From Insights to Action: How AI is Transforming Moderna
We're diving deep into how AI is totally transforming the way Moderna operates. Brice Challamel, the head of AI at Moderna, shares some seriously eye-opening insights into how the company has integrated AI across its processes. He emphasizes that AI isn't just about automating tasks; it's about empowering people to innovate and thrive in their roles. For instance, we talk about how AI tools are being used to connect data across various therapeutic areas, enabling scientists to learn from each other's experiments. Plus, we explore the importance of a supportive environment where employees can leverage AI to elevate their work, rather than seeing it as a threat. This conversation is all about understanding the holistic impact of AI on the future of workâdefinitely a must-listen if you're curious about the intersection of technology and humanity in the workplace.
Companies mentioned in this episode:
- FeedForward
- Moderna
- OpenAI
- ChatGPT Enterprise
- M Chat
- GPT 4.1
- Deep Seq
- Mistral
- Claude
- Gemini
- Llama
Takeaways:
- Moderna has seen an impressive AI adoption rate of 80-85%, enabling transformative uses across the company.
- The integration of AI at Moderna allows for innovative solutions in biotechnology, impacting various therapeutic areas.
- Brice emphasizes that AI is not just a technical tool, but a holistic approach to empower people within the organization.
- The use of AI fosters a culture of curiosity and creativity, leading to new ways of thinking and problem-solving at Moderna.
Transcript
Hi, this is Adam Davidson.
Speaker A:I am one of the co founders of FeedForward and your host on the FeedForward podcast.
Speaker A:I think you probably know that already.
Speaker A:This is a really fun conversation today.
Speaker A:This is Bryce Chalamel.
Speaker A:He's a feedforward member and is the head of AI at Moderna.
Speaker A:I have to say, of all the feedforward members, I think that Moderna is probably the most advanced in its use of AI, a truly integrated part of almost everything they do.
Speaker A:And Bryce is both a big cause of that and in a way, an effect of that.
Speaker A:As we'll learn, he came to Moderna specifically because they wanted to embrace AI so fully and he was only willing to work at a company like that.
Speaker A:So I think you'll get a lot of this conversation and I love how Bryce fluidly moves between big ideas and really practical implementation.
Speaker A:So here's our conversation with Bryce Shalamel.
Speaker A:Hi, Bryce, it's so great to see you.
Speaker B:Hey, Adam, thank you for having me.
Speaker A:When people ask me, you know, of, of the people I've met through FeedForward who's really using AI wisely, you're, I've told you this.
Speaker A:I'm.
Speaker A:I'm not flattering you just to flatter you.
Speaker A:Your name is one of the first I mentioned.
Speaker A:I would say the same with Ethan.
Speaker A:I want to get into some specifics.
Speaker A:So you're at Moderna now.
Speaker A:Can you walk me through ways AI is being used at Moderna?
Speaker A:How many people work at Moderna?
Speaker B:So Moderna has about 6,000 full time employees.
Speaker B:We have just now reached 5,000 active quarterly users and we are 4,000 something every month.
Speaker B:So it's slightly a floating number because there are a few thousand people working with us who don't have access to digital devices at work because they are either doing research in clean rooms or manufacturing in clean rooms also.
Speaker B:And so those I wouldn't call knowledge workers in the usual sense of the term.
Speaker B:And we need to factor them in our adoption rate because we have an 80 to 85% adoption rate at Moderna, meaning active people who use it, let's say once a week or once every other week at least.
Speaker B:And most of them use it several times a week.
Speaker B:But that is very close to 100% of the people who could use it.
Speaker B: level of adoption in January: Speaker B:And we see this as a utility by this I mean something that if you take it away from people, they're going to complain.
Speaker B:And to the same effect that if you take Internet from people, they're going to complain, or if you take laptops from people, they're going to complain.
Speaker B:And that really, what is a utility?
Speaker B:And so all this to say that beyond the level of adoption and then the proficiency and then the ways of working and we can decompose the journey from access to full AI empowerment.
Speaker B:There is a question mark because I don't know exactly how AI is used at Moderna, the same way that no one knows exactly how Internet is used at Moderna or how computers are used at Moderna.
Speaker B:I know a lot of use cases because people have bragging rights.
Speaker B:And we have a great forum.
Speaker B:We have a forum on teams.
Speaker B:It has 3,000 active users, which is half the company.
Speaker B:It's by far our most active forum.
Speaker B:And once every so often I launch a message to the effect of who wants to brag about what they're doing with AI right now?
Speaker B:And we get hundreds of answers with people who have genuine bragging rights.
Speaker B:And I interview some of the most interesting things that I see.
Speaker B:So I go to them, interview them, and I surface those use cases.
Speaker B:So I can give you a few concrete examples that 99% of the usage of AI I get either through scalable governance tools and processes that we've put in place because of the sheer number of messages and GPTs and agents, or through an idle feedback like this on the forums.
Speaker B:So I'll give you an example.
Speaker B:Let's be very concrete.
Speaker B:We have teams that work on industrial processes.
Speaker B:It's really hard because just to situate Moderna for everyone.
Speaker B:We are known for the COVID vaccine, but we are not exactly a vaccine company.
Speaker B:We are a biotech company that works on nanotechnology with the purpose of translating messages within the cellular environment to ribosomes, those little factories which produce proteins.
