• Strategy
  • Payers
  • AI
  • Cybersecurity
  • Device Development
  • Studies/Published Research
  • Ethics
  • Regulatory
  • Digital Biomarkers
  • Data Science
  • Companion Diagnostics
  • Clinical Care
  • Digital Therapeutics
  • Academic Research
  • Virtual Reality
  • RWE
  • Digital Endpoints
  • Medical Devices (Class II and Class III)
  • Commercialization
  • Clinical Trials
  • Drug-Device Combinations
  • SaMD
  • Remote Patient Monitoring
  • Patient Engagement
  • Profiles/Q&A

Changing the Paradigm in AI Implementation


How pharma companies can reduce the risk of failure with AI-based innovations.

Policastro Marcus

Jeff Policastro and Alan Marcus
LabVantage Solutions

AI analytics are fast becoming essential in laboratories that want to optimize processes and make better decisions. A McKinsey report1 revealed that over a quarter of companies with proactive AI strategies credit at least 5% of top-line profits to artificial intelligence (AI). On the other hand, Gartner found that implementing AI could fail—leading to “erroneous outcomes”—as much as 85% of the time because of bias in data, algorithms, or the teams managing them.2

How can labs get AI right? How do they reduce the risk of failure and access the deep domain expertise needed to get results from their data?

The problem isn’t technology—it’s the approach. To reap the rewards of AI analytics, every lab must define its own unique pathway from its current state to the digitally advanced, data-driven ideal. This article outlines five steps for the successful implementation of AI analytics.

What it takes to succeed with AI

Successful AI implementation requires commitment. Companies that approach AI with an idle attitude of “testing it” or “playing with it” are not likely to see demonstrable success. We recommend asking hard questions about your AI plans:

  • How do we define success?
  • How will I get there?
  • Which talent and what technology will get me there?
  • How much will this cost, and what will be the return?

Working with an industry-experienced technology partner can help you answer these questions and more, starting your AI journey off on the right foot. Once you are committed to implementing AI with a business mindset, there are five key steps we recommend following.

Give AI a practical application

Successful AI implementation begins with an appropriate use case. Unlike mere experimentation, identifying a specific use case establishes a clear intention for your AI integration and links it to measurable returns on investment (ROI) and business results. Define your objectives and secure stakeholder buy-in by developing a detailed plan, including timelines, milestones, and dependencies.

This process of defining a use case provides the structure needed for appropriate financial planning, the creation of actionable steps, and the ability to map out a forward-looking path.

How to design the ultimate AI use case

AI doesn't have a one-size-fits-all application; the most suitable scenario for your laboratory will heavily depend on the specific tasks you're conducting and the business goals you're pursuing.

However, a common objective for labs just beginning to incorporate AI is to maximize output while minimizing effort and risk. Within this context, there are several functions that will further this objective, including:

  • Assessing lab performance. AI-driven performance analysis allows you to sift through the intricacies of your lab's operations swiftly, pinpointing problematic areas such as quality issues, time delays, or overall efficiency bottlenecks.
  • Integrating statistical models. Integrated modeling enables you to conduct statistical analyses, such as calibration curves, immunogenicity assessments, and stability evaluations, without the need for constant data transfer, preserving crucial process insights.
  • Predictive formulas. Harnessing AI to extract formulas from existing data substantially reduces the number of physical studies required, potentially cutting them down to a fraction of the original count.

In this vein, there are a number of practical use cases a lab might begin with.

Pharmacokinetics (PK) analysis

This supports drug development, individualized treatment planning, and research on the factors influencing drug metabolism, ultimately contributing to the improvement of healthcare and the development of safe and effective drugs. PK analysis could take the form of:

  • Bioequivalence studies. For generic drug manufacturers, a PK modeler can be used to demonstrate the bioequivalence of their products to brand-name drugs, ensuring that generic medications are as effective as their counterparts.
  • Clinical trial optimization. During clinical trials, the solution can assist in dose selection, helping to strike the right balance between therapeutic efficacy and safety. It can also aid in designing adaptive trials that adjust dosing regimens based on ongoing data analysis.

Pharmacodynamics (PD) analysis

This allows researchers to quantitatively analyze the relationship between drug concentration and its effects, leading to more effective drug design, optimized dosing regimens, and better-informed decisions throughout the drug development process. PD analysis could take the form of:

  • Drug dose optimization. By selecting the appropriate model type, researchers can determine the most effective dose that achieves the desired therapeutic effect while minimizing side effects.
  • Pharmacological mechanism exploration. Researchers can use different models (e.g., linear, Emax, sigmoidal) to explore the underlying pharmacological mechanisms of a drug's action.
  • Clinical trial design. Researchers can use modeling to estimate the sample sizes needed, choose appropriate endpoints for efficacy evaluation, and design dosing schedules that maximize the chances of detecting a therapeutic effect.

Immunogenicity analyses

In this analysis, automatic cut points assist in enhancing precision and optimizing assay performance. Immunogenicity analyses could help to:

  • Reduce fake results. Comprehensive analysis, including multiple cut points and sensitivity calculations, reduces the chances of false positives or negatives in immunogenicity testing. This is vital for making informed decisions about drug development and patient care.
  • Optimize data and streamline workflows: The module's integration into lab workflows simplifies and automates the cut point calculation process. This increases efficiency, reduces human error, and speeds up the overall testing process.

