Pharmatizing AI

Scott Snyder
Chief digital officer
EVERSANA

There is no shortage of enormous predictions around the potential impact of AI since the recent uptake of Generative AI (GenAI) with its array of mesmerizing, human-like, and easily accessible capabilities. Top analysts and consultants have thrown out numbers like $4 trillion of economic impact and 40 percent of all work activities having the potential to be automated by GenAI solutions. In the pharmaceutical industry alone, the estimated benefit could be between $60 and one hundred billion annually.

Yet, while 79 percent of executives expect GenAI to drive significant transformation in their organization and industry, less than 40 percent of leaders believe they have the right technology and data infrastructure, strategy, governance, and talent in place to fully capture the benefits of GenAI. Seventy-one percent of companies are experimenting with GenAI, but few of these use cases seem to be translating into commercial impact.A recent survey of 300 global business leaders by MIT Technology Review and Telstra revealed only 9 of the surveyed companies are significantly using AI.

So how can companies in a highly regulated industry like pharmaceuticals overcome this type of adoption challenges with all of the fanfare and uncertainty surrounding GenAI?“Pharmatizing” AI means balancing the superpowers offered by the technology such as content creation, summarization, knowledge-assist, and self-service with the real risks that need to be managed such as hallucinations, model bias and drift, and privacy and security.

Using the 3 Rs to guide AI innovation

To help organizations and leaders make informed decisions on where and how to invest in their own AI transformation, I would like to use 3 Rs as a framework: Responsibility, Reliability, and Return on Investment (ROI).

First, let’s talk about responsibility.It is the cornerstone of ethical AI deployment and every organization should have a responsible AI development policy in place to provide the guardrails to protect both the organization, its employees, and customers/patients.

Responsible AI is built using algorithms that are transparent and allow users to understand how decisions are made. The outputs must be explainable to stakeholders – especially in a regulated industry and trace the reasoning behind AI-generated outcomes.

Responsible AI must also actively combat bias, ensuring fairness across diverse user groups. To ensure this is accurate, regular audits and bias assessments are essential to maintain ethical standards.

And finally, privacy and data security are paramount. AI models, and especially in the pharmaceutical industry, must adhere to privacy regulations such as HIPAA. Critical data and patient information cannot be compromised.

The second R of AI readiness is reliability.AI solutions by their nature are trained on available data and will have prediction errors and even hallucinate as in the case of Large Language Models. This lack of reliability can directly impact trust, adoption and effectiveness. To address the reliability challenge, organizations must first select use cases where AI’s reliability matches the required accuracy. For a use case like content generation to speed up the creative process, the error tolerance may be high whereas a use cases for patient self-service may demand a highly predictable and controlled set of outputs for experience, compliance and safety reasons. In addition, organizations need to build trust in AI systems through rigorous testing and “red teaming” along with on-going monitoring to ensure models perform consistently across various scenarios before deploying against a specific use case.

Finally, the third R of AI is return on investment or ROI.

Yes, AI can crank out many variants of an image or video in seconds or write an article or proposal one hundred times faster than a person, pulling from data sources across public and/or private platforms. But if an organization invests in AI, how can they effectively estimate and measure the overall impact? And how can they use this to allocate investment across competing AI opportunities? Is it saving hours and costs, creating new revenue streams, improving speed, or reducing the risk of human errors?

Each organization should develop its own AI ROI model. This model should include both direct and indirect benefits such as cost savings, revenue growth, customer and employee satisfaction, speed to launch/market, societal impact and accuracy. The model should include direct and indirect costs such as solution design and development, technology and data infrastructure, model licensing and API costs, machine learning operations, support, change management, training, ethics and legal support, among other variables. The model must be adapted for different use cases and scenarios and may need to rely on estimates from similar use cases in other companies and industries where there is limited experience inside the company. These ROI models will help leaders make the best decisions across their portfolio of AI opportunities by connecting investments to tangible results.

The new AI Wave is here–are you ready?

Using the 3Rs to guide your AI opportunities will ensure that you take practical steps that advance potential innovations while managing risk and return. Regulated industries like Pharma demand even greater focus on pursuing lower risk use cases that can drive near-term impact, while planting the seeds for higher risk, higher payoff cases in the mid and longer term. Pharma companies that can navigate the 3R’s and unlock the full potential of AI in their organizations will have the opportunity to leap ahead in transforming their operating model as well as patient experiences and outcomes.

Scott Snyder is EVERSANA’s Chief Digital Officer, driving digital transformation for employees, clients, and the patients. Contact him at scott.snyder@eversana.com.