Speaker B:And so it's nanotechnology applied to expression of genes and proteins within the cellular environment, which can help with oncology to cure cancer.
Speaker B:It can help with rare disease and actually anything hereditary, almost.
Speaker B:And it can help triggering the immune system, which is what we do with the vaccines.
Speaker B:So what I mean to say here is that we are straight out of a science fiction book.
Speaker B:I am working today on artificial intelligence for a nanotech firm that re expresses messages within the human cell to cure cancer.
Speaker B:That's what I'm doing.
Speaker A:Crazy.
Speaker A:Like you couldn't even explain any of this to someone 20, 30 years ago, right?
Speaker B:Yes.
Speaker B:And Sometimes my children ask me, you know, how should I think about my future job?
Speaker B:I have children, they are 12 and 13, so they are at that very interesting frontier where they know what they like to do at school, but they haven't yet fully formulated an idea of how that's going to play out in active life.
Speaker B:And I tell them, do you really think that at 13 I thought I would be working on artificial intelligence for nanobiotech?
Speaker B:Because neither of those two things existed when I was 12 or 13, so I couldn't have dreamt to be doing this.
Speaker B:And this is my just way of letting them know that for now they should follow what brings them energy, what gives them joy, because that's where they're going to over invest time and gain that little extra proficiency that makes a difference.
Speaker B:And the rest is very directional because the world moves very fast those days.
Speaker A:Yeah, absolutely.
Speaker A:Yeah.
Speaker A:That's why my son who's 13 is.
Speaker A:He's not always the best at math, he's not always the most diligent, but he's so curious and he's always trying to figure stuff out.
Speaker A:And he uses AI a lot to figure things out.
Speaker A:And I think, okay, that's good.
Speaker A:That curiosity feels like one of the killer traits for the 21st century.
Speaker B:I completely agree.
Speaker B:And the curiosity and also more and more achieving with and through AI.
Speaker B:Right.
Speaker B:We have a summer project right now with my children.
Speaker B:We are coding games together on low code platforms with AI and even though they are 12 and 13, we've done some pretty amazing games.
Speaker B:And the topic that is the most interesting to them is they are middle schoolers, is their crush in middle school.
Speaker B:So we have created a role playing game which sandboxes how you approach a crutch.
Speaker B:It's called Love levels and you go levels of 10 to 10 increments from 10 to 100.
Speaker B:And it uses GPT 4.1 which is really good at this on the backend through the model platform and they've coded it entirely in loveable.
Speaker B:They've done all of it, the design, the events, the affinity matrix with characters, the illustrations, the interface, all of it.
Speaker B:This was or big summer project and now we have this amazing sandbox which is a great game to play and I love playing it where you just can choose from a gallery of Persona or make them yourself and customize yourself your crush and then it throws you into an environment which you can choose and from there you go on and the model responds.
Speaker B:It's completely open field, it will take anything you propose and respond.
Speaker B:And we Organize several calls.
Speaker B:One calls for a story building model that actually is going to propel the story forward.
Speaker B:But another one looks at how, how you're behaving and gives you feedback.
Speaker B:And that one is actually permanently giving you a grade that allows you to go through the levels or not.
Speaker B:And if, let's say you are three interactions in and for the sake of argument, if you are a girl and you have a crush with a boy and you kiss the boy after three interactions, it's game over.
Speaker B:You've crossed a boundary.
Speaker B:And the AI that looks into this, this is way too early for the kiss.
Speaker B:But if you're at 70% and you don't kiss the boy, it's also game over.
Speaker B:Because you also broke his heart.
Speaker B:Because at 70%, at some point you got to kiss the boy.
Speaker B:And that's the whole narrative they built.
Speaker B:And I share this little cute story because designing interfaces and illustrations and storytelling, which is really world building, like the environment of that middle school, the teachers, the parents, the students, their personality, the characters, their motivations, their goals and so on, then leveraging a recent model, GPT 4.1, to bring the whole thing to life within an interface that is a huge expansion of the conversational interface of Gemini or ChatGPT.
Speaker B:But suddenly, for role playing, it's been a complete blast.
Speaker B:I've had a fantastic summer so far doing this with them.
Speaker B:We are all learning a lot.
Speaker B:I'm learning a lot, they're learning a lot.
Speaker B:And we are not just getting insight out of AI, we're achieving something with AI.
Speaker B:And that will lead me to something I want to share about how I'm seeing it used.
Speaker B:At Moderna, there are really three levels.
Speaker B:The first is give me knowledge.
Speaker B:And it can be knowledge about expertise.
Speaker B:Knowledge.
Speaker B:It can be knowledge about my own behavior, like coaching or mentoring.
Speaker B:Knowledge, insights, you know, how did I behave in this meeting?
Speaker B:Based on the transcript, it can be knowledge to the effect of ideas and being a creative partner to you and yes, and you with new ideas.
Speaker B:Or it can simply be extracting insights out of data in the way that an assistant would.
Speaker B:So you have all the different, I would say the team members of AI around you.
Speaker B:Right?
Speaker B:The assistant, the expert, the coach, the creative partner around you, this team of five, including you.
Speaker B:And the first level is bring me knowledge.
Speaker B:Right?
Speaker B:AI is an insights engine.
Speaker B:The second level really is let's get things done.
Speaker B:It's AI as a work orchestrator.
Speaker B:Who are my stakeholders?
