This article looks at how tasks may change in medical writing, regulatory affairs and clinical operations, without assuming a fixed forecast for jobs or productivity. The aim is to help leaders build a view of change from the work their teams actually do.
Assess tasks rather than predict whole jobs
AI may change parts of a role without removing the need for that role. Drafting, searching or assembling information can sit alongside review, judgement and communication that remain important.
Assess the actual workflow, intended use and limits of the available tools. Broad predictions about how many jobs will change are less useful than a clear view of the tasks your teams perform and the evidence needed to change them.
A task-level assessment usually starts by breaking a role into its recurring activities and asking three questions of each. How much time does it take today? How much judgement does it require? What would go wrong if it were done badly? Activities that are time-consuming, rules-based and low in consequence are reasonable candidates for assistance. Activities that carry scientific, regulatory or patient safety weight need far more care, even when a tool appears capable of producing a plausible output.
This exercise is most reliable when it is done with the people who perform the work. Process documentation often describes how work should happen rather than how it does, and the informal steps, such as checking a colleague’s interpretation or chasing a missing reference, are frequently where the real effort sits.
Medical writing and scientific communication
AI tools may assist with drafts, summaries and information retrieval. The resulting material still needs appropriate checking for source accuracy, scientific meaning, completeness and suitability for its audience.
A useful assessment separates drafting from review and measures both. Faster first drafts do not necessarily mean a shorter overall cycle if the output creates additional correction or checking work.
Consider a medical writing team producing plain-language summaries. An AI tool might produce a readable first version quickly, but reviewers then need to confirm that every statement is supported, that nuance about limitations has survived simplification and that the reading level suits the audience. If reviewers lack access to the sources the tool relied on, checking takes longer than writing from scratch. In a case like this, the useful design question is not how to draft faster but how to make verification efficient, perhaps by having the tool cite the passage behind each statement.
Writers’ roles may shift towards specifying the brief precisely, curating sources and exercising editorial judgement. Those are skills experienced writers already have, but they may need to be made explicit and supported with training.
Regulatory affairs
Document assembly and information management may offer opportunities for assistance or automation. The appropriate approach depends on the submission process, systems and requirements involved.
Regulatory judgement, responsibility for the submitted material and communication with authorities need clear ownership. Assign any new AI-related responsibilities within the operating model rather than assuming every organisation needs the same new job title.
Much regulatory work involves locating, comparing and reconciling information across documents and systems. Tools that help find the right version of a document, flag inconsistencies between sections or track commitments can reduce administrative effort. In some cases the right answer is conventional automation rather than AI at all: a structured workflow application that replaces spreadsheets and email trails may deliver more reliable value than a generative tool. The choice should follow the problem.
Whatever tools are used, the organisation remains responsible for what it submits. Validation expectations under GxP, data protection obligations and the organisation’s own quality system all shape what is acceptable, so regulatory and quality colleagues should be involved in deciding where assistance is appropriate.
Clinical operations
Potential use cases include support for planning, information review and operational monitoring. Each needs careful assessment against the intended use, data and oversight requirements.
Consider the work in connected teams as well. Improving one task may move the bottleneck to review, data management or another dependency, so evaluate the whole process.
Imagine a study team that uses a tool to summarise site visit reports more quickly. The summaries arrive faster, but the data management team now receives more queries to resolve, and the overall timeline does not improve. Mapping the end-to-end process before introducing the tool would have shown where capacity was really constrained. This is why process mapping is often the most valuable early step, before any technology choice is made.
Common pitfalls
Several patterns recur when pharma teams plan AI-driven changes to work. Measuring only the step the tool performs, rather than the whole cycle, overstates benefit. Assuming that time saved will automatically be redeployed to higher-value work ignores the need to plan that redeployment. Treating training as a one-off launch activity leaves users unsure how to handle the cases the tool gets wrong. And introducing change without explaining its purpose can create understandable anxiety about job security, which reduces the honest feedback the organisation needs.
Plan changes with the people doing the work
Map tasks, test representative cases and involve business, scientific, technical and people specialists. Identify training needs and agree how quality, workload and user experience will be assessed.
A practical sequence is to select one workflow, map it with the team, agree baseline measures, run a time-limited test on representative material and review the results with everyone involved. Content workflows are often a sensible place to start because outputs are visible and review is already part of the process. Our AI-assisted content workflow with human approval illustrates how drafting support and human sign-off can sit together in a regular cycle.
Where to start
Glyz supports workflow review, digital strategy and capability planning through our AI and digital transformation services, and helps teams build the skills to work alongside new tools through capability and change programmes. The aim is a practical view of what can change, what needs oversight and what evidence should guide the next decision.
