Free tools Windows power users keep installed
One-click scans. No signup required.
AI is being used across drug research for different jobs: finding and ranking targets or molecules, designing biologics, screening possible off-target interactions, retrieving internal research, and supporting clinical-trial work. Those uses are not interchangeable, and none alone proves that a medicine is safe or effective. AI Weekly’s 2026 overview classifies 15 deployments, saying 10 are in production or have results and six have a reported outcome. That is the page’s editorial classification, not an independently audited count; the cases below are the examples for which the available accounts describe a specific workflow or result.
What counts as an AI deployment in drug research?
“AI in drug discovery” can describe anything from a model that ranks candidate targets to a knowledge assistant used by scientists or an AI-enabled clinical-trial workflow. The scientific question, development stage, and evidence differ substantially. A ranked molecule is not the same as a lab-validated candidate; a drug in clinical trials is not an approved treatment; and software adopted by trial teams does not show that AI improved a drug’s efficacy.
As an Amazon Associate I earn from qualifying purchases.
The examples below draw on company and vendor accounts, which are useful for understanding stated workflows but are not automatically independent evaluations. The 2026 review by Huynh, Seal, and the AIA4S Consortium, “AI agents in drug discovery: applications and case studies,” describes applications including literature synthesis, protocol generation, toxicity prediction, small-molecule synthesis, repurposing, and broader decision workflows. It characterizes the field as early and identifies data heterogeneity, system reliability, privacy, benchmarking, and real-world use as continuing challenges. Its survey of applications should not be read as proof that every described agent is a routine production system.
AI for target selection and molecule assessment
Novartis: supporting R&D decisions
In a company account dated July 28, 2026, Novartis says AI and digital tools support target selection, molecule assessment, clinical-trial design, information processing, and earlier decisions. It describes a task-focused approach, reuse and scaling of useful systems, human-centered practice, and embedded safety, quality, and privacy controls. The account describes an organizational strategy and workflow; it does not establish a measured improvement in clinical success attributable to AI.
#1 Best Overall
Novartis also cites fewer than 10% of drug candidates historically reaching the market as background to its R&D strategy. That is company-provided context, not a result of an AI deployment.
Novo Nordisk and Microsoft Research: data and model discovery
Microsoft’s customer story dated October 4, 2024, describes an Azure-based platform supporting regulatory affairs, early research, drug discovery, and trial design. Its components include tools to discover data and models, a research copilot, reusable reasoning templates, auditability, and infrastructure for building models. Microsoft reports initial predictive-model results for cardiovascular risk. Because this is a Microsoft customer story, the results should be attributed to that account rather than treated as an independent comparison.
AI for biologic design and molecular screening
Amgen: proposing and evaluating biologic candidates
NVIDIA’s customer case describes Amgen using generative models to propose biologic candidate designs and predictive models to evaluate them, with BioNeMo and DGX Cloud among the tools named. This is a vendor-published account of the workflow; it does not provide evidence here that a proposed design became a clinically successful medicine.
Innophore: ranking possible alternative targets
An NVIDIA case study says Innophore combines its Catalophore technology with GPU and cloud computing and BioNeMo for molecular interaction analysis and off-target screening. Across 467 screened targets, the case reports that the model identified, on average, 30% of experimentally determined alternative targets within its top ten ranked hits. This is a specific screening metric reported by Innophore through NVIDIA; it is not a general measure of drug-discovery accuracy, and a ranked hit is not itself a validated treatment.
Innophore: accelerating a particular binding-affinity computation
The same NVIDIA account reports reducing one protein-protein binding-affinity prediction workload from 2,000 CPU hours to 15 minutes, at 5% of the original cost. These figures apply to that computation as described in the case study, not to all molecular modeling or drug-development costs. The account does not establish that the speedup translates into faster clinical development.
AI combined with experimental biology
Recursion: turning cellular experiments into computational data
Google Cloud describes Recursion’s platform as combining cellular experiments, image processing, neural networks, and cloud computing. The case page reports five internal drugs advancing through clinical trials and more than 15 partner drugs in early discovery. Those are time-sensitive pipeline figures from Recursion as reproduced by its cloud vendor. Advancement through trials does not show that AI alone caused the progress, nor does pipeline status establish approval or efficacy.
Rank #4
AI for research knowledge and enterprise workflows
Amgen, Almirall, and UCB: finding institutional knowledge
Microsoft’s account dated September 14, 2026, describes Amgen’s Catalyst Copilot, Almirall’s AI research assistant, and UCB’s SKAI platform. The stated functions include retrieving organizational research knowledge and supporting enterprise use with protections for sensitive information. These are knowledge and workflow applications: they may help staff find or use information, but they are distinct from evidence that a model designed or validated a drug.
Bristol Myers Squibb: supporting clinical-trial operations
Accenture says Bristol Myers Squibb implemented more than 30 generative AI solutions and describes Workbench as an AI-enabled clinical-trial accelerator. Its case study reports adoption across more than 30 priority teams and nearly 900 unique users after three months. These are vendor-reported implementation and adoption figures, not measures of trial speed, drug efficacy, or molecular discovery.
How to judge the evidence behind a deployment
Compare like with like. A useful assessment asks what the system did, at which stage, and how the result was checked. Company or vendor statements can identify a deployment and explain its intended role, but their claims should retain clear attribution. Evidence of model performance on a screening task is narrower than wet-lab validation; neither by itself establishes clinical benefit.
- Task: Separate target or molecule ranking, candidate design, off-target screening, research search, computing infrastructure, and trial operations.
- Stage: Distinguish early research and preclinical work from clinical development and operational support for trials. A candidate’s stage is not an approval.
- Validation: Look for the stated test and comparator, including whether model predictions were checked experimentally or clinically. A specific metric should retain its denominator and task.
- Outcome: Adoption, processing speed, pipeline counts, and predictive performance measure different things. They cannot be combined into one general “AI impact” figure.
- Governance: Consider human oversight, privacy protections, data quality, reliability, and whether another team could reproduce the result.
The cited accounts do not establish an independent, cross-company statistic showing that AI generally shortens drug discovery or improves clinical success. The case-specific figures above should therefore be read on their own terms, rather than added together or generalized to the industry.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