Speaker B:What's the right timeline based on all my emails and SharePoint files and documents.
Speaker B:Who the connectors that we have now.
Speaker B:How can you anticipate the right portfolio of stakeholders that I should engage with on this initiative to be successful at what frequency and then help you orchestrate work and help you get things done.
Speaker B:And then the third level, which is so interesting is make me thrive.
Speaker B:I need to travel, I'm going to be jet lagged.
Speaker B:Then I have this podcast interview when I'm back and I want to be in good shape and I'm having this relationship challenge at work or at home that I want to discuss with you or have this small health issue that is kind of nagging at me.
Speaker B:And all those things together, they're part of the overall purpose engineering that we should be doing in life.
Speaker B:Like, why am I here?
Speaker B:What is my purpose here?
Speaker B:How can I help me with that purpose?
Speaker B:And that's how you go from insights to action to really thriving across all categories.
Speaker A:I love that like learn do thrive.
Speaker A:That's.
Speaker B:That's exactly right.
Speaker A:And eventually maybe the thrive is driving the learning and the doing.
Speaker A:Once you have a better sense of what makes you thrive.
Speaker B:I love this and I think that it's a very virtuous circle because you need positive reinforcement.
Speaker A:Yeah.
Speaker B:And, and certainly getting from insight to action and then from action to a form of synergy in the things that you do that give you the feeling to have impact, to be leaving a trace, to be useful to society is as rewarding as it can get.
Speaker A:Yeah, that's awesome.
Speaker A:So let's get at what are.
Speaker A:And I, I mean the thriving.
Speaker A:Honestly, that, that to me is the, the most important.
Speaker A:I'm actually the AI tool I just created was for a group of meditation teachers, how they can structure their meditation teaching business to reach the most people.
Speaker A:I mean, these tend to be people who aren't trying to maximize revenue, they're trying to maximize impact.
Speaker A:This is a group that's highly focused on thriving, but doesn't necessarily think about the pragmatic how do I set up a system to continue to thrive?
Speaker A:You know, if I'm making too little money six months from now, six years from now, am I still going to be doing this?
Speaker B:Probably they're going through all three levels in that because they probably need some insights on their audience, you know, how it's going.
Speaker B:So for instance, then they probably need to organize things to probably think about it together or with their customers and then at some point to align all the plans of their lives.
Speaker B:You know, one of the most interesting things is that let's assume for a second that we're onto something.
Speaker B:And that was three levels of AI usage are meaningful and can be expressed in very concrete terms.
Speaker B:The world of enterprise is very uneasy with the third level, because if I start to talk about how do I organize for my children when I'm traveling for work, which is part of make me thrive, or how do I make sure that I'm in good health and fit my health appointments or my workout routine, even though I'm traveling to this thing abroad, most people in the enterprise environment, in HR or in compliance or at leadership level would be very cautious about how to guide or even influence the way that people deal with those topics.
Speaker B:And this is such an interesting thing to consider for a second, because, for instance, that third level, I have not yet come to an agreement with the AA Academy to make it part of our curriculum.
Speaker B:Because for the head of the AI academy, this is no longer work training.
Speaker B:This is personal training.
Speaker A:You mean within Moderna?
Speaker A:The.
Speaker B:I mean within Moderna.
Speaker A:Yeah.
Speaker B:So even though I express it on social media myself, you would find on LinkedIn things to that effect.
Speaker B:I have not yet had a clear mandate to discuss it within the confines of modernized AI Academy of our corporate forums.
Speaker B:And here is why I think this is a very interesting topic to consider, is that we don't work in a vacuum.
Speaker B:I have some great ideas for work when I'm in my shower or when I'm having dinner with friends.
Speaker B:And reversely, sometimes I'm hampering my ability to deliver at work because I have a personal health issue or relationship issue that is kind of taxing my mind.
Speaker B:We are not a different person.
Speaker B:It's not severance.
Speaker B:We're not a different person in and out of the enterprise entirely with this complete severance from the famous TV series.
Speaker B:And yet AI is something that is so holistic that is going to help us organize the connection between ourselves in and out of work and anywhere in between.
Speaker B:And think of this.
Speaker B:It's in the time when we are more and more working from home, when we are more and more even physically organically intertwining the personal and the professional dimensions of our lives to work together in great synergy.
Speaker B:And I think this is yet another super exciting horizon and a great challenge for AI leaders and leaders in general to consider.
Speaker B:In the age of AI, how far do I want to empower my people?
Speaker B:How far do I want to make them thrive so that they can give their best selves at work?
Speaker B:And when do I stop being comfortable with the way that AI helps them and that I'm helping AI help them if it goes too far within the realm of their personal life.
Speaker B:And I don't know that we have the answers right now, but I think this is something that we're not done talking about in the coming years or decades.
Speaker A:I mean, I think a general blurring of boundaries.
Speaker A:And I think this is typical with new technology where the kind of familiar boxes that we structure our life around start to break down.
Speaker A:And there's fascinating history, at least to me, about the development of the modern corporation, which is really a result of technology of locomotives and telephones and telegraphs, where you suddenly have to coordinate a lot more people in a much shorter period of time.
Speaker A:And human beings had never done this before.
Speaker A:And so we created the modern corporation in the early part of the 20th century.
Speaker A:But some of the logic of that remains, but some of it doesn't.
Speaker A:And I'm not, you know, I think we're already beginning to see the breakdown of traditional silos.
