AI enables a more agile medical affairs team

Medical affairs has evolved into the third strategic pillar in pharmaceutical and life sciences organisations, serving as a critical bridge between R&D and commercial teams. As this role has grown in importance, it has also increased in complexity. Today, medical affairs teams manage an enormous volume of data from various sources that must be translated into medical insights that inform decision-making across the organisation.

The growing volume, velocity, and diversity of data are reshaping how medical affairs teams operate. Manual review processes are time-consuming and inefficient, and disparate data sources heighten the risk of missing insights essential to communicating a drug’s value. But artificial intelligence (AI) enables medical affairs teams to work more efficiently and effectively by reducing the time and effort required to synthesise information.

When applied responsibly, AI can analyse large volumes of data and rapidly generate reliable insights while maintaining human oversight at every critical decision point. As a result, critical information can be identified and conveyed more rapidly to provide timely and impactful support to commercial and R&D teams.

Responsible adoption

Successful AI transformation for medical affairs teams requires a clear understanding of potential obstacles and a thoughtful and deliberate approach to launching and sustaining AI deployments.

Organisations must be cognisant of key challenges associated with AI — particularly large language models (LLMs) — in their strategy, including:

• Data reliability: Challenges like biased input data and the potential for hallucinations to affect data outputs.
• Scale demands: Complex, diverse, and sensitive biomedical data require significant computational scale, which can be costly.
• Regulatory fragmentation: Conflicting AI laws are in place across more than 70 countries, and no harmonised global AI regulations exist.¹

Mitigating these challenges while addressing the needs of medical affairs teams requires a few key elements, too:

1. Trust. Building trust in AI within a medical affairs team is essential to its successful adoption. Talking to teams to understand where they’re spending disproportionate time on manual, repetitive tasks and introducing high-impact, low-risk AI applications to address those pain points can help establish trust and build optimism. A strong change-management strategy, user-friendly AI tools, and training also help build confidence that AI can improve efficiency and effectiveness.
2. The right tools. Not every challenge requires the same AI solution, so, it’s important to choose the right tool for the task. Factors like accuracy, cost, reliability, and ease of integration should all be considered. For some tasks, such as literature reviews that require reproducible, defensible, and transparent results, a hybrid approach that combines AI with other traditional techniques may be the optimal solution.
3. Continuous monitoring. Maintaining the reliable performance, compliance, and long-term value of AI tools requires continuous evaluation. This includes adhering to proven frameworks to review and validate AI outputs and underscores the importance of keeping a human expert in the loop.

Defensible AI use cases

AI can deliver meaningful value across medical affairs operations. The AI use cases that create the most impact for these teams are those that enhance data analysis and insight generation and streamline processes.

AI use cases must also be defensible, meaning they are built on high-quality data, evaluated for reliability, and capable of producing auditable, trustworthy outputs.

Four high-value, defensible AI use cases available to medical affairs teams today include:

1. Scientific literature review

A foundational tool for evidence-based decision-making, scientific literature reviews require robust methodological rigour and adherence to proven guidance frameworks, such as the Cochrane Handbook. They also require significant manual effort as volumes of published literature continue to grow.

Integrating agentic AI into literature reviews enables medical affairs teams to perform this essential work more quickly and efficiently while maintaining accuracy, privacy, and traceability. Autonomous AI agents can support protocol development, document searching and screening, and insight generation to reduce burdens on teams, while keeping an “expert in the loop” ensures output validity. They can also generate audit trails and use reproducible workflows to align with regulatory expectations.

Agentic AI is already delivering measurable acceleration in real-world literature reviews. In one systematic literature review focused on non-small-cell lung cancer, an agentic literature review solution reduced the number of documents that required manual screening by 87%.¹ For a highly complex living review, this solution achieved the same degree of recall as manual extraction while doing the work 2.5-3 times faster.²

2. Chart review

A rich source of patient-level data, chart review has broad application across the entire product lifecycle, from discovery through post-market. It can help identify unmet needs, support real-world evidence generation, and inform strategic decision-making.

