English

Getting more from every study: AI and the future of in vivo research

Andrew Smith Solution Consultant - R&D at Benchling

Each animal study represents a significant investment, scientifically and ethically, and so we need to maximize the insight we get from it. With the 3Rs and changing policy reducing animal numbers, this matters more than ever. AI can play a valuable role, but only when the underlying data is well-structured, connected, and traceable.

This series has tracked AI’s arrival in the in vivo community. At the American Association for Laboratory Animal Science (AALAS) 2024, AI was a prominent theme on the agenda for the first time. At Federation of European Laboratory Animal Science Associations (FELASA) 2025, the conversation turned to the data quality, ethical frameworks, and infrastructure AI would need. This year at AALAS 2026, the questions became practical — where does AI fit in a real study, how much of its output can be trusted, and what does it take to roll it out in a regulated facility?

Benchling’s 2026 Biotech AI Report surveyed around 100 biotech and biopharma organizations already using AI, including teams working with in vivo models. 89% of respondents said copilots or reasoning tools are now their first stop for interrogating and synthesizing data. However, the same report found that data quality is the number one reason AI pilots fail.

Across sessions, the exhibit hall, and conversations with delegates, three themes stood out: AI is moving into everyday in vivo workflows, data foundations decide whether it delivers, and welfare and compliance need to be built in from the start.

Why general-purpose AI falls short in in vivo research

AI conversations at AALAS 2026 were more concrete than in previous years. Attendees wanted to see specific tasks handled well such as drafting and reviewing study protocols, monitoring live studies to gain insights and alerts, structuring results returned from CROs, and summarizing completed studies for reports and presentations.

I spoke with delegates who had used generally available AI tools  and got poor accuracy and strange results, often hallucinated findings. This could be due to these tools lacking scientific awareness and not understanding in vivo work or its terminology.

The tools gaining traction work inside a scientist’s existing workflow. The same pattern shows up in the 2026 Biotech AI Report, where the most widely adopted uses, such as literature review (76%) and scientific reporting (66%), succeed because they fit naturally into daily work and produce output that is easy to verify.

Data foundations decide whether AI delivers

At FELASA 2025, I wrote that without structured, high-quality data, AI is a powerful engine with no fuel. That idea came up repeatedly this year. Facilities hold a large amount of valuable information, including protocols, fully executed studies and reports in different formats, but much of it still sits in paper records, spreadsheets, and disconnected systems. AI can only reason over data it can reach and trust.

Outsourced studies add another layer. Many delegates described CRO results that still arrive as PDFs and spreadsheets and have to be re-keyed before they can be compared with internal studies. Teams that have standardized how study data is captured, with consistent structures linking animals, treatment groups, samples, and results, can now ask cross-study questions and get answers they can rely on quickly.

How do we ask better questions with fewer animals

The AALAS community’s commitment to the 3Rs (refinement, reduction and replacement) shaped every AI discussion. Delegates were enthusiastic about two areas. The first is Refinement such as spotting welfare concerns earlier through continuous monitoring and trend analysis. The second is Reduction  — making better use of existing data, including understanding which studies and compounds have been run in the past, so we don’t replicate work and  so fewer animals are needed. Attendees want AI as a tool to support them in this regulated environment.

The policy backdrop makes this more pressing. In April 2026, the FDA published a one-year progress report on its Roadmap to Reduce Animal Testing, which included qualifying its first AI drug development tool for use in clinical trials. NIH has also asked that new funding opportunities include consideration of new approach methodologies (NAMs). At the same time, as FASEB noted in its response to the FDA report, animal models remain necessary across many drug classes and disease areas. For in vivo teams, that raises the bar: every study needs to be well designed, fully documented, and able to get the most from the data it generates.

The requirements delegates raised were consistent. People need to stay accountable for decisions, AI actions must be recorded in the audit trail, and every output must trace back to its source records, and research data is not used to train models.

Watch AI-assisted in vivo research in action

During the conference, I presented a session on AI-assisted in vivo research and real applications. The aim was to move past the hype and show what AI looks like across real study designed in Benchling In Vivo, where it supports protocol authorship, data capture, and downstream analysis, and where researchers stay firmly in control.

AALAS 2026 showed that AI in the vivarium has moved from promise to practice. The teams getting the most from it pair new tools with structured, connected data and an unwavering commitment to animal welfare. Benchling shares those goals. We’re building In Vivo so that study data is captured once, connected across R&D, and ready for AI from day one.

Thank you to everyone who joined our session and shared their experiences with us in Houston. We’re looking forward to continuing these conversations, and to doing it all again in 2027 at FELASA in Cologne and AALAS in Charlotte .

Curious about how AI agents are at work within in vivo? Read three ways AI helps in vivo research teams, or explore Benchling In Vivo and Benchling AI.

From the bench to your inbox
Our monthly newsletter features science insights, industry best practices, and stories from teams pushing biotech forward.

Powering breakthroughs for over 1,300 biotechnology companies.

Helix Image