Dec 16, 2025
What a year! In October, we closed our $12.1 million funding round and shared our vision: build an intelligent regulatory workflow that connects documentation, data, and decision-making. Two months later, so much has already moved forward, not just for Peer AI, but across the industry.
This past year has been about proving it. Proving that agentic AI can transform regulatory documentation. That expert-in-the-loop design delivers better outcomes than automation or humans alone. And that the industry is ready to move from experimentation to production.
It’s also been a year of learning and surprises along the way. Times where customer feedback influenced product redesigns, not because the vision was wrong, but because production reality is much more demanding than any demo.
Now, with these learnings and proof in hand, I couldn’t be more optimistic about what lies ahead. Here are my reflections on where we’ve been in 2025 and what’s next.
The rapid shift from pilots to scale
A year ago, our customer conversations focused on whether AI could reliably produce regulatory documents. Today, customers aren’t asking if AI works, but how quickly they can deploy and scale it across their organizations. They’re evaluating AI with more rigor, demanding proof of ROI, performance, quality, and speed metrics. The industry mindset has quickly shifted from AI demos to real implementation at scale.
The hardest part of scaling is rarely the technology. It’s governance, validation, change management, and building trust across teams from medical writing, regulatory, QA, and IT — all while keeping auditability intact. It requires full organizational buy-in and collaboration.
We saw this shift firsthand at the CNS Summit this fall, where we presented alongside Biogen on how they moved from pilot to production for AI-powered documentation. Their experience reflects a broader industry pattern: companies are no longer just running experiments. They see a clear runway to broad AI adoption across regulatory documentation and other critical workflows. Moving from pilot to production isn’t a switch you flip. It means rethinking reviews, defining escalation paths, and aligning stakeholders on what ‘good’ looks like – repeatedly.
Every industry is deep in the AI hype cycle. Lots of claims, but not much behind them. But this is changing fast, and we expect this momentum to accelerate in 2026.
We hold ourselves to a transparent measurement framework built on four dimensions: accuracy, completeness, readability, and consistency. We assess every document against these standards and share the framework with customers, because we believe the industry needs benchmarks, not promises.
Our customer results speak for themselves. Time savings of 55% to 94% while improving document quality. Clinical study report drafting dropped from 40 to 17 working days. Protocol turnaround went from 6-8 weeks to one week. These aren’t projections, but real metrics from customers in production.
Regulators move from cautious guidance to action
Perhaps the most significant development this year has to do with global regulators’ support and active AI adoption.
The FDA deployed ELSA for drug protocol reviews and CDRH-GPT for medical device evaluation. In December, they announced agentic AI capabilities for all FDA employees such as tools that plan, reason, and execute multi-step actions with human oversight. By adopting AI internally, while mandating provenance and traceability, the FDA is redefining compliance. EMEA also published reflection papers on AI use and together with the HMA developed a joint workplan on data and AI in medicines extending through 2028.
Regulatory momentum is a strong signal, but not a blanket guarantee. Sponsors still have to earn trust with evidence, traceability, and rigorous controls, and expectations will keep evolving. Auditability has evolved from a quality metric to a fundamental requirement. Transparent data lineage is now a prerequisite for approval.
All this creates urgency for sponsors from both directions: regulators are proving they can use AI responsibly, and boards are demanding efficiency and productivity gains. Companies can't afford to wait. The question now is how to implement AI across workflows with measurable impact, and keep pace with an industry that is already moving.
Experts matter more than ever in the AI era
Nearly every industry, including life sciences, talks a lot about “human-in-the-loop.” We’ve learned this year that not all loops are created equal.
Generic human oversight to review data and outputs often misses the mark. In multi-agent workflows, early errors in the process compound downstream and degrade final quality. What matters most is bringing in human expertise at the right moments: shaping document structure and clinical interpretation upfront, not just reviewing the final draft.
This is why we built Peer AI with medical writers in-house, not as an afterthought but as a core part of how we develop and deploy AI. Our product was designed by medical writers. They lead training and onboarding. They work peer-to-peer with customer teams navigating this transition.
The results validate this approach. We’ve watched skeptical medical writers become champions after hands-on experience with the platform. When an AI tool genuinely requires and values their expertise, even the most hardened skeptics become advocates.
Our director of medical writing Dr. Neel Sheth puts it best: “the role of the medical writer is evolving from writer to reviewer.” This doesn’t diminish their role, it elevates it. In practice, it means medical writers spend less time on first-draft assembly and more time on judgement calls that shape whether a document succeeds: clinical accuracy, regulatory positioning and narrative clarity. We capture the knowledge of medical writers into our product itself so they can focus on providing the judgment and expertise that AI can’t replicate.
Agentic architecture plus the right team
When we started Peer AI, we made a bet on agentic as the future. Our architecture deploys purpose-built agents working together with human oversight at strategic checkpoints.
This year, the industry rapidly advanced toward agentic design patterns, validating the architectural approach we’ve taken from day one.
Advances in reasoning models boost what automation can accomplish. But outputs remain non-deterministic. In practice, that means designing for failure: fallbacks, disagreements between agents, ‘looks right but isn’t’ drafts, and clear rules for when the human checkpoint must interrupt the workflow. Building effective AI requires that medical writer expertise, regulatory requirements, and the realities of how documents actually get created and reviewed all need to be built into the solution.
Success also requires a rare combination of skills: people who understand both AI and regulatory documentation, and can translate between both worlds. We’ve grown the team deliberately, prioritizing these crossover skill sets and hiring people who are as comfortable discussing transformer architectures as they are ICH guidelines.
This work is hard. There is no playbook for applying AI to regulatory workflows. There is no “right” answer. Every day requires rapid iteration, testing, feedback loops and even a level of unparalleled creativity. We’re building a culture where AI engineers and medical writers work together daily because it’s the best way to build products in a new era of technology that truly delivers value for our industry.
Building the infrastructure for what’s next
In 2026, we’ll scale what we’ve proven. More customers, bigger projects, new document types, and further progress toward an intelligent regulatory workflow.
Regulatory success in pharma depends on scientific coherence across multi-year submissions. The industry still has a long way to go. First-cycle approval rates dropped from 35% to 25% over the past decade. Internal inconsistencies were cited as a primary deficiency in 35% of FDA Complete Response Letters (CRLs), prompting the FDA’s push for greater transparency. These coherence failures cost the industry an estimated $8 million per day in delays.
A new approach is needed. We’re building a scalable regulatory workflow that maintains real-time coherence across sprawling submission programs, transforming a significant risk into a competitive advantage. The systems we’re developing now will become the infrastructure for regulatory documentation, with broad applicability beyond any single document type. Our vision has always been to speed regulatory approval, and the documentation layer is where that journey starts.
The tailwinds are undeniable. The FDA is aligned on AI adoption. The technology is maturing. And agentic methods are showing true value and promise.
Ending the year with gratitude
This past year simply would not have been possible without belief from our customers. And for that, we’re deeply grateful. Every contract is a partnership. Your transparency and willingness to learn alongside us has shaped what Peer AI is becoming.
We also owe tremendous appreciation to our investors and advisors. You saw what we saw: agentic is the future and improving how drugs reach patients is worth pursuing with urgency.
Finally, thank you to the entire Peer AI team for believing in our mission and choosing to work on very hard problems that matter for our customers and the industry. I’m grateful to build this company alongside each of you, as peers.
Here’s to a great 2026!
Anita Modi
Co-founder and CEO
