# What Can We Learn from Moderna's AI Strategy?
Source: https://humanspark.ai/learn-from-modernas-ai-strategy/

6k→100k

~6,000 people aiming for the output of 100,000

3,000+

custom GPTs built by employees

2 wks

to build mChat - 80% adoption in months

70%

leaner scientific-data management on AWS

10→5

day production cycle; 80% fewer manual errors

In early 2024 OpenAI published a case study about Moderna's use of ChatGPT - fast adoption, hundreds of internal tools. Early case studies make me cautious; the real test is what happens next. So I kept watching. It didn't fade - it deepened into structural change.

By mid-2025 Moderna had expanded its work with the R&D platform Benchling into a single "AI-ready" environment for hundreds of scientists - re-architecting the foundation of research itself, not just bolting on chatbots.

> AI is creating extraordinary opportunities in science, but realising its full potential requires entirely new ways of working.
> 
> - Wade Davis, SVP, Moderna

// key takeaways for leaders

Moderna's success comes from a CEO-led vision that treats AI as a core business function, not an IT project. They educated the entire workforce before deploying advanced tools, creating demand from the ground up. Their most radical move was merging HR and IT to manage a combined human-AI workforce. And none of it would have worked without a decade of cloud-native, integrated data infrastructure.

Their strategy rests on four ideas:

**1. Vision from leadership, not IT.** CEO Stephane Bancel frames AI as the main way to scale the company's mission - aligning everyone and securing investment. **2. Culture before code.** They educated and empowered the whole workforce (the AI Academy) before rolling out advanced platforms, creating organic, employee-driven demand. **3. Build integrated systems.** A best-of-breed mix (OpenAI, AWS, custom platforms) on a cloud-native foundation laid a decade ago - no data silos. **4. Redesign the organisation.** They merged HR and IT into a single "People and Digital Technology" function, built for human-AI collaboration.

## 01The digital foundation

Moderna's success wasn't overnight - it was a decade-long "digital-first" strategy. Their core philosophy: mRNA is information, the "software of life," so medicine development could become a repeatable platform. That demanded a fully digital infrastructure, so they went **cloud-native from day one**, partnering with AWS over a decade ago for compute, ML tooling, and analytics - standardising on AWS streamlined managing massive scientific datasets by an estimated 70%.

A **data mesh architecture** treats data as a "product" with clear ownership and central governance, instead of brittle redundant pipelines. The payoff: where many companies stall for years cleaning up siloed data, Moderna built its first internal AI tool, mChat, in two weeks. Your return on AI is directly linked to the quality of your underlying data infrastructure.

## 02Leadership, culture and people

Technology is just a tool; its value comes from the people who use it. Bancel positions AI as a force to redesign every process - the goal is scale, not cost-cutting: launch up to 15 new products in five years with a lean team.

> The vision is for a few thousand people to have the impact of 100,000 - AI making people more productive, not replacing them.

In late 2021 they launched the **Moderna AI Academy** with Carnegie Mellon (later Coursera) to build AI literacy across everyone, not a select few - in the first 20 months, 2,000+ learners logged 14,700+ hours, with a 30% average knowledge increase. Then they put tools in people's hands: **mChat** (built in two weeks) hit 80%+ adoption within months. An "AI prompt contest" surfaced the top 100 power users, who became internal champions running office hours and a busy Slack forum. The result: employees started asking for AI solutions themselves.

## 03How Moderna uses AI day-to-day

The strategy is embedded in daily work across every department - not theory, deployment.

## AI in every department

From the research engine to the unlikeliest power users - and the metrics that prove it.

R&D

Algorithms design individualised cancer vaccines (the FDA required the algorithm "locked" before the trial). Benchling unifies lab data into an AI-ready hub.

Manufacturing

A digital-first factory optimises scheduling: production cycles cut from 10+ days to 5-6, and an 80% reduction in manual error rates.

Legal

The first department to hit 100% ChatGPT Enterprise adoption - custom tools automate document review and answer policy questions instantly.

Commercial

"Nitro" (built on Dataiku) analyses unstructured medical inquiries - saving ~40 hours/month and cutting sentiment analysis from months to days.

Corporate

HR and Brand became GPT-building hubs - benefits questions, performance reviews, investor decks and brand messaging.

## 04Why custom GPTs were the game-changer

Choosing ChatGPT Enterprise in early 2024 wasn't about a better chatbot - it was a decentralised *development platform* over a centralised tool. The built-in GPT Builder let any employee become an AI developer, and that solved the "last mile":

- **Hyper-relevance:** a general AI doesn't know your jargon. Legal built a "Contract Companion" on their own templates; HR a "Benefits Assistant" on their handbook.
- **Grassroots innovation:** instead of central IT guessing what 6,000 people need, an analyst could build a "Data Formatting GPT" in an afternoon - no dev queue.
- **Ownership drives adoption:** people use the tools they built. Legal's own contract tool is why they hit 100% adoption first.

