Company case
Moderna: Enterprise assistants / data analysis
What the published sources confirm
Moderna built an internal chatbot (mChat) on OpenAI's API, then deployed ChatGPT Enterprise. Custom GPTs include "Dose ID", which uses data-analysis features to review clinical data and support, not replace, the clinical team's dose-selection judgement.
Business problem and AI concept
- Problem
- Specialists across functions spend time on analysis, summarisation and document drafting.
- Concept
- Enterprise assistants with tool use (data analysis) and human review in a regulated setting.
Reported results
Moderna and OpenAI reported more than 750 GPTs created within about two months of adoption, and over 80% internal adoption of the earlier mChat tool.
Moderna/OpenAI announcement, 24 Apr 2024. Caveat: Adoption and usage counts, not clinical or productivity outcomes.
Suggested workflow
Written for learning, not the company's system. This flow shows how you could build something similar. It is not a description of Moderna's private architecture.
Step 1: Define the analysis question and approved data
Step 2: Load de-identified data into an isolated analysis tool
Step 3: Assistant applies documented criteria and produces charts
Step 4: Assistant cites the data and criteria used
Step 5: Experts review and make the decision
Step 6: Record the analysis for audit
Suggested technologies and alternatives
| Layer | Suggested | Alternatives |
|---|---|---|
| Analysis | Python sandbox with pandas and matplotlib | Managed code-interpreter tool |
| Documents | RAG over SOPs | Enterprise search |
| Governance | Audit log + access control | Validated system processes |
Small example
- Input
- De-identified dose-response table for three doses.
- Output
- Summary table, two charts, and the criteria applied, flagged as decision support for expert review.
Rebuild it yourself
Build an evidence-review assistant over public scientific abstracts that extracts claims and cites sentences.
Definition of done
- Every claim linked to a source sentence
- Missing evidence flagged rather than guessed
- Reviewer sign-off step recorded
Security and privacy risks
- Patient data exposure
- Unvalidated analysis influencing medical decisions
- Over-reliance on generated charts
Sources
Listed by title and publisher. This guide contains no external links; search for the title if you want to read the original.
- Accelerating the development of life-saving treatments (Moderna). OpenAI. Vendor case study.
- Moderna and OpenAI collaborate to advance mRNA medicine. BioSpace (Moderna press release), 24 Apr 2024. Company press release.