Lire en français

$250 billion were invested in AI last year. In pharma, the market alone is expected to grow from over $4 billion in 2025 to $25.7 billion by 2030 (McKinsey). And yet — drug makers have seen neither clinical development accelerate, nor success rates improve.

So is AI failing pharma?

No. It is simply not delivering where the spotlight is pointed. The major breakthroughs in molecule discovery are still ahead of us. What AI is already transforming is the part no one showcases: the operational engine room — regulatory, technical writing, quality, vigilance.

At the JP Morgan Healthcare Conference, AstraZeneca, Roche and Pfizer described the same reality: thousands of pages of clinical, safety and manufacturing documents to compile, verify and keep consistent across geographies — work that AI is now shaving weeks off (Reuters). Teva’s CEO Richard Francis calls it “the unsexy stuff” — the process work that actually moves the needle.

And the concrete results are appearing exactly there.

Sanofi aims to become “the first pharma company powered by AI at scale”: its plai platform supports GxP quality decisions, its generative tool cut Product Quality Review writing time eightfold, and the company created a dedicated Generative AI Regulatory Lead role. Novo Nordisk, for its part, built NovoScribe, an automated authoring platform for clinical study reports, first-human-dose documents and stability reports — with around 70% faster production.

What I keep coming back to — living this from the regulatory and compliance side — is where the value actually materializes: medical writing, quality systems, pharmacovigilance, submission preparation. Far from the moonshot that captures imagination.

But it only works under three conditions, none of which the tool provides: clean, structured data; clear governance; and humans firmly in the loop. The tools are impressive — but nothing is autonomous, and many of these gains are still early or only partially quantified.

So the real question isn’t whether AI will discover the next blockbuster. It’s whether our data and our processes are ready for it to help us file — faster, and better.

That, to me, is where responsible leadership meets real value creation.

Where do you see AI delivering first in your organization — in the science, or in the operational engine room?