How Bayer Uses
Verisian AI
to Automate Submission Document Generation
Bayer
Customer story
"The Verisian AI infrastructure is enabling Bayer to take big steps towards delivering our data caterer vision, where at least 80% of the busywork is automated by trustworthy and evidence-based AI"
In addition to planning, programming, and validating clinical trial analyses, statistical programming functions are responsible for generating critical submission documentation to ensure regulatory compliance. The Study Data Reviewer's Guide (SDRG) and Analysis Data Reviewer's Guide (ADRG) help reviewers understand the structure, content, meaning, and contribution to the statistical results (Tables, Listings, and Figures, or TLFs) of the SDTM and ADaM datasets.
The Define-XMLs, created for both SDTM and ADaM, are machine-readable metadata files that provide a comprehensive technical map of each dataset, variable, and codelist used in the study. Taken together, high quality Reviewer's Guides and Define-XMLs are crucial components of any type of submission to regulatory authorities to deliver 100% transparency and 100% traceability, enabling faster review and reproducibility of the submission results.
These complex and large documents require a lot of time and effort to generate. They must be maintained as programs and analyses change over time. However, most organizations do not have the capacity to update documentation continuously, leading to two crucial issues:
Any inconsistencies or mistakes can compromise submission quality - increasing internal timelines, confusing regulators, and slowing the review - thereby ultimately delaying time-to-market.
Bayer and Verisian are collaborating to evaluate how AI can automate the clinical trial analysis lifecycle, starting with statistical programming first. Verisian provides Bayer with a rich Information Infrastructure powered by full study traceability, built deterministically by analyzing specifications, code, data, and other crucial clinical trial information (e.g. Statistical Analysis Plans, Study Protocols).
The Verisian Information Infrastructure, grounded in traceability, makes Verisian AI outputs high quality and reliable. Submission documentation is no longer drafted from the specifications (what was planned to happen). Instead, it is generated directly from the source of truth of the trial: the code that reads in raw data, processes SDTM and ADaM data and creates the results (TLFs), thereby guaranteeing consistency between code, data, results, and submission documentation to ensure reporting integrity. This is made possible by Verisian's code traceability, which connects all components of a clinical trial using deterministic and fully reproducible graph theory.
In February 2026, Bayer applied Verisian AI to a non-CDISC clinical trial, automating the generation of the ADaM-level Define-XML and crucial Reviewer's Guide sections for a NMPA Full Approval Submission. Cornelia Fulgenzi, Compound Lead at Bayer, spearheaded the effort:
“While the computational methods for complicated efficacy endpoint variables are not perfect and require a review, overall about 90-95% are good enough to submit as-is. The important sections in the ADRG that needed an update from the last NMPA submission, mainly those concerned with explaining dataset lineage and summarizing their inputs, content, and ultimate use, were generated at submission quality as well. Given the short timeframe and low resource availability, this would not have been possible without Verisian.”
Despite this being an initial pilot, and more complex due to it being a non-CDISC study, the drafts were completed in under 2 days. After a thorough review with only minor adjustments, the documents met Bayer's usual high-quality standards, leaving the statistical programming team submission-ready far earlier than existing processes, and are now pending final regulatory review. Once Verisian is fully integrated and the process established, draft generation will take under 5 minutes.
After completion of this first submission, the focus has shifted to the future. As Cornelia Fulgenzi puts it: “Given our initial success, we are already applying the same process for another full submission this month.” Driving the Bayer and Verisian partnership is Sascha Ahrweiler, VP, Head Statistical Programming & Analysis - Strategy & Operations, who concludes: "Rapid prototyping and agile development in 90 day mission cycles is at the heart of our new SPA organization. We all understand that our next mission is to scale this to the entire Bayer community."
.svg)
