PharmaOpinion

Novartis Shaves Six Months Off Trial Enrollment

Novartis’s Intelligent Decision System runs thousands of trial-planning simulations in minutes, with executives reporting 50% more accurate enrollment forecasts than its previous process.

Answer
Novartis Says AI Shaved Six Months Off Trial Enrollment

The brief

  1. Novartis utilized AI to reduce patient recruitment time for a drug program by six months through improved trial-site selection.
  2. The company reported a 25% reduction in time spent on authoring and reviewing clinical study reports, with 90% of report content being AI-assisted.
  3. AI simulations improved enrollment timeline predictions by 50% and are planned to be scaled to all studies by 2027.

Deployment

Company
Novartis
System
Intelligent Decision System
Stage
Live
enrollment forecast accuracy
50%

Novartis says AI helped one drug program recruit patients six months sooner by improving trial-site selection. For CEO Vas Narasimhan, those time savings are the measures that matter.

The brief

  1. Novartis utilized AI to reduce patient recruitment time for a drug program by six months through improved trial-site selection.
  2. The company reported a 25% reduction in time spent on authoring and reviewing clinical study reports, with 90% of report content being AI-assisted.
  3. AI simulations improved enrollment timeline predictions by 50% and are planned to be scaled to all studies by 2027.

Deployment

Company
Novartis
System
Intelligent Decision System
Stage
Live
enrollment forecast accuracy
50%

“I want to be more focused on outcomes than things like token utilization or number of people using an application,” Narasimhan said at the company’s investor event in London on Thursday (November 19, 2026).

The Swiss drugmaker reported shorter candidate-selection timelines and a 25% reduction in the time spent authoring and reviewing clinical study reports. Its next test is whether AI also helps researchers develop medicines more likely to succeed in patients.

Planning Trials Before Opening Sites

Max Lawson, who leads AI and Process Innovation for Development at Novartis, said the company is embedding AI across study design, execution and reporting.

Its Intelligent Decision System moves trial planning from a largely manual process that took weeks to a computational workflow capable of running “thousands of simulations about a potential study in minutes,” Lawson said. He reported that the simulations predicted enrollment timelines 50% more accurately than the previous process.

Novartis’s investor presentation describes those forecasts as potentially 50% more accurate and sets out plans to scale the approach to all studies in 2027.

The company also said 90% of clinical study report content is AI-assisted, reducing authoring and review time by about 25%.

Novartis reported 90% faster statistical programming in selected 2026 trials, with rollout to all new studies planned for 2027.

Getting to a Drug Candidate Sooner

Fiona Marshall, President of Biomedical Research at Novartis, said the average time from starting a research program to selecting a drug candidate had fallen from three to four years to less than 2.5 years. She attributed much of the improvement to AI-assisted generative chemistry.

The technology helps researchers explore molecular designs beyond the areas typically investigated in a chemistry program. Marshall described finding potentially useful binding sites on targets where earlier screening and conventional design approaches had failed, including through Novartis’s collaboration with Isomorphic Labs.

Some AI-driven programs have reached a one-year cycle time, she said, while the company continues collecting comparative data. Those measurements cover candidate selection, with laboratory safety work and human trials still to follow.

Around eight or nine molecules produced through the generative-chemistry and predictive-safety workflow are undergoing dose-range-finding toxicology studies. Marshall said their progress could provide enough evidence within a few years to assess whether the approach reduces safety-related attrition.

Making Earlier Research Useful Again

Christian Diehl, Chief Data and Digital Officer for Novartis Biomedical Research, said the company has accumulated more than 150 petabytes of research data. Its Data42 initiative has spent roughly a decade organizing records and connecting information across disease areas.

Marshall gave a practical example of how teams use that history. When a program produces a safety signal, researchers can examine whether a similar event appeared previously and whether it later proved consequential in patients. She said those comparisons have informed decisions to stop programs or continue development.

Diehl described predictive safety as an effort to identify potential failures earlier and incorporate that knowledge into molecular design, an approach he called “safety by design.”

Novartis’s technology stack combines internally developed models with external proprietary and open-source tools. Diehl said the company generally seeks to bring external models behind its firewall for training and retain versions fine-tuned on Novartis data.

Its partners span different research tasks. Alongside Isomorphic’s drug-design work, Marshall cited Relation for human-tissue data and Deciphex for digitizing slides used in safety research.

The Clinical Test Still Ahead

Novartis’s reported gains address time spent planning studies and preparing evidence. Establishing that its AI-supported discovery process improves the probability of clinical success requires following the resulting candidates through development.

Marshall said the company has identified three novel neurodegeneration targets through foundation models combining clinical and preclinical information. She emphasized that computational outputs must be tested in relevant cell-based systems and, where appropriate, animal models before progressing further.

“A lot of the things that come out just do not translate into even the cell models, let alone into the clinic,” Marshall said.

Written by Amit Raja Naik

Join the discussion

0 answers

People who run these systems answer here. Your verified role is shown, never your employer unless you choose.

Sign in with LinkedIn to answer

Read next

MachineCon USA · from AIM Media House

Where enterprise AI is decided

An invitation-led day for the CIOs, CDOs and Chief AI Officers who own enterprise AI strategy.

  • New York5th edition · 583 Park Avenue
  • Dallas1st edition

Register for MachineCon USASee MachineCon USA →

  • Keynote on the main stage, MachineCon New York 2025
  • AI150 award winners on stage, MachineCon New York 2025
  • At the tables, MachineCon New York 2025
  • Networking, MachineCon New York 2025