Novo Bets on AI Drug Discovery as GLP-1 Patents Near Expiry

AI drug discovery has become part of corporate strategy at Novo, the maker of GLP-1 obesity and overweight treatments Wegovy and Ozempic.

Novo executives now cite measurable gains in target selection, trial planning, and reporting. Yet the industry still awaits a landmark AI-designed medicine, and Novo faces patent pressure on its GLP-1s.

Image

What Does Novo Say About AI Drug Discovery?

On 22 September 2026, Novo leaders briefed reporters on how the company uses artificial intelligence in research.

Axios reported the briefing, which gave a rare look at the scale of the effort.

Novo, which recently shortened its corporate name from Novo Nordisk, cited more than 30 strategic AI partnerships.

It also employs more than 600 AI and digital specialists who work beside its scientists.

Chief scientific officer Martin Holst Lange made the central claim. Novo has found new targets, and new drugs, with AI support, he said.

He added that conventional pharmacology and chemistry still run in parallel.

That second point is actually the more important part. Novo does not frame AI as a replacement for the laboratory.

It does frame AI as an amplifier of existing science, which is a more credible stance for a regulated industry.

The hybrid model also reflects how discovery works in practice.

Computational teams propose targets and compounds. Medicinal chemists and pharmacologists then test them.

Each loop teaches the models something new, and each failure narrows the search. Scale matters in that loop, and 600 specialists give Novo the people to run it.

Why the claim deserves scrutiny

The same Axios report noted that the sector has not yet delivered a major AI-driven breakthrough. The coverage does not name an AI-derived candidate either.

R&D teams should therefore read the briefing as a signal of process maturity, not proof of a clinical win.

This careful approach protects decision-makers from hype. A company can build excellent AI infrastructure and still wait years for a candidate to clear the clinic.

Partnership counts, headcount, and tool rollouts measure effort. Approved medicines measure outcomes.

View interview How Should Pharma Companies Audit AI Vendors?

How Should Pharma Companies Audit AI Vendors?

“This is one of the riskiest industries there is because you're dealing with human lives, you're dealing with experimental protocols, and you're dealing with regulatory bodies where you might not get another shot at that clinical trial."

Image

The Milestones Behind the GLP Headlines

Novo’s investor presentation lists several 2026 milestones from its AI programmes. They span discovery and development. Each one is a distinct claim, so it helps to separate them:

• Threefold improvement in target rate: Novo prioritised genes in virtual assays to reach this gain.
• 90 per cent less time spent selecting trial sites and analysing patient populations.
• 10-fold faster generation of clinical study reports.

Discovery gains versus development gains

Two of the three gains sit in development, not discovery.

Trial site selection and report writing are operational tasks. They deliver real savings, but they do not show that AI designs better molecules. Only the target rate speaks to early research.

That specific metric deserves the closest attention, because early choices shape every later cost. A weak target can consume years of work before a trial exposes the flaw.

But, the pattern of gains is still sensible. Development tasks rely on structured, repeatable data, so AI tools perform well there. Discovery depends on biology that is noisy and incomplete. You have to transfer data somewhere between the two.

R&D teams should expect gains to arrive first where the data is clean. Trial planning, for example, draws on site histories, enrolment records, and epidemiology. Models can rank those inputs quickly and explain their choices. Early biology offers fewer such anchors.

Target selection is also becoming an ever-increasingly crowded field. Pharmatica’s Insights into target prioritisation shows how fast ranking tools are maturing. Novo’s virtual assays sit in the same movement.

What do virtual assays change?

Virtual assays test hypotheses in software before anyone reaches for a pipette. They help rank genes by their likely role in disease.

Teams then send only the strongest candidates to wet-lab validation.

If the approach holds, it saves time and reagents, and it frees scientists for harder molecule questions.

The key test is predictive accuracy.

A virtual assay earns trust only when its rankings match laboratory results again and again. Novo has not yet shared their validation record in public.

Questions to ask before copying Novo

• How does Novo define and measure the target rate?
• Which virtual assays drive the gene ranking, and who validated them?
• Does faster trial site selection change trial quality or cost?
• Do shorter report cycles shorten the path to regulators?

Novo did not publish answers to these points in the coverage. Peers should ask them before they set similar internal targets.