Speaker A:So this is interesting.
Speaker A:Like is the very boundary between the sphere of work and the sphere of, you know, it's not that we want our AI at work to tell us how to be a better parent or to give us advice on, but.
Speaker A:But thriving is clearly something that spans every minute of our days.
Speaker A:And to compartmentalize it simply as non work, it does seem extreme.
Speaker A:So that I find that it'll be fascinating how the categories of work transform.
Speaker B:And I think we always had that problem before.
Speaker B:It's just going to become more and more prominent because when we talk about work, life balance, what kind of a strange expression, as if I wasn't alive when I was at work.
Speaker B:So on one side you have work and what, I'm dead.
Speaker B:And then the other side there's life.
Speaker B:So work life.
Speaker B:And in the expression work, life balance makes no sense.
Speaker B:And the fact that we have tolerated this expression for so long is an indication of the form of mental illness that we have this schizophrenia about.
Speaker B:I shouldn't be alive when I'm at work, right?
Speaker B:I shouldn't be at work out of a dead environment with no life, no family, no health, no background, no education, no personality.
Speaker B:This is not the way things work.
Speaker B:So I think we have a life balance.
Speaker B:And this life balance surfaces at work for sure.
Speaker B:And then of course, to your point, this was there with previous technologies.
Speaker B:When you think of how people, for instance, interact with social media, most companies have social media guidelines for how people express themselves on those platforms.
Speaker B:It can come back to harm them.
Speaker B:When they do this poorly, they can become viral for the good or the wrong reason, which is always something out there.
Speaker B:And I think with AI, it's going to be.
Speaker B:Was even more important because then again, there's a lot of context that goes into those AI prompts.
Speaker B:If you're doing AI right, you give it way more information than it needs because it will surface very interesting insights out of the overflow of information.
Speaker B:And at some point, you're going to need to give information about yourself, who you are as a person, what you're about, what are your constraints, what are your boundaries, and that always going to start to include personal topics.
Speaker B:So.
Speaker B:Yeah.
Speaker A:All right.
Speaker A:But let's get back to some of the practical ways people are using it.
Speaker A:And I just want to underscore something you said.
Speaker A:So you have decided, or Moderna has decided not to peer in.
Speaker A:You.
Speaker A:You don't see what people are typing in to.
Speaker B:Yeah, it's very important.
Speaker B:We.
Speaker B:It's not only this is that we have intentionally decided to seal everything off for two reasons, and they both appeared at the same time.
Speaker B:The first is that we have a benefit assistant that works on our AI platform.
Speaker B:And that benefit assistant, if you use it correctly, you're going to share a lot of private health information or private information, period.
Speaker A:Like, I'm thinking of having a kid or I was just diagnosed with a disease or whatever it is, things you might not be ready to tell your boss or may never want to tell your boss about.
Speaker B:That's exactly right.
Speaker A:Or I want to quit this stupid job and I want to figure out what to do.
Speaker A:Like how much money I can walk away with.
Speaker B:That would be sad, but that's also a possibility.
Speaker B: enefit assistance was done in: Speaker B:Right.
Speaker B:It was done on the heels of us having access to the API to GPT4 with our own client called mchat.
Speaker B:It's the very first agent we put out there was the benefit assistant for open enrollment.
Speaker B:And immediately when we started working on it, we realized the danger of not sealing the locks.
Speaker B:And at the same time, my leadership team was all over it.
Speaker B:Because you know this, at Moderna, our leaders are very proactive in AI policy and AI expansion.
Speaker B:So our CEO, Stephane Bancel uses it all the time, our chief legal officer, our chief financial officer.
Speaker B:And we can't afford for their ideas or their questions or their prompts to go in the public because they are material information.
Speaker B:They would be pointing towards insider information, insider trade.
Speaker A:So if the CEO, and I'm really just making it up, I was like, oh, I'm thinking of acquiring this company, or we're thinking of dropping this line of business and opening another.
Speaker A:Or give me whatever it might be.
Speaker A:Yeah, I hadn't thought of that.
Speaker A:That instantly.
Speaker B:Here's the part that is the most troubling is that the worst that could possibly happen is that when my CEO is thinking to acquire another company, he can't use AI to guide him.
Speaker A:Yeah.
Speaker B:So I need.
Speaker A:You want him to be able to use AI to say, what am I missing?
Speaker A:What are some questions I should be asking?
Speaker B:What can you tell me about the valuation of this company?
Speaker B:What are the signals?
Speaker B:So for those highly critical needs, those are the ones that you should never go about out of your intuition or your common sense or even your expertise without having the super human expertise and reasoning abilities of AI by your side.
Speaker B: and more, as early as August: Speaker B:And so it is impossible for anyone at Moderna or for anyone at OpenAI to know what you are sharing with AI or to know what AI is sharing back with you.
Speaker A:So it's literally technically impossible.
Speaker A:It's not just not a problem.
Speaker B:Technically impossible.
Speaker B:We made sure that it was technically impossible on both sides.
Speaker B:It can be done because we're a discoverable company.
Speaker B:It's part of our compliance requirements.
Speaker B:So there is a secret to call it a secret, technically speaking, that is split in three parts and that is handled by our head of data engineering, our head of compliance and our head of legal.
Speaker B:And they will have to reassemble the key to be able to access the logs if we were subpoenaed to that effect.
