Five Key Steps to Successful AI in the Lab

Opportunity, When Done Right The efficiencies and insights of artificial intelligence (AI), machine learning (ML), predictive analytics, and other emerging technologies offer abundant opportunities to laboratories. A McKinsey report revealed that more than a quarter of companies with proactive AI strategies credit at least 5 percent of top line profits to AI.1 It’s not just the top line that benefits — rapid improvements can be seen in turnaround times, lab throughput, and labor costs as a result of successful AI implementation. Yet despite the benefits, most labs aren’t capitalizing on the potential of AI. Many have yet to embrace the technology at all, and those that have typically approach it in ways that are destined for failure. Although Gartner found nearly half of CIOs planned to implement AI, it also found that it could fail — leading to “erroneous outcomes” — as much as 85 percent of the time because of bias in data, algorithms, or the teams managing those.2 But the problem isn’t in the technology — it’s in the approach. To help our customers leverage digital transformation and prepare for the lab of the future, LabVantage Solutions has devised a five-step process for successfully implementing and profiting from AI in the lab and across the enterprise. We also offer an AI-readiness workshop to prepare you to leverage this technology. But First, What is AI? Accenture defines AI as “a constellation of many different technologies working together to enable machines to sense, comprehend, act, and learn with human-like levels of intelligence.”3 It reminds us that AI “remains an extension of human capabilities, not a replacement.” While many envision a robot with human-like tendencies, AI is realistically the solutions that leverage data and digital assets to make processes more intelligent. In a laboratory setting, that can mean using data from connected systems to make decisions, predict future outcomes, or reduce bottlenecks. ABOUT LABVANTAGE A recognized leader in enterprise laboratory software solutions, LabVantage Solutions helps you transform data into knowledge. We support more than 1500 global customer sites across lab-centered industries with the most technologically advanced informatics platform incorporating LIMS, ELN, LES, SDMS, and advanced analytics. LabVantage’s 40 years of experience and investments in AI/ML, mixed reality technology, semantic search, and more makes us the ideal partner for forward-thinking labs.Five Key Steps to Successful AI in the Lab | 2 What is the ‘Lab of the Future’? The ‘lab of the future’ has yet to be defined, though many people have many visions of it. Central to our future lab is digital connectivity, where all instrumentation is connected and the potential for a “digital twin” of the physical lab exists virtually — enabling predictive modeling, optimization, automation, and, going a step further, autonomation — defined as automation with intelligence that enables humans to interact or intercede with automated processes. Monitoring and visualization of lab activity will lead to increased control; from that emerges the ability to optimize lab performance and scientific research; then the opportunity to automate and eventually autonomize more lab functions. As labs become digitally empowered, they — and importantly, the scientists working in them — become central to solving our most pressing humanitarian challenges: climate change, unmet medical needs, food safety, and more. To do that, organizations must commit to technologies that support the lab of the future. Commitment Precedes Success True success with AI begins with a genuine commitment to the process. Many labs are attracted to the promises of new technology but underestimate the effort required to make these technologies pay off in the long run. Labs that view AI through a testing lens rather than a commitment lens are setting the technology up for failure. Sophistication with AI requires a development mentality, rather than a purely research mentality. Instead of ‘playing around’ with the technology, commit to it by answering these questions: • Can I define what success looks like? • Do I have a plan for achieving this success? • What resources do I need to pull this off? • How do I deploy this into production? • How do I integrate this technology into my business plans? • How do I streamline this technology into my business processes? • Who can I partner with to gain access to those resources? In effect, labs that ‘experiment’ with AI yield similar results to those who ignore the technology altogether. In fact, an experimentation approach can often lead to negative results in the form of sunk time and resources. DON’T FALL INTO THE ‘TEST’ TRAP Just as they have with other technologies in the past, many labs approach AI as a research project. We often hear phrases like: • “We tried it. It didn’t work.” • “We’re testing it out.” • “We thought we’d play around with it.” • “It’s in R&D at the moment.” • “We’re building a prototype.” A “test” mentality is not a commitment, which is necessary to win with AI.Five Key Steps to Successful AI in the Lab | 3 The Five Steps to Successful AI Transformation 1. Start With a Use Case The first key to commitment is selecting a clear use case. Unlike simply ‘playing around,’ developing a use case sets a clear purpose for your AI implementation and assigns an