How Beam Therapeutics cut report writing from many days to under 3 hours with Benchling AI
The slowest part of regulatory writing often isn't the writing itself. It’s all the tedious data prep that comes before it. At Beam Therapeutics, the team is tasked with producing up to nine dosage concentration reports in a two week period. After using Benchling's Deep Research Agent, Beam writes quality first drafts for regulatory reports in about 3 hours, instead of a multi-day process.
The challenge: Data retrieval took time and brainspace
For Kenton Hetrick, a regulatory writer at Beam, writing reports started with retrieval: remembering where data lived, manually pulling details from notebooks, cross-referencing study protocols, and tracking down gaps. Before a single word went on the page, he had already spent hours just assembling the raw material. That preparation work is unglamorous but critical. Under tight deadlines, the narrative and review process gets rushed, and accuracy can be compromised.
The solution: Let Benchling Deep Research handle the manual work
Kenton built a Deep Research prompt that takes a protocol, study number, and example table, and fills out the report structure for any given study. After just a few iterations, the workflow has changed how he works entirely. "It gets the first draft 80% of the way there," Kenton says. "I use it as a collaborator writing the initial draft. Now I'm not spending hours scrolling through notebooks."
Because Kenton isn't racing to assemble content, he has time to step away and come back to review with fresh eyes. He also runs finished reports back through Deep Research as a final QC check, proactively flagging issues that would have surfaced later in formal review.
Key benefits: Faster drafts, sharper review, better outcomes
First drafts in 3 hours instead of many days: Benchling Deep Research handles retrieval so writers can focus on narrative, consistency, and regulatory polish.
Data gaps caught before they become problems: Deep Research flags missing records, inconsistencies, and deviations across notebooks that are easy to overlook under deadline pressure.
Better work at the end of the day, not the middle of the night: "I can deliver at 5pm instead of 2am," Kenton says. "And the quality is better because I'm less tired and more focused on the report as a whole."
QC built into the workflow: Running finished reports back through Deep Research adds a fast review layer before submission, catching issues sooner in the process.
How Beam implemented Deep Research in practice
Kenton shared advice for teams getting started:
For repetitive reports:
Invest time upfront — it pays off quickly once optimized
Track your prompts systematically in a spreadsheet
Rate what works and document what doesn't
Refine based on actual results, not assumptions
For one-off specialized reports:
Still helpful but requires more pointed questions
Won't write detailed unique assessments perfectly
Can tell you what reagents were used and which notebooks to check
Reduces research time and provides helpful first drafts
Universal principles:
Talk to power users and copy their approaches
Keep an open mind — if it fails the first time, correct it and give it grace
Expect iteration, not instant perfection
Document what works so you don't start from scratch next time
A new baseline for report writing at Beam
Deep Research helped make Kenton faster; but more importantly, it changed the nature of his work. The hours once spent on retrieval are now spent on judgment that requires his unique expertise: building a consistent narrative, confirming accuracy, and catching what only an experienced regulatory writer would notice.
Learn more about Benchling AI and report writing. Already a Benchling customer? Visit the Deep Research help pages to get started, or get ideas from peers in the Benchling Community AI Resource Hub.
Just getting started with AI for R&D? Download our free guide — including example prompts for using AI in your everyday workflows.
