Moderna is a Cambridge, Massachusetts-based biotechnology company that has built its operating model around messenger RNA (mRNA), a technology it has used to develop vaccines and therapeutics for infectious disease, oncology, and rare disease. The company is best known for Spikevax, its COVID-19 vaccine, but it now markets three approved products and is advancing a pipeline aimed at up to 10 additional product approvals through 2027.

Moderna reported approximately $1.9 billion in revenue for 2025, down sharply from prior years as COVID-19 vaccination rates normalized, while research and development spending totaled roughly $3.1 billion. The company employs around 4,700 people, a workforce it has explicitly chosen not to scale to match its product ambitions. Instead, Moderna executives have stated that they intend to bring up to 15 new products to market over five years — a goal they argue would require a headcount in the hundreds of thousands under a traditional operating model.

That gap between ambition and headcount is the throughline connecting Moderna’s AI strategy. The company has invested in artificial intelligence both to multiply the output of its existing employees and to compress the scientific timelines that have historically made drug development slow and expensive.

Moderna is not unique among large biopharma companies in pursuing this path — Merck, Pfizer, and Sanofi have each disclosed comparable internal generative AI deployments in recent years — but Moderna’s data-first origins and its willingness to publish detailed adoption figures make its program one of the most thoroughly documented in the industry.

This article examines two of Moderna’s internal AI use cases:

• Enterprise-Wide Generative AI Deployment: Scaling knowledge work productivity across legal, research, manufacturing, and commercial teams without proportional headcount growth.
• AI-Driven mRNA Sequence Design: Compressing the design, testing, and optimization cycle for new mRNA vaccine and therapeutic candidates.

We begin by examining how Moderna applies generative AI to address enterprise-wide productivity constraints.

Enterprise-Wide Generative AI Deployment

Moderna’s leadership has been direct about the scaling challenge behind its AI investment: the company wants to launch far more products than its current headcount could historically support. The company employs roughly 4,700 people today. Executives have argued that bringing 15 products to market under a traditional model could require a workforce on the order of 100,000 employees.

That left the company with a straightforward strategic question: how do you multiply the output of a relatively small, specialized workforce across functions as different as legal, clinical research, manufacturing, and investor communications?

Point solutions for individual departments would not scale quickly enough, and Moderna’s leadership wanted a tool flexible enough for each team to adapt it themselves rather than one built and maintained by a central IT function alone.

The pressure was compounded by industry-wide caution around generative AI. A survey of more than 200 life sciences professionals found that roughly half of life sciences companies, including a majority of the largest pharmaceutical businesses, restricted employee use of consumer chatbot tools like ChatGPT even as a majority of individual workers were already using them informally.

Moderna’s leadership treated that gap between restriction and informal use as a governance problem to solve rather than a reason to avoid deployment altogether, which shaped the security and data-residency requirements it built into its eventual rollout.

Moderna’s approach began in early 2023 with mChat, an internal chatbot the company built on OpenAI’s API in about two weeks — a pace made possible by data infrastructure Moderna had already standardized on AWS for years. After mChat reached strong internal adoption, Moderna’s AI team ran comparative user testing against Microsoft Copilot and OpenAI’s ChatGPT Enterprise before selecting ChatGPT Enterprise as its enterprise-wide platform.

The distinguishing feature of that rollout is the custom GPT Builder, which lets any employee create a specialized assistant trained on their own team’s documents and terminology without engineering support. Within about two months of launch, employees across the company had built more than 750 of these custom GPTs. Examples include a “Dose ID” GPT that uses ChatGPT Enterprise’s data-analysis features to evaluate optimal vaccine dosing against standard clinical criteria, a “Contract Companion” GPT that summarizes legal documents in plain language, a “Policy Bot” that answers employee questions about HR and safety policy by searching internal documentation, and GPTs used by Moderna’s brand team to translate technical or regulatory material into language suited to investors and the public.

Governance was built in rather than added afterward. Because ChatGPT Enterprise does not train its underlying models on customer data or conversations, Moderna was able to give employees latitude to build their own assistants while keeping proprietary research, contracts, and manufacturing data out of any externally shared training set. The company paired that technical safeguard with mandatory AI ethics training and internal review requirements for any GPT that touches clinical trial design.

The workflow change is less about replacing tasks and more about giving every function its own AI-built specialist:

• Legal teams use GPT assistants to summarize contracts and answer policy questions instead of manually searching through documents.
• Clinical teams use the Dose ID GPT to analyze large trial datasets and generate a documented rationale for dose selection, with a human team still making the final call.
• Corporate communications teams use dedicated GPTs to prepare earnings-call materials and translate scientific findings for non-expert audiences.
• Individual employees, not just IT, build and maintain their own GPTs, which keeps the tools tied closely to real day-to-day workflows.

Moderna reports that each ChatGPT Enterprise user averages 120 conversations with the platform per week, and that 40% of weekly active users have built at least one custom GPT.

This use case is mature and enterprise-wide rather than experimental. Moderna’s legal department has reached 100% adoption of ChatGPT Enterprise, the highest of any function in the company, and platform access has expanded from an initial rollout reported at around 3,000 employees to deployment across legal, research, manufacturing, and commercial functions company-wide.