associated ROI and business outcome. Developing a use case allows you to budget accordingly, develop an action plan, and forecast a roadmap. AI adopters with a proactive strategy achieve significantly higher profit margins than those who simply experiment with the technology. DESIGNING THE IDEAL USE CASE There is no universal use case for working with AI; the best case for your lab will be highly relevant to the operation you are running and reflect the business outcomes that matter to your lab. That said, a common goal for labs starting out with AI is to leverage functions that would provide the highest yield for the least amount of effort or risk. There are several functions that fit that description including lab performance analysis, integrated modeling, and predictive formulas. Each of these areas provide a rich value to the organization. A rigorous AI-driven performance analysis allows you to look through the complexities of your lab operations and quickly identify the hotspots causing problems, whether it’s problems with quality, turnaround time, or overall performance. Integrated modeling allows you to do statistical modeling like calibration curves, immunogenicity, stability, and others without moving data back and forth while losing valuable information about the process. Finally, leveraging AI to derive formulas from existing data will cut the number of physical studies required to a tenth or less. Example use cases of these include: • Instrument Data Analysis Setting up a real-time data ingestion pipeline for laboratory instruments for downstream data analysis and predictive maintenance of instruments • Lab Resource Scheduling Managing better utilization of lab resources (raw material, equipment, and manpower) through operation research modeling DRAMATICALLY REDUCE PHYSICAL STUDIES Leveraging AI to derive formulas from existing data will cut the number of physical studies required to a tenth or less. 2 DATA ECOSYSTEMS 3 TECHNIQUES AND TOOLS 4 WORKFLOW INTEGRATION 5 OPEN ORG CULTURE 1 USE CASES / SOURCES OF VALUE • Scan use-case horizon • Articulate business needs and create business case • Break down data silos • Decide on the level of aggregation and pre-analysis • Identify high-value data • Identify fit-for- purpose AI tools • Partner, or acquire to plug capability gaps • Take an agile, “test and learn” approach • Integrate AI into workplace processes • Optimize the human/ machine interface • Adopt an open, collaborative culture • Build trust in AI insights • Reskill the workforce to ensure complementarityFive Key Steps to Successful AI in the Lab | 4 • Quality Management Statistical process control and quality-related analytics to identify drivers of poor quality and recommend real-time intervention strategies • PK-PD Modeling Accelerating pharmacokinetic and toxicology studies through statistical tools and machine learning models that enable researchers to perform sophisticated analyses • Immunogenicity Analyses Facilitating immunogenicity cut point analyses and calculations using parametric and non-parametric approaches through a set of OOB models • Formulation Studies Leverage AI-based algorithms on existing data to predict a recipe that uses specific raw materials and meets the desired specification 2. Get a Handle on Your Data Data still presents a significant roadblock for many labs. Successful AI implementation requires a commitment to sourcing the right data and transforming it into useful, readable formats. Most labs see their data problems as technology problems, but data problems often begin higher up the chain. They tend to be problems of corporate vision, not of lab technology. Better data starts with better data stewardship and data design. It’s important to choose one person who will be responsible for resolving data-related friction, and who can help to design a data ontology that represents the best case for the organization. Many labs have yet to bring their LIMS, ELNs, and other digital assets together with their financial and production systems; your data needs to flow through a network that reflects the organization inherent in your business systems. Understanding these connections and building an ecosystem to facilitate data flows is crucial to success with AI. Companies need a clear idea of what’s being measured and the parameters defining it. Most of all, they need a firm grasp on the metrics that matter. The technology required to manage data already exists; what’s missing is the business knowledge and the structure to apply the data in useful ways. We encourage clients to start with a project leader and work through the organization to build a ‘digital twin’ of their laboratory — recording everything digitally to enable monitoring. With clean, up-to-date, and prepared data coming in, labs can quickly move to the next step. 3. Assess Techniques and Technologies The next step in successful AI implementation involves identifying fit-for-purpose AI tools and partnering with the right talent to plug existing capability gaps. Navigating the world of artificial intelligence alone can become a slippery slope to project failure. The technology landscape changes rapidly and staying ahead of the latest tools and technologies can be difficult and time-consuming. Total commitment to AI requires finding and working with a knowledgeable partner who can guide you through the technology, the models, the methodology, and the