Author: Panos Karelis, VP of Commercial, Intelligencia AI
In a recent AI in Business Podcast episode, I discussed the key challenges and opportunities in AI in drug development, an area where innovation is rapidly transforming pharma.
During my conversation, we explored several critical aspects of integrating AI into drug development processes, from overcoming adoption challenges to maximizing AI’s potential in pharma.
Navigating the Challenges of AI Adoption in Drug Development
AI in drug development is still in its early stages. While its potential is undeniable, adoption comes with many hurdles. One major challenge is the “black box” nature of many AI models: without transparency, stakeholders struggle to trust AI-driven insights. Building trust requires AI systems that not only deliver outcomes but also offer explainability and validation. From my experience, companies that successfully adopt AI invest in such systems, ones that provide clear reasoning behind their predictions.
Another significant barrier is data fragmentation. The pharmaceutical industry houses data in scattered formats and systems, making harnessing AI’s full power difficult. High-quality, harmonized data is essential for reliable AI-driven predictions. Paraphrasing a quote from James Clear, author of Atomic Habits: you don’t rise to the level of your technology; you fall to the level of your data.
Lastly, we can’t overlook the cultural and mind shift needed for AI adoption in drug development. Many teams hesitate to move away from traditional workflows, and transitioning to AI-augmented processes requires strong organizational buy-in. In some cases, comprehensive training programs are necessary to overcome resistance and close skill gaps. However, AI should always be seen as an enhancement to human expertise, not a replacement. Its true power lies in augmenting decision-making, ensuring the best possible outcomes.
Why Data Quality is the Foundation of AI Success in Drug Development
At Intelligencia AI, data sets the foundation for everything. Without high-quality data, AI-driven predictions can crumble. Data quality is the critical first step toward achieving accurate and reliable outcomes.
For us, high-quality data encompasses several key attributes:
- Recency: Incorporating the most up-to-date information to reflect current realities.
- Comprehensiveness: Covering a wide array of relevant data points.
- Harmonization: Standardizing data from diverse sources to ensure consistency.
- Structure: Organizing data systematically for efficient processing.
We apply an engineering mindset to biological data, transforming fragmented datasets into a single source of truth that fuels better decision-making. This approach prepares the data for machine processing, enabling analysts to extract meaningful insights.
Addressing the AI “Black Box” Problem in Drug Development
Transparency is crucial in fostering trust in AI systems. To tackle the black box issue, we focus on making both our data sources and model outputs explainable. We also provide detailed explanations of the factors influencing our predictions. This means that alongside each prediction, we offer a comprehensive list of the drivers identified by our machine learning models.
We are meticulous in evaluating and monitoring our algorithms’ performance over time. In addition to standard data science techniques like retrospective analysis and specificity assessments, we use a prospective approach. This involves comparing our AI predictions to real-world clinical trial outcomes. While this is a time-consuming process, it is essential in building trust and ensuring our models perform reliably in real-world settings.
Educating Business Leaders: AI is Not a Magic Bullet
Managing expectations around AI in drug development is crucial. AI’s rapid advancement has led to significant excitement and overhyped expectations. While AI offers transformative potential, it’s not a magic bullet.
One of the most significant risks is AI hallucinations, where models generate outputs that seem plausible but are actually incorrect. This is particularly relevant for large language models (LLMs) and generative AI. Without proper validation, generative AI-created insights can mislead decision-making. That’s why human oversight remains essential.
The good news is that so-called classical or mature AI, e.g., machine learning algorithms—like the ones used for Intelligencia AI’s Portfolio Optimizer—do not suffer from hallucinations. They are reliable, accurate, and very close to being deterministic.
Another challenge is bias in AI systems. AI models are only as good as the data they are trained on. Poor-quality or biased datasets can lead to flawed predictions. At Intelligencia AI, we invest heavily in clean, structured, and diverse data to ensure our models are as unbiased and reliable as possible. (Find out more here, here and here).
To help business leaders navigate these challenges, I recommend:
- Start small – Pilot projects can demonstrate AI’s capabilities before full-scale deployment.
- Foster AI literacy – Leaders should understand AI’s limitations and potential risks.
- Maintain expert human oversight – AI should support, not replace expert judgment.
Organizations can build confidence in AI by setting realistic expectations and ensuring robust validation while avoiding common pitfalls.
Promising AI Opportunities in Pharma
Beyond augmenting decision-making, AI holds immense promise in several areas of drug development, particularly in:
- Drug discovery and target identification – AI can analyze vast biological datasets to uncover new therapeutic targets that traditional methods may overlook. This enhances precision and accelerates research.
- Drug repurposing – AI can identify new indications for existing drugs, offering a faster, cost-effective approach to therapy development.
- Clinical trial optimization – AI improves patient selection, clinical trial design, and data analysis, making trials more efficient and reducing costs.
Looking Ahead: AI Drug Development to De-Risk and Accelerate
At Intelligencia AI, we are committed to leveraging AI to de-risk drug development and make the process more cost-effective. AI is already transforming pharma, and its impact will only grow. As an industry, we must ensure responsible AI adoption, balancing innovation with scientific rigor to unlock its full potential.
Want to Listen to the AI Drug Development Podcast Interview? Here’s How to Tune In
If any of these ideas resonate, please listen to the full conversation on the AI in Business podcast. You’ll also hear about how to mitigate bias in AI systems, more real-world applications of AI in pharma, and the future of AI transforming drug development.