Video:  OpenAI: Moderna partners with OpenAI to accelerate the development of life-saving treatments

Moderna has not published a single audited productivity figure tied to dollar savings, so the evidence of impact here is adoption-based rather than an outcome metric: consistent usage across nearly every department, a fully converted legal team, and continuous expansion of the GPT library, rather than a pilot that stalled.

Moderna’s own framing — that AI lets the company scale like a company of one hundred thousand with its actual headcount — is a purported benefit rather than an audited result, but the underlying adoption numbers support that the tool is genuinely embedded in daily operations rather than symbolic.

AI-Driven mRNA Sequence Design

Long before generative AI became an enterprise mainstay, Moderna’s core scientific bottleneck was speed: designing, testing, and optimizing an mRNA sequence used to take significant time because each candidate had to be reasoned through and validated largely by hand.

For a company whose stated advantage is treating mRNA as a programmable, information-based molecule rather than a biological material to be discovered through trial and error, that manual bottleneck undercut the platform’s entire premise. As Moderna has expanded from infectious disease vaccines into oncology, rare disease, and individualized cancer therapies, each new modality multiplies the number of sequences that need to be designed and tested.

The problem is strategically important because Moderna’s business model depends on running many modalities at once rather than betting the company on a single drug. Traditional pharmaceutical R&D typically takes 10 to 15 years per candidate and can cost billions of dollars, with the vast majority of candidates failing before approval.

Moderna’s argument to investors has always been that treating drug design as an information-processing problem, rather than a wet-lab discovery problem, changes those economics — but that argument only holds if the design tooling can actually keep pace with a growing, diversified pipeline.

Moderna’s answer is a proprietary system it calls the mRNA Design Studio, part of a broader internal platform the company refers to as its Scientific Intelligence Engine. The Design Studio’s Sequence Designer module takes a target protein — any protein in the human proteome, or a novel designed protein — and automatically converts it into an initial optimized mRNA sequence, which scientists then refine across the 5′ untranslated region, coding region, and 3′ untranslated region. The system runs on cloud computing infrastructure built on AWS, enabling Moderna to design multiple mRNA candidates in parallel rather than one at a time.

Beyond sequence design itself, Moderna applies machine learning models to the large volumes of preclinical and animal-study data its programs generate, looking for patterns such as the relationship between mRNA dose and immune response, and using those patterns to help flag which candidates are worth advancing. The company describes this as a system that learns continuously: data generated in one program is meant to improve the algorithms used for the next one.

Moderna’s leadership has specifically pointed to individualized cancer treatment as the modality that stresses this system hardest. Its Individualized Neoantigen Therapy program requires designing a distinct mRNA sequence for each patient based on the specific mutations found in that patient’s tumor, which means the design pipeline has to run at production speed for what is effectively a one-patient batch size.

For Moderna’s scientists, the workflow shift shows up at both ends of a program:

• At the design stage, a request for an mRNA construct targeting a chosen protein starts as an automatically generated sequence rather than a blank page, with scientists tailoring the details.
• At the analysis stage, machine learning models sort through large preclinical datasets to surface which formulations or dosing regimens performed best.
• The system integrates directly with Moderna’s automation platforms, so an approved design can move into the company’s high-throughput production facility without a separate manual handoff.
• Individualized programs, such as personalized cancer therapies, depend on this automation because each patient’s treatment requires its own unique sequence on a compressed timeline.

This use case has moved well past the experimental stage and has a well-documented, if now several years old, proof point: during the COVID-19 pandemic, Moderna’s platform completed the sequence design for its vaccine candidate mRNA-1273 within two days of the viral genome being published, and the company delivered its first clinical batch to the National Institutes of Health for a Phase 1 trial 42 days after that initial sequencing. That timeline compression remains Moderna’s clearest evidence that its digital design, automation, cloud computing, and machine-learning infrastructure can accelerate the movement from concept to clinic.

Since then, Moderna has extended the same underlying platform beyond infectious disease into oncology, rare disease, and individualized neoantigen cancer therapies. The company frames this as a “virtuous cycle,” where data from manufacturing and clinical programs continuously feeds back into the design algorithms. Moderna has not disclosed audited cycle-time figures for its current, more diversified pipeline, so the infectious-disease result should be read as proof of concept for the platform’s ceiling rather than a guaranteed benchmark across all therapeutic areas.

Video: AWS Summit Online ANZ | Powering Moderna’s digital biotechnology platform to develop new vaccines

This analysis examines the following lessons enterprise leaders can draw from Moderna’s AI adoption:

• Bottom-Up Tooling Beats Centralized Mandates – Moderna’s fastest-adopted AI tool succeeded because employees, not a central IT team, built the specific assistants their departments needed, keeping every tool tied to a real workflow rather than a generic capability nobody used.
• Prior Digital Maturity Is What Makes Speed Possible – Moderna could build and roll out mChat in two weeks and its mRNA design tools at production speed because it had already standardized its data and cloud infrastructure years earlier; companies without that foundation should expect AI initiatives to take much longer.
• Adoption Metrics Can Substitute for ROI Figures, But Only Temporarily – In the absence of audited dollar savings, sustained usage data such as 100% legal department adoption or 120 weekly conversations per user is a legitimate signal of real embedding, but leaders should still push toward measurable outcome metrics as programs mature.