Executive Summary: Elevate and De-Risk Your Drug Portfolio Strategy With AI

With the right data, the pharmaceutical industry can benefit from artificial intelligence (AI)-based tools in many ways, including decision-making for scientific and business challenges.

INTRODUCTION

In this Pharmaceutical Executive webcast sponsored by Intelligencia AI, industry experts* discussed leveraging AI to quantify better the risks associated with crucial decisions in drug development and the value of consistently applying the same set of objective criteria in decision-making. The panelists covered a range of topics, including the use of AI to assess and calibrate the probability of technical and regulatory success (PTRS).

The panelists included moderator John McGrath of Technological University Dublin,  who has over 25 years of experience consulting on project management for companies and government agencies; Panos Karelis of Intelligencia AI, an expert in management science who helps customers apply analytics to data; Ivan Kugener,  who supports EMD Serono’s decision-making for its pharmaceutical pipeline; and  Joshua Hattem of ZS, who assists pharmaceutical companies in developing strategies for product development and commercialization.

CURRENT STATUS OF AI IN PHARMA

To kick off the webcast, moderator John McGrath, who has worked with more than 150 organizations to improve their project portfolio management, pointed out the popularity of AI and noted that some industries need to turn this interest into action. While pharma has started to embrace AI, there are still untapped opportunities. 

The conversation began with each panelist sharing their AI journey. Kugener described AI as a tool to improve productivity and manage risk that is already being used to select sites for clinical trials and signals of drug safety.1 Hattem mentioned companies taking two different approaches to AI adoption; some companies start applying it to defined, bite-sized projects in a bottom-up approach, while others have a corporate mandate to apply it to decision-making. Karelis emphasized that AI is intended to “help  [the industry] become smarter and work smarter; AI technology is a conduit, but some companies still need to overcome the challenge of using this technology, which can be simplified by starting with the proper use case(s).” 

Overall, the panelists agreed that organizations must embrace  AI to augment decision-making or risk being left behind. In pharma, for instance, Kugener described the application of AI  as a race, stating, “If we don’t innovate, we disappear.” 

Several questions were posed to the audience to gauge their perspective on the topics the panelists discussed.  

The first poll question asked participants to describe their use of AI for de-risking their company’s drug portfolio. 

As shown in FIGURE 1, about 40% of the respondents indicated that they are actively using AI-based tools to de risk drug portfolio strategy or are planning on exploring the use of AI within the next 12 months. More than 50% of the respondents selected: “It’s very exploratory, and no substantial investment has been made yet.” Overall, this poll suggested that industry stakeholders must determine how to apply AI to enhance current risk assessment practices and applications. 

Hattem noted, “Predicting approval of a new product is ambiguous and requires gathering data from many sources not typically connected in an organization today. The first big step companies need to take is on the data side by synthesizing it all into one place. Those data sources aren’t even digitized in some companies. There are often PowerPoint decks that asset teams have put together for governance meetings. A lot of groundwork needs to be put in place.”  

BUILDING A SOLID DATA FOUNDATION TO MAXIMIZE AI POTENTIAL

Acknowledging that the AI journey is underway in the pharmaceutical industry, McGrath explored how companies built the foundations for using this technology. “We all know pharma organizations have been collecting data for at least the last 10 years,” he said. “We always hear about the power of big data,” and AI might be the tool to harness that data if companies can make data AI-ready. 

According to Hattem, quality is the key to AI-ready data. The data must be clean, indexed, and organized. “That’s the very costly and labor-intensive piece,” he noted.

As Karelis emphasized, a company must decide what data is needed to solve a particular problem. He stated, “It’s more than just data. It’s data plus. It’s really about the infrastructure that supports the data. It’s about harmonizing the data so that everything becomes part of a single database, a single source of truth. It comes down to the why. It comes down to defining the problem you want to tackle and then collecting the data in the right form. This is critical for feeding your machine-learning models. You need both data and structure. It’s about bringing that engineering mindset to the data.”

Hattem added that how a company invests in data is even more important than how it invests in AI. Instead of a company owning all of its data or relying exclusively on an  ecosystem of data partners, he said that, in the industry’s current state, “It has to be a balance—you have to own some yourself, but you can’t own it all.” 

Kugener explained that EMD Serono started with the belief that  the company could collect comprehensive data independently. However, doing that would have required several full-time employees to focus exclusively on this task. If they chose this path, EMD Serono would have had to put the development of  AI models on hold or significantly slow down the AI transition. Instead, the company explored working with industry partners, including Intelligencia AI. As Kugener emphasized, only a few existing industry partners can provide the service EMD Serono needs – the data, the AI tech and the expertise. Kugener encouraged webcast attendees to thoroughly vet and research potential external partners to find the right provider of advanced AI-based solutions. 