Lab performance analysis

AI can measure and enhance the efficiency, quality, and financial sustainability of a clinical testing lab. For example:

  • Pre-defined performance dashboards enable lab managers and analysts to quickly access critical metrics and key performance indicators (KPIs).
  • Analysis of on-time completion of testing requests can identify which tests are consistently delayed and why, helping the lab improve its workflow and meet patient expectations.
  • Analyzing pending requests and average delay statistics can help in resource allocation, ensuring that critical tests are prioritized and completed promptly.
  • Tracking on-time deliveries and providing insights into billed and unbilled costs can improve the testing and reporting process while also assisting in financial management and identifying areas where cost-saving measures can be implemented.

Ultimately, the best use case is the one that solves a pressing pain point—one that, with the right results, can demonstrate to leadership and the rest of the team that AI can be of practice use in lab settings.

Establish robust data systems

Data challenges can hamper even the most promising AI projects. Data is the most important element of any new AI system but is simultaneously the biggest challenge for pharmaceutical companies, many of which maintain data logs that go back more than a century.

The purity and organization of your input data determines the quality of your results. High-quality data is data that has been thoroughly cleaned, formatted, and prepared for analysis.

Most data challenges are not technology problems. Rather, they originate further up the organizational hierarchy and stem from broader corporate vision issues. This means better data begins with improved data management practices and strategic data design. Labs must have a clear understanding of the metrics they’re tracking and the parameters that define them.

A substantial number of labs have not yet integrated their library information management systems (LIMS), electronic lab notebooks (ELNs), and other digital resources with their financial and production systems. Grasping these interconnections and constructing an informatics ecosystem that facilitates seamless data movement is vital for AI success.

The technology required to handle data effectively already exists; what's lacking is the business acumen and the framework to utilize the data effectively. We recommend that clients initiate the process with a designated project leader and collaborate across the organization to construct a “digital twin” of their laboratory—a comprehensive digital record of all activities for ongoing monitoring. Laboratories can rapidly progress through their AI journey by ensuring that data is clean, current, and well-prepared.

Find the most appropriate tools and technologies

With data protocols in place, look for the knowledge and skills gaps that need to be filled—most likely through a mix of technology and human expertise. Most labs won’t have the in-house skills required to implement AI from scratch and guide it through to success, so it’s important to seek out a knowledgeable partner at the start of your AI journey.

The right partner can help you understand the technology, models, methodology, and language your project requires. In fact, your use case’s specific needs and requirements will lead you to fit-for-purpose AI tools and techniques, such as statistical analysis, machine learning, or data visualization. The AI landscape is constantly evolving. A partner can help you keep up with the latest tools and technologies while ensuring that your data quality, privacy, and security measures are up to standard.

Look for a partner at the intersection of AI and your industry. An organization familiar with your industry's distinctive tools, processes, and applications can help you bridge the gap between technical implementation and your specific business objectives—it can mean the difference between success and failure with AI.

Embed AI into existing ways of working

To acquire the best results from AI, you must first identify the points within your lab’s workflow where AI can add the most value. This is where the strength of your commitment will get tested as it may require organizational level changes and a certain degree of training and up-skilling.

From there, examine how AI models can be incorporated into your processes and workflow, and start integrating them into existing software and tools, or developing new, complementary interfaces. Keep in mind that optimizing the human/machine interface will go a long way. Seek out platforms that offer user-friendly, intuitive interfaces and dashboards to simplify how your team executes tasks with the help of AI.

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 on your AI-assisted journey, we recommend monitoring the effectiveness of your digital workflows at regular intervals. This will enable you to continuously optimize and reconfigure for increasingly better results.

Build an AI-positive culture

Often the biggest obstacle for AI adoption in the lab comes down not to data or workflows but rather to the humans on your team. AI has become dinner table conversation for many knowledge workers, who harbor genuine fears of being “replaced” by this technology or significantly altering their roles and daily work.

It's important to involve your team. Clearly articulate the objectives you aim to achieve through AI and provide reassurance that their positions are secure. Leaders should contextualize AI implementation within a broader framework of change management, considering the team's concerns and perspectives. Collaborate with your team to cultivate the skills required to work in tandem with AI. Encourage your team members to shift toward higher-level, more strategic work, and look for ways to let AI handle routine tasks.

AI rewards the early adopters

Despite AI’s remarkable autonomy, true success with this technology requires clear planning, well-defined use cases, and effective leadership and guidance. Once you’ve seen initial success with a first use case, you can expand your use of this technology, applying it to a wider array of use cases and integrating it even further into daily lab operations.

AI is likely to become ubiquitous in our daily working lives, but the labs that pioneer its implementation will claim the competitive advantage. Taking the right steps now will fast-track your desired outcome, keeping you ahead of the competition for years.

Alan Marcus is Chief Growth Officer and Jeff Policastrois Director, SaaS Business Development; Both with LabVantage Solutions.


  1. Chui, M.; Hall, B.; Singla, A.; Sukharesky, A. The State of AI in 2021. McKinsey Global Survey. December 8, 2021. https://www.mckinsey.com/capabilities/quantumblack/our-insights/global-survey-the-state-of-ai-in-2021
  2. Gartner Says Nearly Half of CIOs Are Planning to Deploy Artificial Intelligence.Gartner. February 13, 2018. https://www.gartner.com/en/newsroom/press-releases/2018-02-13-gartner-says-nearly-half-of-cios-are-planning-to-deploy-artificial-intelligence
© 2024 MJH Life Sciences

All rights reserved.