Speaker B:It hasn't never happened yet.
Speaker B:So there is technically a way to.
Speaker B:It's very common.
Speaker B:You would do the same thing for emails or things like this, messages.
Speaker B:We have to keep records of everything.
Speaker B:It's part of our obligations.
Speaker B:So we do have a way to go back to the docs if we need to, but so far we've never used it.
Speaker B:It's been two years and it would really take a watershed moment for us to break the glass and use this because the certainty of the best privacy we can give you is a core component of how fast people have adopted AI module now.
Speaker B:And I guard it with all of my energy and my dedication against anyone who wants to question it or to reconsider it.
Speaker A:That makes a lot of sense.
Speaker A:I want to get back to practical uses.
Speaker A:But you've talked a lot about.
Speaker A:You started early with OpenAI product, what's.
Speaker B:Available now internally, now we have ChatGPT Enterprise as a horizontal platform which is available to all of our full time employees and a good number of our contractors.
Speaker B:Not all of them, but a good number of our contractors.
Speaker B:We have mchat, which was issued our client for the OpenAI API has now become a pass through which no longer has a user interface, but can orchestrate and redistribute APIs to every model out there.
Speaker B:Every model out there.
Speaker B:So we test them, we make sure they are safe, we make sure they are reasonably efficient.
Speaker A:So Mistral, or obviously Claude, Gemini, Llama, even the Chinese models as well.
Speaker B:Well, so this is a two level thing because you're talking about Deep Seq most probably here.
Speaker A:Yeah.
Speaker B:And Kamek 2, we do have an API to Deep Seq, but it's an API to a self hosted version of Deep SEQ which we've reviewed entirely because Deep SEQ is an open source model.
Speaker B:So we have here to be very, very clear that the one that is hosted by the Chinese company that created Deep SEQ is not the one we're using.
Speaker B:We're using the open source code that they published.
Speaker B:It has been reviewed through and through.
Speaker B:It took us weeks to do the full security review of that code and make sure that there was no trojan horse or backdoor to it.
Speaker B:And when we knew line by line, with all the expertise that we could have access to that this was safe, then yes, we integrated an API to Deep SEQ in our portfolio.
Speaker B:We don't use it a lot, but it's a matter for us to be always at the forefront of everything that technology can bring us because we are a company with a purpose and our purpose is to save millions of lives and we will use every technology that is safe and reliable to do it.
Speaker B:And actually the open source version of a Deep SEQ that we have self hosted and we're pointing to is to that effect.
Speaker B:And it's very important for us because that's how we can do the compliant part of AI usage in pharma.
Speaker B:So here I need to explain something for our audience.
Speaker B:In pharma, when you are working, let's say on the meeting minutes of your next off site, you can use whichever AI you want.
Speaker B:But if you're working on a standard operating procedure for manufacturing for a drug that is in production, you need to go through a system called validation in which you have to declare everything that you're doing, every little thing that you're doing, and if you change anything, it's called the deviation and it has to be published and declared to the regulator.
Speaker B:And so when you're on a normal AI platform, those models, they change all the time.
Speaker B:There's like almost daily changes of the models and so you would have such a high frequency of deviations that this would be untenable for us.
Speaker B:So we need a stable AI which we can have in a compliant environment.
Speaker B:And then we validate the use case, we document this use case and how the model has demonstrated its reliability for that use case and it becomes part of our regulated workload on AI.
Speaker B:And we can only do this with self hosted models because those are the only ones which we can guarantee are not going to change or to evolve or to be updated with our knowledge.
Speaker B:Now we could use Llama, but Meta is currently forbidden in Italy, which makes it irrelevant for Europe.
Speaker B:And that's when we did Mistral and I will say this, I think we are weeks or months ahead of a topic on model residency and model sovereignty.
Speaker B:We've heard those things on data.
Speaker B:They are not completely part of the landscape.
Speaker B:A lot of countries require data residency and data sovereignty, which means that data has to be stored in data centers in that country and handled with a capability for the government to take over if needed.
Speaker B:And I expect that we're going to see the same thing with models that, that some countries will not permit allow to use models that were not made in that region of the world or in that country for things that are highly sensitive to their geopolitical stability or future.
Speaker A:Wow.
Speaker A:So you might have dozens of models around the world wherever there's moderna staff and true for every company is that, that, that.
Speaker B:That's exactly right.
Speaker B:Like or use models that have, that have a form of sovereignty and that have a form of resiliency that works with the governments.
Speaker B:But I can tell you already that the French, Italian or German governments will not tolerate that models that were built in China or in United States and that run with data that is not hosted in their countries would operate on things that are vital to their future.
Speaker B:That's for sure.
Speaker A:Another kind of foundational question I have.
Speaker A:We have a wide range in the feedforward membership and then even more so with people at other companies that aren't in feedforward of compliance issues.
Speaker A:I was talking to some folks at a different global pharma company that essentially uses as little AI as possible.
Speaker A:I mean essentially they, they aren't allowed to use AI almost at all.
Speaker A:Some companies seem locked into copilot but getting permission for others.
Speaker A:Often we find this that the IT department or the legal department is Is are the ones that have like a risk and control focus.
Speaker A:Why did that not happen at Moderna?
Speaker B:I don't know about other companies because they're not there and I won't speak to them.
Speaker B:But I can tell you why that worked well at Moderna.
