productivity of their supply chain organizations, improve product availability, and reduce the overall cost of goods. We have seen such use cases create a 2 to 3 percent decline in supply chain costs, a 15 percent increase in forecast accuracy, and a 20 to 30 percent workload reduction for planners.

Implementing gen AI: Building the foundation for success

Biopharma companies, with their vast proprietary data sets for products, facilities, and technologies, are in an optimal position to capitalize on gen AI. As the technology becomes integral to the industry, those that adopt a value-focused approach will begin to drive transformational change. For a successful gen AI transformation, organizations must know the limits and risks of the technology before deploying use cases so they can prepare the right foundations.

The challenges to implementing gen AI in the biopharma industry include the risk of errors, commonly known as hallucinations, particularly when dealing with numerical data. Human oversight and verification are essential to address this. Companies should deploy gen AI only sparingly in complex tasks, such as batch release and forecasting, for which numerical accuracy is critical.

Additionally, gen AI isn’t suitable for high-volume, time-sensitive tasks, such as real-time monitoring and control of the supply chain and production. Moreover, successful gen AI adoption requires extensive training for engineers, operators, and technicians to ensure that they can spot errors, prompt appropriate queries, and make informed decisions.

Gen AI initiatives are also vulnerable to risks, such as intellectual property (IP) theft and algorithmic bias. For instance, a gen-AI-created design might infringe on another company’s IP, or the company’s own IP might leak into the public domain. Also, a gen AI tool that is trained on biased historical data may exhibit demographic bias when assigning tasks.

Finally, the biopharma industry’s highly regulated environment demands rigorous risk assessment and robust guardrails to secure compliance when using gen AI for tasks such as batch record reviews, quality audits, and batch releases. Companies should enforce the appropriate guardrails and policies to ensure that they’re using the tools for the intended and validated tasks.

To overcome these constraints and transform sustainably at scale requires foundational capabilities in six areas (Exhibit 6). Rewired: The McKinsey Guide to Outcompeting in the Age of Digital and AI (Wiley, 2023) describes these areas in more detail. With such capabilities in place, it becomes easier to accelerate decision-making, transform an operating model, and migrate talent and resources to the highest-potential opportunity areas.

The biopharma industry stands on the cusp of a gen AI revolution. By strategically selecting use cases, building foundational capabilities, and fostering a culture of innovation, companies can harness the power of gen AI to accelerate drug development, optimize operations, and ultimately improve patient outcomes. While navigating the complexities and risks associated with gen AI is crucial, the potential rewards for organizations that embrace this transformative technology are immense. The future of biopharma operations is intelligent, automated, and driven by the power of gen AI.

The post "Gen AI: A game changer for biopharma operations" appeared first on McKinsey Insights