Generative AI’s Impact on Medical Writing: Insights from Occupational AI Research

Recent research from Microsoft offers one of the most comprehensive looks yet at how generative AI is already reshaping work activities across industries. Analyzing 200,000 anonymized Bing Copilot conversations, the authors developed an “AI applicability score” for occupations, measuring where AI is already demonstrating relevant capabilities. While their study spans the entire U.S. workforce, the findings have direct implications for medical writing—particularly when thinking about AI-powered document authoring and medical writing automation platforms. 

How the Study Relates to Medical Writing 

User Goals vs. AI Actions 

The research distinguishes between user goals (what a human wants to achieve) and AI actions (what the AI actually does). In many cases, these do not overlap—users may request research, while AI provides explanations or training. 

This distinction matters in medical writing, where AI can: 

• Assist writers in gathering and synthesizing scientific evidence (high-frequency, high-success activities in the study). 

• Perform drafting or editing tasks, from automated CSR creation to protocol section generation. 

• Provide structured feedback or training for junior writers, helping standardize quality and style. 

High-AI-Applicability Tasks in Medical Writing 

Occupations involving high levels of information gathering, writing, editing, and explaining—core components of regulatory and clinical documentation—scored among the highest in AI applicability. This suggests that medical writers and publication teams are already working in domains where AI’s demonstrated strengths align closely with day-to-day responsibilities. 

Practical AI Uses for Medical Writing Teams 

AI Strengths Highlighted in the Study 

In the study, writing and editing activities had among the highest satisfaction and completion rates. For medical writers, this reinforces the potential of AI not just as a time-saver but as a quality enhancer when paired with structured authoring frameworks and human oversight. 

For example: 

• Speeding literature searches for faster context setting in regulatory sections. 

• Producing consistent, submission-ready document preparation outputs across multiple deliverables in a single trial. 

• Integrating AI for protocol development with downstream authoring tools to maintain alignment from trial design through final submission. 

Biases and Limitations Relevant to Medical Writing 

The authors acknowledge several limitations that matter when translating these findings into a medical writing context: 

Platform Bias 

The data comes from a single AI platform (Bing Copilot) over a defined 9-month period. Usage patterns and strengths may differ across tools such as ChatGPT Enterprise or domain-specific AI for life sciences. 

Occupation Decomposition Bias 

The mapping from tasks to occupations relies on O*NET’s classification system, which may not fully capture the nuanced, cross-functional nature of medical writing roles that blend regulatory, scientific, and editorial expertise. 

Work vs. Leisure Ambiguity 

The study can’t always distinguish between conversations done in a professional context vs. personal use, which could over- or under-estimate applicability in specialized professions. 

Task Scope and Quality Variance 

Even when AI demonstrates capability in a work activity, completion rates are not 100%, and the scope of impact is often “moderate” rather than “complete,” underscoring the importance of human oversight in high-stakes deliverables. The results seem to indicate that people are using AI for what they’ve discovered it’s currently good at, not for what it could be in the future to optimize efficiency. 

Key Takeaways for Medical Writing Leaders 

This research reinforces that generative AI is particularly well suited for the information-heavy, writing-centric nature of medical writing. However, adoption strategies must go beyond “AI can do X” toward “AI can reliably assist with X in our regulated, quality-driven workflows.” Leveraging medical writing automation platforms designed for compliance and accuracy—while being aware of technology and data biases—will be key to unlocking sustainable productivity gains.