Pharma scientists tell Reuters AI is speeding up early drug discovery tasks


A Reuters investigation based on new interviews with senior scientists at global pharmaceutical companies found that artificial intelligence is reducing the time required for certain early drug discovery tasks. Executives at seven major companies reported notable acceleration in molecular design and preclinical work, while human trials and regulatory approvals remain necessary.
Testimonies from senior scientists at global pharmaceutical companies document a practical shift beyond drafting administrative documents toward compressing molecular design cycles from weeks to shorter periods, while a fundamental gap remains between accelerating lab work and approving a safe, effective medicine for patients.
Reuters reported in an investigative article published on November 20, 2026, at 11:02 UTC (15:02 GST), a new series of interviews with leading pharmaceutical industry scientists indicating that artificial intelligence is shrinking the time required to complete certain early-stage drug discovery tasks.
Executives at GSK, Bristol Myers Squibb, Novartis, Eli Lilly, Novo Nordisk, AstraZeneca, and Sanofi described practical applications that go beyond merely drafting administrative documents, estimating that selected stages of molecular design and preclinical development work have become measurably faster.
It is important to draw a clear distinction between accelerating molecular design in the lab and achieving an officially approved medical treatment. A computational model can help a research team generate or prioritize chemical structures, identify promising biological targets, or interpret complex experimental data.
However, drug candidates must still pass laboratory testing, toxicity assessments, human clinical trials, and rigorous regulatory review. Therefore, shortening the early research cycle does not demonstrate a proportional reduction in the total years required to prove therapeutic safety and efficacy.
Historical experience in the pharmaceutical industry indicates that the largest share of therapeutic compound failures occurs during human clinical testing stages due to safety concerns or limited therapeutic effect within the complex human body.
Reuters highlighted specific examples including Sanofi’s research on the investigational asthma treatment lunsekimig, Novartis’s evaluation of biological targets linked to kidney disease, and changes in design cycles at Novo Nordisk. Some companies described significantly compressing iterative cycles that previously took several weeks.
These accounts represent company-reported experiences tied to specific workflows and should not be treated as broad, generalized clinical outcomes. The investigation also noted a growing pipeline of AI-linked drug candidates, most of which remain in early stages and may never gain final approval.
Biological models can detect useful patterns across proteins, molecular structures, and other scientific data, but AI predictions remain highly sensitive to training data quality, laboratory conditions, and the biological question under study. Researchers also need to verify whether a prediction reveals a true causal mechanism or merely a passing statistical correlation.
For this reason, major pharmaceutical laboratories continue investing in parallel in wet-lab experiment automation and building high-quality biological databases, recognizing that computational prediction quality depends directly on the accuracy of the experimental measurements feeding the model.
The pharmaceutical sector’s interest in companies like Anthropic reflects a broader convergence between foundation models and scientific computing. Commercial AI providers assist with reasoning and data analysis while specialized molecular systems tackle different technical problems, without proving that a general-purpose chatbot alone can invent an effective drug. Future evidence will come from reproducible preclinical results, independently measured development timelines, completed clinical trials, and regulatory approvals.
Based on these factors, the real impact of AI in the pharmaceutical sector is measured by its ability to raise success rates for compounds advancing to late clinical stages and reduce the cumulative long-term cost of failed research.
Question: Which major pharmaceutical companies did the Reuters investigation document using AI?
Answer: The investigation included interviews with scientists and executives at GSK, Bristol Myers Squibb, Novartis, Eli Lilly, Novo Nordisk, AstraZeneca, and Sanofi.
Question: Does accelerating molecular design with AI guarantee faster drug approval?
Answer: No, because candidate compounds are still required to pass laboratory tests, toxicity evaluations, human clinical trials, and regulatory approvals.
Question: What are the main research examples mentioned in the Reuters report?
Answer: The report cited Sanofi’s research on the investigational asthma treatment lunsekimig, Novartis’s evaluation of kidney disease targets, and accelerated design cycles at Novo Nordisk.
Question: What role do foundation model companies like Anthropic play in pharmaceutical research?
Answer: Foundation models support logical reasoning and scientific data analysis alongside specialized molecular systems, without replacing laboratory experimentation.