Moderna's AI academy was built for every employee

Two lines a company already repeats will not teach anyone a new tool. They do show what the company thinks the work is for. Dave Johnson, Moderna's chief data and artificial intelligence officer, names two of those mindsets in his account of the academy: "obsess over learning" and "digitize everywhere possible."

His post, dated 9 December 2021, is the company's own description of the Artificial Intelligence Academy, built with Carnegie Mellon University. The aim he states is to educate and empower employees at all levels to identify and integrate AI and machine learning into every Moderna system and process, so that mRNA medicines reach patients. He is not describing a specialist track with a general briefing beside it. He says the curriculum goes to all employees, "not just our executives," and he calls that an untraditional approach.

The post is an announcement of a design. The first cohort was due to start the following week, after months of planning, iterating and testing. He does not report a completion rate, a skill score or a cycle time. The forward-looking notice at the end says the statements are not promises. A people leader can still use the design, because the choices are specific, and they are the choices most academies skip.

He ties the academy to work the company had already run, rather than to a tour of tools. Investment in AI, machine learning and a digital, cloud-based, highly automated manufacturing site had helped speed the COVID-19 vaccine: 42 days from sequence identification to the release and shipment of the first batch for a Phase 1 trial. As trials grew, the company used data and predictive models, including US and Canadian census data, to compare who was enrolling with the demographics around trial sites, so the trials included communities of colour and reflected the wider society. Those are examples of AI and data already inside a process. The academy is how he proposes to make that kind of use available to 2,400 people, not a claim that the academy itself produced the 42 days.

The curriculum he names is the part worth copying onto a page. Carnegie Mellon's platform, in person and online, would deliver a programme structured for working professionals, covering data quality and data visualisation, statistical thinking and models, machine learning algorithms and AI ethics. Breadth and depth, he writes, and application, are what should set it apart. He also says the number of possible AI applications is infinite, which is why the point of the investment is to help people apply AI and focus on work that matters. An infinite catalogue is not a curriculum. A short list, tied to systems the company already runs, is.

Four questions follow for anyone about to fund the same kind of academy.

1. Who is in it. Johnson's line is every employee, not an executive cohort and not a data team alone. If the people who run a trial, a batch record or a quality step are absent, the academy will not reach the system he says it is for.

2. What it teaches. Data quality, visualisation, statistical thinking, models, algorithms and ethics is a sequence a non-specialist can be held to. A login and a prompt gallery is a different programme. Write the list down before a vendor fills it.

3. Which system it has to enter. He does not stop at awareness. The sentence is identify and integrate, into a system or process that already exists. Name one. If the academy cannot point at it, the learning has nowhere to land.

4. What you will not claim yet. This post is the design and the start date of a first cohort. Treat a later result as a later measurement. Do not borrow the vaccine timeline, or the equity work on trials, as proof that the academy worked. Those came before it.

An AI training session is a practical place to put those four questions on one live workflow, with the people who run it and the person who would own the academy in the same room.