Without shared definitions, one company’s threefold gain may mean little to another. Benchmarks only help when the underlying measures match.

Image

Why Does Patent Pressure Raise the Drug Discovery Stakes for Novo? 

Novo faces the expiry of key GLP-1 patents in the early 2030s.

Wegovy and Ozempic carry much of the company’s growth. A thin pipeline would leave that growth exposed.

AI offers one route to a wider portfolio, and potentially a faster one.

Speed matters because patent clocks keep running. A 10-fold gain in report writing helps only if molecules reach the clinic. Novo therefore needs AI to add candidates, not just to tidy paperwork.

The next two years will show whether the discovery side keeps pace with the development side.

Investors will watch for named candidates, not slide decks. Competitors will watch for evidence that virtual assays cut the attrition that hurts pipelines.

Three partners, one platform strategy

Novo has built its AI position through several alliances in 2026:

• OpenAI (April): A partnership with OpenAI covers discovery, manufacturing, supply chain, and commercial work.
• Amazon Web Services (August): An AWS agreement created a London hub that applies agentic AI and cloud tools to discovery.
• Anthropic (September): A collaboration with Anthropic tests Claude Science for scientific reasoning and agentic software engineering.

Novo says the Anthropic work runs with data governance and human oversight.

The multi-vendor approach also limits dependence on any single model provider and it lets Novo compare tools on real tasks, from biological reasoning to software development, and move work to whichever performs best.

For a deeper look at one partner, read Pharmatica’s Insights into Anthropic’s biology lab.

What to watch next?

Three signals will show whether the strategy works.

First, Novo may name programmes that began with an AI-identified target. Second, the company may report how many of those programmes reach the clinic.

Lastly, the company may share how human reviewers check model outputs. Each disclosure would turn a promising narrative into evidence that peers and investors can test.

Image

The workforce question

Novo CEO Mike Doustdar said AI will change the company’s workforce, as reported by Axios.

Some tasks will become automated, and the skills mix will shift. His remark that Novo should be honest about this stands out, because few pharma leaders say it so plainly.

Decision-makers should plan reskilling alongside tool rollout.

Faster science still needs people who can judge model outputs, challenge assumptions, and own the results.

Companies that invest early in these skills will adopt tools with fewer errors. Those that delay may find the gap hard to close.

Training budgets, clear role definitions, and honest communication all help teams adapt without losing trust.

Regulators will ask for credibility

The U.S. Food and Drug Administration (FDA) has issued a draft guidance with a risk-based credibility framework for AI models that support regulatory decisions.

The guidance centres on the context of use and on evidence that a model performs reliably for that purpose. The recommendations are non-binding.

Scope matters here. The guidance focuses on AI outputs that inform safety, effectiveness, or quality decisions. Teams should check whether each use case falls inside it, especially as trial planning and reporting tools mature.

Early engagement with the agency also helps.

Documented validation, clear human review, and traceable data will matter whichever way the final text lands.

View interview AI in Pharma: Why the Future of Healthcare Starts With Patients, Not Tech

AI in Pharma: Why the Future of Healthcare Starts With Patients, Not Tech

Kate O’Reilly, President & Chair of the Healthcare Businesswomen’s Association (HBA) Dublin-Ireland Chapter & Healthcare Transformation Partner at Roche, discusses AI in pharma, patient engagement, and the future of healthcare innovation.

Image

What Should Drug Discovery Teams Take From Novo’s AI Drug Discovery Push?

Novo’s briefing shows where AI drug discovery stands in late 2026.

The gains are real, but they cluster in structured, operational work. The target rate is the one metric that touches early science, and it needs more detail.

Patent pressure explains the urgency, and a multi-partner strategy explains the method.

Four lessons follow for pharma executives:

• Separate discovery gains from development gains when you benchmark.
• Demand definitions for every AI metric before you adopt it.
• Diversify AI partners, and keep human oversight in every workflow.
• Prepare for regulatory scrutiny and workforce change at the same time.

At Pharmatica, we focus on the systems, strategies, and technologies shaping Pharmaceutical R&D. Our Insights connect data, Regulation, and market dynamics, so decision-makers can judge AI claims with confidence and act where impact is measurable.

Pharmatica: Insight. Connection. Impact.