Speaker B:And I want to use what you just said, AI the way that we are speaking about it right now.
Speaker B:Because there was machine learning and neural networks, but that's not what we're talking about.
Speaker B:We're talking about the conversational interface and the ability for everyone to use it.
Speaker B:Machine learning was a five star hotel at the end of a dirt road.
Speaker B:It was almost impossible to get there, impossible to come back from it.
Speaker B:It was super hard to use it when you were there because it was its own world.
Speaker B:And generative AI in a lot of way has improved the whole thing.
Speaker B:But it has also paved the road to IT and from it.
Speaker B:And that ease of access is the revolution.
Speaker B:And because that's what we're talking about, AI is not a technical or a digital topic.
Speaker B:It is profoundly a leadership and an HR topic, a people topic.
Speaker B:So this was always a decision from the top, from the CEO that was immediately enforced by our head of hr, Chrissy Franklin, who is now head of People and Digital Technologies.
Speaker B:What most other companies have, a separate HR and IT departments, we have as one global leadership organization which oversees work and the future of work under all of its dimensions.
Speaker B:And our head of Legal was actually one of the first to have 100% adoption in her organization and shared about it in a video that we did with OpenAI a year and a half ago and that was published on their YouTube channel and then served as a source for an article in Wall Street Journal.
Speaker B:And so the fact that the CEO wants it and that the head of people and the head of legal are the fiercest champions for it immediately creates the, I would say the regulatory framework that goes with the highway.
Speaker B:Right, like the mandate to go on the highway, the speed at which you should go on the highway, which is actually a high speed in our case.
Speaker B:And that this is the right place to be.
Speaker B:That if my chief HR officer and my chief legal officer are both on the highway, and that they're going high speed on that highway.
Speaker B:If I'm not, I'm probably somewhere behind.
Speaker B:So then again, I won't speak to the other companies.
Speaker B:Every company does its own thing.
Speaker B:I would never have joined a company that would think this way.
Speaker B:By the way, I knew Stefan and the leadership team of Mudarna from my time at Google and from meeting them as part of my role as global transformation lead at Google.
Speaker B:And they were always a very forward thinking, tech driven, futuristic leadership team.
Speaker B:This is what attracted me to join Moderna.
Speaker B:So I knew exactly where I was going and that this would not be an issue.
Speaker B:And they have never ceased to amaze me by how thoughtful, diligent and relentless they were about empowering people.
Speaker B:Because we're only 6,000, but the 6,000 people are worth 100,000 in any other company.
Speaker B:And so they are all heroes in their own right.
Speaker B:Most of them are the most brilliant people that I've ever met or heard of.
Speaker B:They come from all over the world and we will give them every technology they need to achieve the impossible and to respond in as little time as possible to any new threat that we are facing.
Speaker B:So this is the spirit of Moderna and this is why it's a complete no brainer for us to deep dive into AI, to be thoughtful and secure it, but once it's secured, to not have, I would say, too many analysis paralysis problems, give ourselves a chance to see what people can do with it and then they amaze us back with what they do with it.
Speaker B:And probably that these two, because you've been waiting for these two concrete use cases and examples, right?
Speaker A:Yeah.
Speaker A:Let's get into some more.
Speaker A:I am interested in the everyday uses, but is it being used to develop, to identify new molecules, develop new treatments, think of new mechanisms of impact?
Speaker B:So that's a great question.
Speaker B:First I want to say that we are an ML native company, an a native company without machine learning there's no MRNA to the grade that we have it and we would not exist.
Speaker B:So there are more ways to code an MRNA that it was going to express one specific protein than there are atoms in the universe.
Speaker B:And so this is kind of the go game theory all over again in different field, which is now biotechnology.
Speaker B:And the way that we achieved the breakthroughs necessary for us to be reliable and efficient in producing those MRNA's, especially the single stranded MRNA's which are the only ones that are going to work and do this in the time that we do this has required machine learning and AI through and through.
Speaker B:So we are an AI native company born out of an AI breakthrough and discovery applied to biotechnology.
Speaker B:And we've never stopped using AI ever since our origination.
Speaker B:And of course we keep doing this.
Speaker B:But the thing I want to say is that around the core eureka moment of AI discovery are a thousand eureka moments of access to knowledge, of access to new ways of thinking.
Speaker B:And this is so quintessential in how we should think of AI right now.
Speaker B:So let me start and give you one concrete example.
Speaker B:There are a lot of ways that we can share experimental data in a pharma company.
Speaker B:They can be on a PowerPoint, they can be in a SharePoint site, they can be on a dedicated platform like Viva or Benchling.
Speaker B:And the truth is they live simultaneously and at various degrees of their maturity in all of those platforms together.
Speaker B:And we are a platform company, meaning we have one chemistry, nanolipid and messenger rna.
Speaker B:And that one chemistry, as I explained early on, applies to a whole different set of therapeutic areas, to rare disease, to oncology, to virology and so on and so on.
Speaker B:We have seven therapeutic areas and every time someone runs an experiment and learns something in any of those areas, they are learning something that's useful to the others because we have a single platform and it's expressed through all of them.
Speaker B:And that's very unique.
Speaker B:Every other pharma company out there has about as many chemistries as they have therapeutic areas.
Speaker B:And so they deal with hundreds of different chemical types of therapies, while they deal with hundreds of different therapies.