The difference between a customisable platform and a fixed tool was the difference between deploying 10 tools and sparking **3,000+**. Here are ten representative examples.

| Custom GPT | Department | What it does |
| --- | --- | --- |
| Dose ID GPT | Clinical Dev | Analyses clinical datasets to recommend optimal trial doses with rationale and charts - humans make the final call. |
| Contract Companion | Legal | Summarises contracts and answers questions about them, cutting hours of manual review. |
| Policy Bot | HR / Ops / All | Instant, accurate answers on internal policy - less admin, better compliance. |
| Earnings Prep Assistant | Brand & Comms | Automates parts of the quarterly earnings slide deck for consistent investor materials. |
| Brand Storyteller | Brand / IR | Translates biotech terminology into clear language for investors and media. |
| Self-Review Assistant | HR | Helps employees summarise accomplishments for annual reviews - consistent, faster. |
| US Benefits Assistant | HR (US) | Guides employees through annual benefits elections, cutting HR support tickets. |
| Equity Comp Explainer | HR | Explains options vs RSUs and grants - better understanding, fewer queries. |
| "Ask HR" Gateway | HR | A virtual HR router - directs questions to the right specialised GPT or a human. |
| Performance Mgmt Guide | HR / Managers | Targeted help on reviews, goal-setting and evaluation standards. |

## The rollout, year by year

From a two-week chatbot to an org redesign and an AI-ready research platform.

Early '23

mChat launches (OpenAI API)

80%+ employee adoption.

Late '23

Custom GPTs for specialised roles

Cross-departmental productivity boost.

Early '24

Company-wide ChatGPT Enterprise

Wider access; 3,000+ GPTs in use.

Mar-Apr '24

HR/IT integration begins

Key functions automated; faster operations.

Late '24

HR and IT formally merged

A digital-first organisational shift.

May '25

Benchling AI-ready research platform

Unified research data, scalable science.

## 05The radical move: merging HR and IT

In late 2024 Moderna merged HR and IT into one function - **People and Digital Technology**. In an AI-powered company, the silos separating people strategy from technology strategy are a barrier: choosing an AI agent (IT) directly reshapes roles and skills (HR). So they shifted from "workforce planning" to **"work planning"**: what work needs doing, and what's the best mix of human talent and machine intelligence to do it?

They created a new C-suite role - Chief People and Digital Technology Officer - and gave it to Tracey Franklin, former head of HR. Putting an HR leader in charge signals that the central challenge of the AI era is human-centric organisational design, not just technology. It structurally supports treating AI agents as **teammates**, not passive tools.

> It forces a level of integrated thinking that's impossible in a siloed organisation - designing the company for the next era of work.

## 06Governance, risk and ethics

Giving powerful tools to every employee is fast - and risky in a regulated industry. The core tension is empowerment versus control: empowerment drives innovation; central control offers oversight but slows progress. The risks are real - hallucinations, misuse in critical processes (trial dosing, performance reviews), algorithmic bias, and data privacy. Moderna manages them with a multi-layered framework:

- **An AI Code of Conduct** grounding responsible use in corporate values.
- **Human-in-the-loop** for high-stakes decisions - the Dose ID GPT informs, but humans decide.
- **Structural and cultural guardrails** - the HR/IT merger gives unified oversight; the AI Academy includes mandatory ethics training.
- **Proactive regulatory engagement** - informing bodies like the FDA about AI in core processes.

## 07Seven lessons from Moderna

Start with culture, not technology.

Build AI literacy before you evaluate software - an educated workforce is ready to adopt.

Secure a clear C-suite mandate.

Frame AI as core strategy tied to the mission, focused on what you can achieve - not cost-cutting.

Audit and modernise your digital foundation.

You can't build AI on a weak base. Cloud-native infrastructure is the single greatest accelerator.

Empower your people, then govern.

Give safe, sandboxed tools and encourage experiments; use prompt contests to find your power users.

Target high-value, cross-functional problems.

Build momentum on specific high-impact wins across business units.

Rethink your org chart for a human-AI future.

Ask whether traditional silos - especially HR and IT - are holding you back.

Build a human-in-the-loop governance model.

AI informs, humans stay accountable. Publish an AI Code of Conduct and engage stakeholders early.

The foundational decisions for AI success are made now.

Free for leaders: the Complete AI Toolkit - Opportunity Paper, Risk Framework, and a four-stage Adoption Roadmap. No opt-in required.

[Download the AI Toolkit](https://humanspark.ai/toolkit)