Improving Pharma Launch Performance With AI-Driven Decision Support

Pharmaceutical Commerce: Deloitte cites that about one-third of drug launches miss expectations. What concrete metrics or case examples show how new decision-support tools can improve launch performance compared to traditional approaches?

Lance Hill: Many launches fall short because launch teams lack a complete view of the market. Before any drug or therapy reaches launch, you need to understand who will prescribe it, which patients need treatment, where they’re located, and how access will be covered. Traditional market research can provide these insights, but the information is often weeks or months out of date.

Real-time data and AI reveals critical gaps much earlier. In one launch, physicians reported that patients with comorbid mental health conditions were frequently delayed in receiving GI diagnoses. When those findings were combined with congress monitoring and field insights, the company quickly identified an overlooked HCP population, broadened education, and improved alignment across functions. The result was faster diagnosis, stronger engagement, and better patient access.

This validates how modern decision support transforms outdated intelligence into timely, actionable strategy.

PC: When integrating inputs such as field activity, HCP engagement, claims, and social sentiment, how can companies ensure the data is normalized, weighted appropriately, and not skewed by noisy or biased sources?

Hill: You want to be working with a partner who specializes in that kind of work. It takes years of experience working with each of those data sets individually to know what really matters, and how it should be weighted in different circumstances, particularly when you’re combining different data sets. At Within3, we've spent years developing our algorithms and data integration approaches to be able to do this effectively. We also have years of client feedback that tells us what we need to adjust to get to a point where the precision level is high enough that launch teams can reliably make good decisions.

Each of these data sets is different in its own right. When looking at global social listening data, for example, it really is a needle in a haystack. It's about managing that noise to understand what matters versus what's irrelevant.

PC: What types of independent validation or audits are important to confirm the accuracy and reproducibility of AI-driven insights in the context of product launches? How transparent should the outputs be for commercial, medical, and compliance teams?

Hill: It matters what you're using AI for, because certain use cases require different levels of accuracy than others. For example, analyzing markets for launch intelligence requires that we always ensure we provide traceability back to the underlying data. It’s like how in the publications world you have citations that bring you all the way back to the source content. We do the same thing back to the underlying data elements that tell us why we're seeing a signal that suggests you could be doing something different in any particular aspect of the market.

If an organization is as serious about responsible AI as we are, it should be able to publish material that tells people exactly what it's doing—it's not smoke and mirrors. You need a transparent AI governance structure and a regulatory compliance framework that shows how you intend to comply with relevant local, national, and international regulations. You need clear processes for incident management, both how incidents are reported, and the steps for resolution. And you need ongoing training and education to keep critical staff informed on AI best practices.

When building an AI system, you're combining human expertise and analysis with quality assurance processes, data science processes, ever-improving algorithms, and ever-cleaner data. Those things come together so that six months from now, the system is going to be vastly more agile and accurate than it was today.

PC: With sensitive healthcare data in play, how can organizations balance real-time analytics with HIPAA/GDPR compliance, data segregation, and robust model governance practices?

Hill: The good news is that in 20-plus years, the types of data we're handling haven’t changed, perhaps with the exception of social data. The industry has been dealing with clinical data, scientific data, and claims-based data for years and years, and has already determined which people and roles within the organization should have access to what data, and at what level of granularity. As a technology company, all you're really doing is overlaying your technology in a way that matches those established rules, and that allows different organizations to access the same system in different ways.

There are also data sets that are used reliably across the industry and are not company-specific. For example, the fact that I was prescribed an antibiotic yesterday is not unique to Pfizer, AstraZeneca, or any other company—it's an established fact about me as a patient. That’s not proprietary data that a pharma company should worry about getting out into the wild. But if you are using sensitive data, then it’s important to be transparent. How is that data being used? Is it being used to train AI tools, or not? In our case, the answer is always no. The Within3 system is set up using the highest compliance safeguards.

However, I think there's a little bit of an overcorrection on the part of pharma companies at this point, because the technology is still new. Right now, we’re seeing a bias towards being very cautious. I believe that will open up over time as compliance teams in particular get more familiar with the domain they're dealing with. It’s the same change resistance we’ve seen with every technology wave over the last 30 years.

By adopting AI-driven decision support, pharma organizations can address drug launch failure rates through precision and compliance. Transparent governance and a commitment to excellence allows companies to navigate complex markets with confidence and integrity. Together, these factors ensure that impactful therapies reach the right patients swiftly and responsibly.