Speaker B:And we are very different in this because we have one chemistry for all the applications that I just described.
Speaker B:And so you would want ideally for someone in oncology to learn from a very critical experiment that was successful or unsuccessful in rare disease.
Speaker B:And if you combine those two topics, the topic that it's fragmented across a number of platforms and that it is somehow siloed by the therapeutic area, because those are different organizations, you come immediately to the point that, wait, there's a systemic challenge here on how do I give access to the right data to the right person for them to create the conditions for the eureka, right, for them to be exposed to the data that is new to them, that provides them a new perspective on things and that suddenly takes them to a different track of to lead to the discovery.
Speaker B:So we don't need right now AI models by themselves in their own self reflection to come up with a scientific breakthrough.
Speaker B:We come with scientific breakthroughs all the time.
Speaker B:We don't need AI to come up with a revolutionary new therapy, because we came up with a revolutionary new therapy.
Speaker B:It's called messenger rna.
Speaker B:What we need is to create an even better environment for the geniuses that work at Moderna to thrive and express their skills and their ability to invent net new concepts and therapies faster, safer, better, in a more productible way, in a more accessible way than ever before.
Speaker B:And in this field, we have hundreds of different AI applications.
Speaker B:One of them, my favorite, we have an AI librarian.
Speaker B:It's actually an agent orchestration platform that runs four agents.
Speaker B:One of them surfaces and indexes every piece of knowledge that's relevant and tags them.
Speaker B:Another creates abstracts out of those elements that redact them to some effect to protect the IP that might be enclosed in them.
Speaker B:A third one can organize and query them through the generation of scripted SQL queries.
Speaker B:And the fourth one is going to give you the means of permission to access or not the files that are inventory by the librarian.
Speaker B:And this is so important because you can't give the entire keys to the castle, to everyone.
Speaker B:And there you go, open all the files that you want, go and prod all the sharepoints.
Speaker B:That's not sustainable.
Speaker B:So we don't do it.
Speaker B:But you don't want either to have a thousand little small cubes of data files and data folders that no one even knew existed and that people have long lost the key of.
Speaker B:So you have to find the right balance.
Speaker B:And the librarian platform is very efficient at this.
Speaker B:We are still having debates and discussions about this.
Speaker B:This is not a technical topic.
Speaker B:How much do you want to give access?
Speaker B:How much do you want to protect data?
Speaker B:There's a fine line there.
Speaker B:And it's a political decision.
Speaker B:It's a decision for leadership, not for technicians.
Speaker B:And in this, again, AI is not a technical topic.
Speaker B:AI and the way that it shares knowledge is a political topic and it's an HR topic, and it's a legal topic and it's a compliance topic and it's an IP topic, and so on and so on.
Speaker B:Right?
Speaker B:So it's basically a leadership.
Speaker A:So if I'm working in infectious disease at Moderna and I'm trying to come up with therapies for deadly infectious diseases, I might want to know, what do our oncologists, what have they figured out that might apply?
Speaker B:You might be thinking, all right, I know of an excipient that's going to help me in the process of formulation for nanolipids.
Speaker B:That excipient specifically, has it ever been tested before?
Speaker B:And if yes, in what conditions?
Speaker B:And what is the right dosage of the excipient to be the most efficient in the way that I'm running the industry process to generate the nanolipid.
Speaker B:So you see, and that's where you're going to find, maybe through the librarian surface, the fact that there have been seven experiments, three in oncology, two in rare disease and three in virology.
Speaker B:And that those seven experiments point to the exact right dosage and also show that there is a fantastic synergy with two other excipients which you haven't thought of.
Speaker B:Right.
Speaker B:So I'm trying to not give away names or specific topics here, but these actually come from real experiences that I know of and that led to major advancements and breakthroughs and understanding from our scientists.
Speaker B:And that would not have been possible without the librarian.
Speaker A:And I know from work I've done with pharmaceutical companies, there's the bench science, and then there's what's going on among patients.
Speaker A:And in a global population, there might be wildly different kinds of treatments at, you know, major medical centers in capital cities versus rural clinics and poor countries.
Speaker A:And, you know, so there's.
Speaker A:There's all that information, like if you're creating the most beautiful treatment for an infectious disease, but it needs to have cold storage all the way through.
Speaker A:And most people who have that disease live in the developing world anyway.
Speaker A:There's all this other information that has to come in.
Speaker A:And then there's, of course, the connector, the sales operation, the marketing operation, the just thinking through the business side, this shipping and logistics, all of that.
Speaker A:So I'm beginning to see a picture of how AI could plug into all of these methods.
Speaker A:Is that the right way to think about how it's being used?
Speaker B:Yes and yes.
Speaker B:And I will tell you the wrong way it is to start by saying, okay, there's this new thing called AI.
Speaker B:Let's not give it to our people, because who knows what they're going to do with it?
Speaker B:And I don't trust them.
Speaker B:I don't believe in them.
Speaker B:And I think they're going to try to slack and find an easy way out.
Speaker B:And I think of them as children or under humans compared to me, the great leader.
Speaker B:So I'm not going to give them the technology.
Speaker B:It's most often what's really in the.
Speaker A:Back of our mind.
Speaker A:I'm going to cut just that and say that was your quote.
Speaker A:Okay, I'll just take.
Speaker A:I'm just teasing.
Speaker A:I'm teasing.
