Pharmacovigilance
ISCO-08 2262, 3344, 4110, 2641Drug safety officer, Pharmacovigilance case processor, Safety physician, Signal detection scientist, Pharmacovigilance manager, QPPV / deputy QPPV
What this job is made of
The main activities of the job family, the share of working time each one takes, and how much of that time generative AI can automate.
- Case intake and coding in the safety database30% of working time80% exposure
AI reads incoming reports, extracts the fields, proposes MedDRA codes and checks duplicates; the processor reviews exceptions.
- Case narratives and follow-up15% of working time70% exposure
Standard narratives and follow-up letters to reporters are drafted automatically from the case data.
- Medical assessment of cases and causality15% of working time20% exposure
Seriousness and causality assessment stays with the safety physician; AI surfaces similar past cases for comparison.
- Signal detection and evaluation15% of working time35% exposure
Statistical screening and data compilation are automated; deciding whether a signal is real and what to do remains human.
- Periodic safety reports (PSUR, DSUR)15% of working time55% exposure
AI compiles the tables and drafts descriptive sections; the benefit-risk conclusion is written and signed by the team.
- Regulatory interactions and inspections10% of working time15% exposure
Answering authorities, hosting inspections and QPPV accountability stay with people; AI helps retrieve documents.
Headcount need, in FTE
Target AI skills
level 1 to 4Job skills
Growing in value
- Medical review of complex cases
- Signal evaluation and benefit-risk assessment
- Validating automated case processing
- Inspection readiness for AI-assisted processes
Losing value
- Manual case entry in the safety database
- MedDRA coding of reported events
- Writing standard case narratives
- Duplicate checks and literature screening
How the job will change
Most pharmacovigilance headcount sits in case processing: reading incoming reports, entering them in the safety database, coding events in MedDRA, writing narratives, checking duplicates. These are the steps AI handles best, and intake and narrative automation is already on the market. Literature screening for adverse events and the first draft of periodic safety reports are following the same path.
What remains is medical judgement and accountability: assessing causality on a difficult case, deciding whether a signal is real, arguing a benefit-risk position before an authority. The team also has to prove to inspectors that automated steps are validated and controlled. A good pharmacovigilance professional tomorrow combines clinical reasoning with the ability to audit a system they did not build.
- 2026-2027
- Automated case intake and narrative drafting, with full human review.
- 2028-2030
- Review moves to sampling; case processing teams shrink, often at providers.
- 2031+
- Smaller teams centred on signals, benefit-risk and system validation.
Case processing has been the entry route into drug safety. If it disappears, companies will struggle to grow the safety physicians and signal experts they will still need. See the seniority outlook below.
What if you hired fewer juniors?
Your 2036 seniors are the juniors you hire today.
Advanced settings modified
2026 2036
The AI Cookbook 2026
albert's guide to cut through the noise around generative AI and turn it into workforce decisions.
- What AI actually changes in jobs and skills
- Why the junior pipeline matters more than most companies realise
- How to integrate AI into workforce planning

Run these scenarios on your actual workforce
AI Impact Diagnostic: 6 to 8 weeks, your data in albert, three costed scenarios and the projected seniority mix for each job family.
How the numbers are built
Every headcount figure combines four assumptions. Three come pre-filled from public research and job family defaults. The strategic ceiling is yours to set.
Seniority outlook
A flow model with three levels of experience in the profession. Promotion and exit rates are set so that today's mix stays stable when hiring does not change: any gap you see comes from the junior hiring cut alone.
Sources
- International Labour OrganizationGenerative AI and jobs: a refined global index of occupational exposure (2025)
- ILO datasetTask-level GenAI exposure scores by ISCO-08 occupation
- AnthropicAnthropic Economic Index (June 2026 release, April and May 2026 usage data)
- Stanford Digital Economy LabCanaries in the Coal Mine? Six facts about the recent employment effects of AI