5 Common Mistakes That Risk Your Pharma AI Efforts

By Kelly Waller

While it’s true that artificial intelligence can occasionally falter, the last thing we want to do is compound these errors with misinformed data strategies. The risks are significant, particularly in the context of AI engagement.

This can be a tall order as the global pharmaceutical industry races to adopt AI technologies for patient engagement. Generative AI alone could unlock $110 billion in economic value annually for pharmaceutical and medical product companies, according to a 2024 McKinsey & Co. report.

Indeed, AI tech can produce blockbuster results. Novo Nordisk, the maker of the Ozempic, uses AI to process large volumes of data for clinical trials. These insights could improve drug adherence by pinpointing patient motivations and compliance challenges.

However, the urgency to incorporate AI into an overall data strategy can cause pharmaceutical companies to miss potential hazards, resulting in patients receiving the wrong messages. In our experience, five strategic factors can make a company vulnerable to these mistakes.

The 5 Top Data Mistakes Pharma Companies Make In AI

Personalizing for drug patients is uniquely complex because it must adhere to industry-specific regulatory standards. A data architect can guide your organization through these standards while preventing missteps from derailing AI strategy.

  1. Cutting corners on data consolidation. AI technology is only as good as the data that feeds it. The most reliable data is unified. Correctly executed, unified data strategies manage the entire drug lifecycle, from clinical development to commercialization, while integrating data management processes, including generation, enhancement, storage, and application. If the data is insufficient, the insights will be unreliable and biased. Because of these intricacies, many pharmaceutical firms partner with external data experts.
  2. Not giving potential bias its due. It’s well-established that a medication’s effectiveness varies based on gender, race, behaviors, and socioeconomic factors. Consequently, potential biases exist in all data sets, threatening the best personalization efforts. For example, biases in AI applications can throw off patient adherence predictions throughout the drug lifecycle. Tools for identifying and mitigating biases include systems that seek out diverse datasets, algorithms that detect and correct biases, and programs that regularly monitor and update data.
  3. Ignoring real-world evidence (RWE): RWE—clinical evidence of a drug’s efficacy and safety—is revolutionizing data strategies in the pharmaceutical industry. Achieving it entails integrating data from electronic health records, medical claims, patient registries, and other sources, then supplementing it with deep intelligence based on behaviors, intent, psychographics, and technographics. The resulting demographic insights can expose deficiencies in clinical trials and inform decision-making across a drug’s lifecycle, significantly improving patient outcomes.
  4. Not shining a light on dark data. The issue of dark data—data collected but not recognized, used or accessed—is a looming threat to pharmaceutical engagement. According to CIO magazine, 40% to 90% of organizational data is dark, presenting a considerable waste of time and money and missed opportunities in AI applications. In particular, extensive dark data can contribute to patient bias and related inaccuracies, neutralizing an investment.
  5. Failing to prioritize security and governance. The data strategy discussion, especially in the context of AI, covers many shiny features, but none should outshine security and governance. The extent to which an organization’s regulatory infrastructure can identify, filter and categorize data can make or break its ability to avert an information leak and maintain compliance. AI has the tools specifically crafted to bolster such security and governance.

These Mistakes Risk De-Personalizing the Patient Experience

We’ve seen firsthand that pharmacy companies earnestly want to avoid these mistakes. Their primary struggles stem from a lack of time to develop a thorough strategy and attempting to tackle too much at once.

This is what those companies risk losing.

  • Smoother healthcare access. A patient support program that meets individualized needs requires meticulously segmented and analyzed data, but that’s just the start. The data also must be comprehensive and unbiased to eliminate access barriers such as language, technological and healthcare illiteracy, skepticism, transiency, and challenges faced by people with disabilities.
  • Understanding patient motivations. Insightful data can interpret patient psychology and behaviors, including the reasons behind consistent or sporadic medication adherence. These insights tap into motivations that can inform solutions, such as health incentive programs that better align with patient needs and preferences.
  • Better healthcare, all around. Thanks to effective data strategies, medical professionals can diagnose conditions like diabetes up to six months in advance. That’s because these strategies combine the potential of each data point with AI tech to automate processes and forecast results. By investing the time, researchers aren’t merely opening doors to what’s possible; they’re building the portals.

Know When to Seek Help

Data architects have the expertise, experience, and resources to develop advanced data strategies, freeing pharmaceutical visionaries to focus on their core competencies. The journey towards better patient outcomes begins with a solid data strategy. Learn more about how Harte Hanks can help your company on the journey to better patient outcomes.


Kelly Waller is the global Senior Vice President of Sales and Marketing at Harte Hanks.

The editorial staff had no role in this post's creation.