Moderna: 750 internal GPTs in two months and 120 AI conversations per employee per week

• Moderna and OpenAI partner to accelerate the development of life-saving treatments — OpenAI (customer story), 2024-04-24
• Collaboration with OpenAI: Transforming the Way We Work and Innovate Through AI — Moderna (корпоративный блог), 2024-04
• Moderna banks on OpenAI to accelerate mRNA research (~3 000 сотрудников на ChatGPT Enterprise, штат ~6 000) — Pharmaphorum, 2024-04-26

Background

Moderna is a biotech company that has worked at the intersection of science, technology, and health for more than a decade; its mission is mRNA medicines, and its best-known product is the COVID-19 vaccine. The strategic ambition is stated bluntly: bring up to 15 new products to market within five years — from an RSV vaccine to individualized cancer treatments.

Meanwhile the company deliberately stays lean: Moderna employs a few thousand people (publications around the partnership cite roughly 6,000 employees — Pharmaphorum), whereas, per CEO Stéphane Bancel, by traditional biopharma standards such ambitions would require about a hundred thousand. The company intends to close the gap between ambition and headcount with technology — which makes its case an unusually clean experiment: AI here is not 'optimizing costs' but a load-bearing element of the operating model.

Importantly, Moderna arrived at generative AI prepared: it spent the previous decade building its own tech stack and data platform (earlier technology partnerships include, for instance, quantum computing with IBM for mRNA research). The company has worked with OpenAI since early 2023: first the internal chatbot mChat on the OpenAI API, adopted by more than 80% of employees, then a full move to ChatGPT Enterprise, formalized as a public partnership in April 2024, when about 3,000 employees gained platform access (Pharmaphorum, citing the WSJ). Bancel framed the philosophy plainly: 'We believe very profoundly at Moderna that ChatGPT and what OpenAI is doing is going to change the world. We're looking at every business process — from legal, to research, to manufacturing, to commercial — and thinking about how to redesign them with AI.'

The talent side of the strategy is public too: in its own blog the company stresses that embedding AI into everyday work 'requires more than technology — it necessitates a strong AI-focused culture,' and points to labor-market recognition — 9th on LinkedIn's 2024 Top Companies list in the U.S. and 2nd among healthcare companies. For this case that is not decoration: the ability to hire people ready to work 'AI-first' is part of the same operating model.

Problem

The goal was set hard and measurable: 100% adoption and proficiency in generative AI by all employees with access to digital tools — within six months. By corporate standards that is an extreme bar: typical rollouts settle for 'licenses issued, let's check activity in a year'.

The key risk of any enterprise rollout is that the tool gets bought and people don't use it. In pharma the risk is amplified by the industry's specifics: heavy regulation, the high cost of errors in clinical data, and the conservative culture of functions like legal and regulatory. Add the classic employee fatigue with 'yet another transformation' — and it becomes clear why Moderna treated the AI rollout not as an IT project but as a change management program with a dedicated team of experts.

There was also a technology fork. By the time ChatGPT Enterprise launched, the company already had a successful mChat on the OpenAI API with 80%+ adoption. The question was posed like adults pose it: keep developing the in-house tool, adopt Microsoft Copilot, or give everyone ChatGPT Enterprise? For a science-centric company the answer could only be experimental — 'as a science-based company, we research everything,' says Brice Challamel, Moderna's Head of AI Products and Platforms.

Solution

Moderna ran platform selection as an experiment: Challamel's team performed extensive user testing of mChat, Copilot, and ChatGPT Enterprise. NPS decided it: 'We found out that the net promoter score of ChatGPT Enterprise was through the roof. This was by far the company-favorite solution, and the one we decided to double down on,' Challamel says. Betting on measured user preference rather than brand or an integrator's discount is this case's first signature move.

The rollout ran as a three-level change program. Individual: research and listening programs, training in person, online, and with AI learning companions — 'using AI to teach AI was key to our success.' Collective: a prompt contest whose top 100 power users were structured into a cohort of internal Generative AI Champions; office hours in every business line and geography; an internal AI forum that grew to 2,000 active weekly participants. Structural: engaging the CEO and executive committee through leadership meetings and town halls, incentive programs, and events with internal and external experts. The program's philosophy, as the OpenAI case study puts it: collective intelligence means 'everyone together, everyone with a voice and nobody left behind.'