language of your project. We recommend looking for a partner with deep technical experience in AI and analytics, combined with broad knowledge of your specific industry. AI service providers are appearing seemingly overnight, but a partner with knowledge of the unique instruments, processes, BUIID A DIGITAL TWIN Organizations should build a ‘digital twin’ of the laboratory — recording everything digitally to enable monitoring. With clean, up-to-date, and prepared data coming in, labs can quickly move to the next step.and use cases in your industry can mean the difference between success and failure with this technology. When searching for a partner, make sure they demonstrate the industry experience to be able to connect technical implementation with your specific business objectives. Additionally, it’s important to ensure that your data quality, privacy, and security protocols are adequate. Once again, a trusted partner can make all the difference. 4. Integrate AI into Workflow The best results from AI come from first identifying the points within your lab’s workflow where AI can add the most value. This might even require changes at the organizational level and involve some degree of training and up-skilling. From there, look for ways to incorporate AI models into your workflow by integrating them into existing software and tools, or developing new, complementary interfaces. It’s also important to optimize the human/machine interface. Look for platforms that offer userfriendly, intuitive interfaces and dashboards that make it easy for your team to execute tasks with the help of AI. The use of hardware tools such as mixed reality, robotics, and digital assistants can help to bridge the human/machine gap and assist in many lab functions, including training, onboarding, manufacturing, and more. As you continue to work with AI, we recommend monitoring the effectiveness of your digital workflows at regular intervals. This way, you will continually identify ways to optimize and reconfigure for best results. 5. Adapt a Culture for AI Sometimes the biggest challenge is not that of technology, but of human nature. Today’s cultural resistance to AI is much like the resistance seen with other technologies (telephones, computers, the internet) introduced in the past that are now an integral part of our daily lives. In many ways, these technologies enter our lives like waves; those who willingly ride the technology wave take it safely to shore, whereas those who resist will be tumbled and spun to the ocean floor. AI has the potential to replace many of the tasks currently performed by humans in laboratory settings, leading many to fear that they will be ‘replaced’ by the technology. We encourage our clients to adopt an open, collaborative culture that educates staff on the opportunities of this new technology and reassures them of their continued value in the workplace. Labs will see greater impact from AI when they focus on up-skilling their teams, ensuring a complementary relationship between computer work and human workers. Where AI might now be able to take care of some of your team’s more manual and laborious processes, your team can now focus on higher-level, more strategic work. As with most new technologies, it can be difficult for your team to fully accept and trust the insights generated by AI technology. This can take time and education, but we believe the benefits of AI eventually become obvious to all. Five Key Steps to Successful AI in the Lab | 5 TO SUCCEED, BUILD A COLLABORATIVE CULTURE Adopt an open, collaborative culture that educates staff on the opportunities of this new technology and reassures them of their continued value in the workplace. Labs will see greater impact from AI when they focus on up-skilling their teams, ensuring a complementary relationship between computer work and human workers. FOCUS ON HIGHER-LEVEL WORK Where AI might now be able to take care of some of your team’s more manual and laborious processes, your team can now focus on higher-level, more strategic work.ABOUT LABVANTAGE SOLUTIONS A recognized leader in enterprise laboratory software solutions, LabVantage Solutions dedicates itself to improving customer outcomes by transforming data into knowledge. The LabVantage informatics platform is highly configurable, integrated across a common architecture, and 100% browser-based to support hundreds of concurrent users. Deployed on-premise, via the cloud, or SaaS, it seamlessly interfaces with instruments and other enterprise systems – enabling true digital transformation. The platform consists of the most modern laboratory information management system (LIMS) available, integrated electronic laboratory notebook (ELN), laboratory execution system (LES), scientific data management system (SDMS), and our advanced analytics solution (LabVantage Analytics); and for healthcare settings, a laboratory information system (LIS). We support more than 1500 global customer sites in the life sciences, pharmaceutical, medical device, biobank, food & beverage, consumer packaged goods, oil & gas, genetics/diagnostics, and healthcare industries. Headquartered in Somerset, NJ., with global offices, LabVantage has, for four decades, offered its comprehensive portfolio of products and services to enable customers to innovate faster in the R&D cycle, improve manufactured product quality, achieve accurate record-keeping, and comply with reg