Benchling's Science Foundries puts AI agents in charge of whole R&D workflows

Benchling has launched Science Foundries, a new way for R&D teams to run entire scientific processes with AI agents doing the legwork and scientists approving the decisions that matter. The product was announced on 15 October and is available now to Benchling customers.

From single tasks to whole processes

Until now, most AI in lab software has meant a chat assistant or an agent that does one job. Science Foundries goes further. A team picks a process or goal, such as producing proteins, optimising a lead or improving a manufacturing process, then decides which steps agents run and where a scientist has to sign off.

From there, the agents carry out and monitor the work, hand tasks to one another and flag key decisions, all working from the same scientific data inside Benchling.

Benchling's own example is a high-throughput antibody programme run by four agents: one designs and scores sequences, one plans cloning and places orders with a CRO, one sends instructions to lab workcells and runs QC, and one analyses results and feeds them back into training data. In practice, an assay that finishes late at night can be analysed before the scientist is back at their desk.

Early customers and positioning

Benchling says agents are "the fastest-growing capability the company has ever released", though it hasn't published usage figures. Clinical-stage biotech Parabilis Medicines is named as an early adopter, using Foundries for protein production, screening and lead optimisation.

"We're moving into a world where scientists can run more experiments, explore more ideas, and advance more programs." Ashu Singhal, co-founder and President, Benchling

Beyond discovery, Benchling is pointing Foundries at preclinical work (study design review, CRO data processing) and process development (DoE, bioreactor data, analytical requests). Pricing hasn't been disclosed.

The bigger picture

This lands in a busy week for agentic R&D software. Revvity Signals unveiled its own AI workflow product days later, and the big AI labs are moving into life sciences directly. The pitch across the market has shifted from "AI that helps scientists" to "AI that runs the process", and buyers will judge vendors on whether they can actually get it working inside a regulated lab.