Kugener suggested a three-step approach to initiate the use of AI in a pharma company. “First, thoroughly research potential partners. Second, select a partner to adjust its services based on your specific needs. Third, communicate extensively inside the organization and try to find pockets of value where this service can be leveraged.” Finding the applications will require exploration and speaking with different divisions within the company, like clinical and regulatory affairs. 

While companies need to invest in data capture and management, companies like Intelligencia AI have already done much of the heavy lifting, simplifying AI integration into current pharma workflows. 

THE VALUE OF AI IN DECISION-MAKING FOR PHARMA

Decision-making in pharma is complex, with inflection points at various stages. Panos shared, “Strategy is a multifactorial, complex task. There is often the misconception that you can press a button and an AI model will spit out the correct answer. This just is not reality. There are always trade-offs when making decisions. And AI can help by feeding the decision-making, strategy and mitigating the risk-laden drug development process. Think of AI as a tool to augment and support and not a replacement for the expertise of individuals.”

The biopharma’s oncology program failed during the engagement period with Intelligenica AI. If they had access to the more accurate, AI-powered prediction, the company could have discontinued the program earlier, saving significant resources. Comparing the PTRS assessments, the low PTRS and the approval failure validated the accuracy and credibility of Intelligencia AI’s methodology.

In another poll, the audience was asked about their go/ no-go decisions in drug development. Attendees selected a range for the number of decisions that could have been better informed over the last two years; more than a quarter of the attendees indicated that more than six decisions could have been better informed regarding go/no-go drug portfolio decisions (FIGURE 2). 

A company needs to track performance to measure the impact of applying AI to drug development. According to Karelis, “Many companies [are not] taking the time to develop a structural approach to tracking and documenting the performance of their decision-making and their assessment.” Tracking outcomes is critical, as the lessons learned can be fed back into a predictive model to improve future risk assessment.  

Today, many companies need help making those portfolio management decisions. An audience poll revealed that more than half of the participants felt that subjectivity— human biases and lack of objective, data-driven processes—was the main challenge when conducting portfolio risk assessment (FIGURE 3).  

This result didn’t surprise Karelis, who stated that AI applications provide a more “sophisticated way to interact with data and [drive] science.” More objectivity is needed; access to the right combination of data, AI and technology as well as working with an external partner can make that a reality.  

AI-based tools can provide more clarity and confidence in such decisions, especially when evaluating two options. As Kugener  explained, the key is using AI to assess relative value, such as deciding between two options in an internal pipeline. “AI is very good at telling you A is better than B or vice versa,” he said. Additionally, he pointed out that even a small benefit from AI can be a good return on investment. “If you just move the needle by a fraction of a percent of the probability of success, you’ve paid for your service 10 times, 30 times, 50 times,  100 times.” The risk of not investing in AI is far greater than investing in it and learning. 

More objectivity could also improve PTRS. For more on PTRS, Kugener recommended an article by scientists at Novartis.3 The article highlights that people and companies tend to be overconfident and set probabilities too high, which applies in both experimental and business situations. “Using an AI-based approach to PTRS is not going to give you the answer, but it can suggest the direction of a better decision, and we have to be okay with that,” stated Kugener.

Hattem added, “When it comes to testing and measuring the impact of AI, companies need to strike while the iron is hot and not wait—you need to take a leap of faith; it may take some time for us as an industry to see the longer-term impact of the benefits of AI on the business of pharma. Five years form now, we will have seen some poor decisions associated with AI. It will cause some companies to fall off the map, and the persistent and most strategic ones will ultimately generate the most value.”

CONCLUSION AROUND AI, DATA AND DE-RISKING DRUG DEVELOPMENT

The pharmaceutical industry continues to evolve in leveraging AI-based methods. One important application for these advanced tools is de-risking decision-making in drug development. The panelists discussed the critical steps companies must take to implement AI-based decisions successfully. The key steps include selecting the proper use cases and then collecting, cleaning and harmonizing the correct data, ideally with an experienced external partner.  

REFERENCES

  1. Askin, S., Burkhalter, D., Calado, G., El Dakrouni, S. Artificial intelligence applied to clinical trials: opportunities and challenges. Health and  Technology. 13(2):203–213. (2023).
  2. GlobalData. Big data in the pharmaceutical industry: analyzing  innovation, investment and hiring trends. (2024). https://www. pharmaceutical-technology.com/data-insights/big-data-in-pharma 
  3. Hampson, L.V., Holzhauer, B., Bornkamp, B., et al. A new comprehensive approach to assess the probability of success of development programs before pivotal trials. Clinical Pharmacology &  Therapeutics. 111(5):1050–1060. (2022).