The role of mChat as an 'early win' deserves its own note: the chatbot launched on the OpenAI API in early 2023 gave the company a year and a half of practice, data on real usage scenarios, and 80%+ of employees with basic prompting skills — before the question of buying an enterprise platform even arose. Then the construction-kit effect kicked in: once employees could easily assemble their own GPTs, they started building. 'We were never here to fill a bucket, but to light a fire. We saw the fire spread, with hundreds of use cases creating positive value across teams,' Challamel describes. The examples stand out for their variety. Dose ID is a pilot assistant for the clinical team: via Advanced Data Analysis it analyzes clinical data, verifies the optimal vaccine dose selection against standard criteria, provides a rationale with source references, and builds charts; final review remains human-led. 'Dose ID has provided supportive rationale for why we have picked a specific dose over other doses. It has allowed us to create customized data visualizations and helped the study team analyze the data from multiple different angles,' says Meklit Workneh, Director of Clinical Development. Contract Companion produces readable contract summaries for any function; Policy Bot answers questions about internal policies without digging through hundreds of documents; the brand team runs a GPT that prepares slides for quarterly earnings calls and another that translates biotech terminology into approachable language for investors ('What would my mother want to know about Moderna, versus a regulator, versus a doctor?' is how Chief Brand Officer Kate Cronin frames the task).

The legal department reached 100% adoption — a rare figure for a conservative function; Chief Legal Officer Shannon Klinger explains the effect simply: the tool 'lets us focus our time and attention on those matters that are truly driving an impact for patients.'

Result

Within two months of adopting ChatGPT Enterprise, the company had 750 GPTs, 40% of weekly active users had created their own GPTs, and each user averaged 120 ChatGPT Enterprise conversations per week — roughly twenty-five interactions per working day, meaning the tool is genuinely embedded in the workflow rather than opened 'for the record'. The legal team reported 100% adoption. The very fact that the case is measured in usage metrics rather than 'projected savings' sets it apart from most corporate announcements.

Important framing. All figures are Moderna's self-reported numbers published in an OpenAI case study — material from interested parties (OpenAI sells ChatGPT Enterprise; Moderna demonstrates technology leadership); there is no independent audit. Dose ID is a pilot explicitly positioned as an assistant to human-made decisions, not an autonomous dose selector. No dollar-value financial impact has been disclosed, and the link '750 GPTs → 15 products in 5 years' remains declarative: the product plan will be tested by clinical trials and regulators, not by the number of chatbots.

In our view, the case's main substance is not the numbers but the method: Moderna showed that 'AI adoption' is 20% platform choice (made, tellingly, via an in-house NPS experiment) and 80% a change program with contests, champions, office hours, and an engaged CEO. The '120 conversations per user per week' metric is the best publicly available indicator that the program worked: licenses can be issued by decree, habits cannot.

A second observation: the pyramid '80% mChat adoption → three-platform NPS test → 750 GPTs' illustrates a sound investment sequence. A cheap API prototype built the skill base and user-behavior data before the enterprise purchase; the platform choice leaned on that base; the mass of GPTs grew on an already prepared culture. We would not expect the same numbers to reproduce at a company that started by simply buying licenses.

Lessons learned

1. Choosing the platform via internal comparative testing (mChat vs Copilot vs ChatGPT Enterprise on NPS) ends ideological debates: data decides, not brand loyalty.
2. 'Conversations per user per week' is a more honest metric than 'licenses issued': 120 weekly conversations means the tool is embedded in work, not a checkbox.
3. A cheap early prototype (mChat on the API) with 80% adoption built the skill base before buying the enterprise product — the migration met no resistance.
4. A prompt contest is a practical way to find internal champions: the top-100 power users became the core of practice diffusion, with the 2,000-participant weekly forum as its carrier.
5. 'Using AI to teach AI': AI learning companions scale training where no headcount of human trainers would suffice.
6. In a regulated industry the AI assistant is positioned as support for human-led decisions (Dose ID provides rationale and references; the team decides) — a deliberate legal and ethical frame, not a weakness.
7. When reading vendor case studies, separate usage metrics (750 GPTs, 120 conversations — measurable) from strategic declarations (15 products in 5 years — to be tested by clinics and regulators).

Frequently asked questions