Novo Nordisk, which now calls itself Novo, put its headcount on slide seven. At its Capital Markets Day in London on September 21, CEO Mike Doustdar’s corporate strategy deck listed “Reduced workforce to ~66,000 FTEs” and, beside it, “-13,000 FTEs vs peak.”
That is 4,000 more than the number Novo gave last September, when Doustdar, two months into the job, announced 9,000 layoffs worldwide. A spokesperson told BioSpace the next morning that the extra 4,000 were not necessarily layoffs. They were attrition, retirements, voluntary exits, and “decisions not to backfill roles.”
Doustdar’s explanation on stage was that the savings are being redeployed. “We’re using the funds coming through this exercise to fuel for the future, fuel the R&D organization,” he said. The deck put a figure on it: more than DKK 10B in savings redirected to commercial and R&D.
What the R&D money is buying
The afternoon R&D session had two presenters. One was Martin Holst Lange, the chief scientific officer. The other was Mishal Patel, Novo’s GVP of AI & Digital Innovation. A company that seats its head of AI next to its CSO for the R&D strategy pitch is telling investors how it now thinks R&D gets done.
Slide seven of the R&D deck is titled “AI is leveraged to improve speed, success and productivity across the research & development value chain.” Its list of 2026 achievements reads like a work-allocation memo:
- 10x faster clinical study report generation
- 90% less time on site selection and population analyses
- More than 3x improved target hit rate, from prioritizing genes found in virtual assays
- More than 3x more molecules tested on average, with a four-week design-make-test-analyze cycle
Slide eight walks through one case. Searching for a once-monthly amylin analogue, Novo ran human-designed and machine-learning-designed sequences through the same active-learning loop. It tested roughly 2,000 molecules, cut time to candidate by 50%, and needed about 60% fewer design cycles.
Five days before the London event, Novo announced a collaboration with Anthropic to test Claude Science in R&D workflows and to use Anthropic’s models for “agentic software engineering.” In that release Doustdar restated the ambition “to become the world’s most AI-driven healthcare company.”
The link Novo didn’t draw
Novo has never tied the 13,000 to AI. Its stated reasons are commercial: Eli Lilly taking share in obesity drugs, a semaglutide patent cliff ahead, a cost base built for a boom that has cooled. Those reasons are real, and investors didn’t find the new plan convincing either. Novo’s U.S.-listed shares closed down nearly 8% on the day.
Read the two decks side by side, though, and the arithmetic lines up. The strategy deck says the money moved to R&D. The R&D deck says R&D now runs faster on AI. And the specific workflows it names are some of the most labor-intensive white-collar work in a drug company.
A clinical study report is a document of hundreds of pages, assembled by medical writers from trial data, statistical output and protocol history, then revised through several review rounds. Generating it 10 times faster does not mean the same team writes ten times as many. For a company whose trial count is fixed by its pipeline, it means fewer writers per report. Site selection and population analysis belong to clinical operations feasibility teams and pharmacometricians. When the time spent on that work drops by 90%, the hours are what go first, and the headcount follows at the next budget cycle.
The org chart points the same direction. The strategy deck lists an “R&D reset” that merged research and development into one organization, and it credits the reorganization with cutting the number of direct reports to the CSO by more than half. Fewer boxes at the top of R&D, a faster workflow underneath, and a budget line that says the savings stay inside the function. None of that requires anyone to use the words “replaced by AI.” The work gets done by fewer hands, and the reorganization gives the reduction a name that is about structure rather than software.
That second step rarely shows up as a layoff. The 4,000 unfilled roles are the mechanism in miniature. No WARN notice, no works council, no press release. Someone leaves and the requisition never opens.
The cohort
We have tracked this shape of shrinkage all year. Globant’s Q2 held revenue to within $200K of the prior year with 2,673 fewer employees, the fifth straight quarterly decline. Oracle booked $19.3B with 21,000 fewer people on payroll. Prudential cut 196 Newark jobs in three rounds without once naming AI.
Novo is the first pharma company in the corpus at this scale. BioSpace, citing S&P Capital IQ, puts its 2025 peak at more than 78,000 employees. Losing 13,000 from that base is roughly one job in six, inside a year. Danish reporting based on Børsen’s figures puts the Denmark share alone at more than 7,000 (The Local); Novo has declined to give a country breakdown.
What this means for pharma white-collar work
Three job families sit directly under the slide.
Medical writing and regulatory documentation. Clinical study report generation is the single workflow Novo quantified at 10x. Submission documents are built from the same source material, so the automation extends naturally.
Clinical operations feasibility, biostatistics and pharmacometrics. Site selection and population analysis are named at 90% less time. These are specialist roles that were scarce five years ago; the scarcity premium erodes first.
Early discovery chemistry and screening. Fewer design-make-test-analyze cycles per candidate means fewer rounds of bench work per program, even as the number of programs rises.
The pressure runs the other way on pipeline volume. Novo’s 2030 R&D targets are more than 15 first human doses a year and at least 10 Phase 3 programs. More programs, fewer people, and the gap between the two is being closed by the workflows on slide seven.
Two things to watch. Novo’s full-year headcount in its annual report early next year will show whether the 66,000 keeps falling. And the next big-pharma investor day will show whether a second company puts a “10x faster reports” line on a slide. If one does, the contraction in medical writing stops being a Novo story and becomes an industry one.