The challenge is that about 80% of electronic medical record data is estimated to exist in unstructured formats, such as clinical notes, scanned documents, and free-text entries.³ This makes manual chart review time consuming and difficult to properly validate.

Augmenting chart review with AI applications employed from discovery to post-market speeds up information extraction from various sources, improves insights, and reduces burdens on staff. For example, AI can help confirm clinical trial feasibility by assessing protocol alignment and identifying eligible patients from historical records. It can also identify adverse events and patient-reported outcomes from clinical narratives to support post-market surveillance.

The tangible benefit to medical affairs teams is gaining deeper insights into patient data much more quickly than is possible in a manual review. In one real-world case, natural language processing (NLP) AI was used to accurately unlock insights from 34 million clinical notes across more than 100,000 patients. This resulted in a more than 100 times reduction in manual effort for site staff.⁴

3. Key opinion leader (KOL) identification

Identifying the leaders who influence scientific dialogue, guide treatment adoption, and shape the understanding of diseases has long been a manual task. It requires time-intensive reviews of publication lists, clinical trial registries, and other documents, as well as additional time spent interpreting and updating information. These efforts can only be applied to certain periods of time and can miss current conversations or overlook emerging voices.

AI empowers professionals to better understand the evolving landscape of scientific, clinical, and digital influence. It can scan vast data sets and identify experts more comprehensively and with greater precision. Additionally, it can capture the context, reach, and emerging ideas of experts. For instance, it can reveal which articles by an opinion leader have the most citations or the topics most often presented at conferences.

With the help of AI, medical affairs teams can spend less time collecting and analysing data and focus on strategic planning and conducting more personalised outreach.

4. Medical communications

Translating complex scientific data into impactful messaging for patients, healthcare providers (HCPs), and other stakeholders is another time-intensive activity. This work requires synthesising large volumes of scientific literature and other structured and unstructured data, then tailoring and sometimes translating content for these diverse audiences.

AI enables medical affairs teams to access and assess large volumes of data to inform integrated medical communications plans and accelerate the creation of personalised, globally accessible medical communications. For example, it can provide more relevant, personalised, and accessible scientific information for HCPs and clearer, more meaningful education that supports understanding and outcomes for patients. Neural machine translation engines enable faster translation of scientific content across languages, providing broader access with scientific consistency.

When using AI to generate content, human oversight is critical to maintaining accuracy, relevance, and trust. Expert-in-the-loop checkpoints help safeguard against bias, hallucinations, and inaccuracies to give medical communications the scientific validity they need.

An evolution, not a revolution

A thoughtful and phased approach to integrating AI into medical affairs can deliver meaningful impact now and in the future, but AI transformation requires proper governance. Responsible use and human oversight are critical to ensuring its reliability and delivering value. By selecting the right AI application for the task and applying expert-in-the-loop validations, AI can greatly reduce manual burden across the medical affairs function and allow teams to focus on high-value work.

References

1. Rath N, et al. Assessing the feasibility of applying natural language processing for systematic literature reviews: A case study in non-small-cell lung cancer. Value Health. 2023;26(12 Suppl):S414.
2. Raghuram A, Ssentongo A, Pereira LM, Kuang Y, Uyei J, Park P. Use of natural language processing to extract published real-world data on a COVID vaccine and antiviral treatment. Open Forum Infect Dis. 2023;10(Suppl 2).
3. Au Yeung J, Shek A, Searle T, Kraljevic Z, Dinu V, Ratas M, et al. Natural language processing data services for healthcare providers. BMC Med Inform Decis Mak. 2024;24:356.
4. Miliard M. How Mercy is using NLP with its Epic EHR to improve analytics for cardiac care. Healthcare IT News. 2018 Jun 8.

About the author

Niamh McGuinness is director of pharma solutions at IVIA Applied AI Science. She leads a diverse group of experts who build AI solutions and utilise a suite of award-winning technology to help pharma customers protect participant privacy and surface critical insights from unstructured data and literature for a vast array of use cases from bench to bedside, including R&D, drug safety, and medical affairs. McGuinness has been with IQVIA for nine years. She received her PhD in Neuroimmunology from Trinity College, Dublin.