Speaker B:Yeah.
Speaker B:Because why wouldn't you empower people?
Speaker B:Like, what's wrong with you not to want to empower your own people who you pledge to support and help and nurture to the best possible extent for the mission of the company?
Speaker B:But let's say for a moment that you're that kind of person, and then you launch a pilot to donate a librarian.
Speaker B:This is dead on arrival.
Speaker B:How can you possibly have all the different stakeholders contribute and Share knowledge and insight and expertise in how to be the librarian and then use it effectively if none of them have been exposed to the overall footprint of AI and have used it hundreds or thousands of times for their own tasks, to the point that they are completely familiar and organically knowledgeable about them.
Speaker B:So running pilots before you provide the utility to your organization is a recipe for defeat.
Speaker B: I through mchat in April, May: Speaker B:Because everyone uses AI.
Speaker B:They understand what they're dealing with, they understand what it can produce, how it works, what are the limitations where it can or cannot be trusted.
Speaker B:They know how to govern, how to guardrail.
Speaker B:They will be thoughtful companions to the science team, the research team, the compliance team and the digital team to create that high value use case that everyone else is looking for through pilots.
Speaker B:And this is not specific to AI.
Speaker B:I think we could have said the same thing about the Internet or about computers.
Speaker B:If you have a utility revolution, don't go wrong piloted.
Speaker B:Because if you had a five people pilot on the Internet, you might not have seen the full value of it.
Speaker A:Yeah, I mean it brings me back to the game you designed with your kids, that doing that allows you to think about AI in a richer way and that may directly apply to your work.
Speaker A:I mean, that's something we encourage at feedforward.
Speaker A:You should play on the weekends, you should find fun ways to use it so that the only time you use it is not, oh my goodness, I have this huge project and I have to figure out how to use AI.
Speaker A:Suddenly.
Speaker B:It's so funny that you say this because that's exactly what happened.
Speaker B:So I came back with that story to the AI Academy team in a Monday morning meeting and we were having a few laughs about it.
Speaker B:And then we started thinking, what if we did a work level role playing game in which you start at 10%, your character sheet is your resume or your LinkedIn profile, your roadmap and your adventure is your job description.
Speaker B:And then the game starts at 10% and you have to go all the way to 100%.
Speaker B:And so we are somewhere on our way to making this happen at Moderna.
Speaker B:Like we are working.
Speaker B:It's for now it's a private project because we haven't share it yet.
Speaker B:We don't Know if it's going to land, how it's going to land.
Speaker B:But we are working on an adaptation of my children's game to the professional world, to something to the effect of work levels.
Speaker B:And if you had a crush on your job, what is too fast, what is not fast, how can you make progress?
Speaker B:And what takes you to be in the best possible relationship with your mission statement, with your why, with what we are expecting you to deliver and to achieve, whichever the means, really.
Speaker B:And those sandboxes are fantastic because you can try a whole lot of different experiences.
Speaker B:They come at no cost.
Speaker B:Like it's a private setting.
Speaker B:They might still reveal things that you're like, oh, that's right, I should have thought about it this way too.
Speaker B:No wonder that in my sandbox that the compliance officer doesn't enjoy this or that the cybersecurity team has second thoughts on my initiative.
Speaker B:And then you can prepare better, right?
Speaker B:Like you've kind of prototyped in a very fast, iterative, very inventive, very holistic way your various initiatives.
Speaker B:And that's even beyond get things done right.
Speaker B:It's like, it makes me think of when we were training models on billions of Go games that they were playing with themselves before they would go play with the humans.
Speaker B:Like, in a way, we are starting to train humans the same way we train models through high frequency sandboxed virtual experiences that give them an alternative set of insights from their actual experience and social life and experimental life in the company, but that can be replicated at such a higher frequency and with a much greater gamut, I would say, of possibilities and then inform you back of what's going to work or not work in ways that you would probably not dare to attempt in real life because you don't know the consequences.
Speaker A:Yeah, that's so cool.
Speaker A:And I love that it starts with being awkward at a middle school dance.
Speaker A:As the father of a recently graduated middle schooler, major life moment of asking a girl, trying to kiss her or asking her to dance.
Speaker A:How many times will you have that?
Speaker A:Three times, four times.
Speaker A:But with AI, your kids can try it over and over and over again and begin to get insight.
Speaker A:And you're saying the same thing about it work.
Speaker A:It's, you know, having these reps.
Speaker A:I love that I want to wrap up with just something that's so implicit to everything you said, you've said, but I want to just make it explicit.
Speaker A:There's obviously a lot of public conversation about AI as productivity tool, which often means as firing tool.
Speaker A:Like, let's have AI do it and let's get rid of people.
Speaker A:And there might be applications where that makes sense, but it sure sounds from everything you've said that you and the senior leadership see AI as a people empowerer.
Speaker A:And that's not just because you like people, it's because you see that as what's right for the business of moderna.
Speaker A:Can you make this idea explicit or tell me I'm wrong if I'm wrong?
Speaker B:No, no, you're absolutely right.
Speaker B:And I will say this.
Speaker B:It is, it is absolutely the wrong thing and the most dangerous thing to only understand the subtractive part of change, as in every level of automation, and we've had hundreds of Those in last 500 years, ever since the beginning of the weaving machines and the textile industry and the awakening of the industrial era.
Speaker B:There is a part of our work that is going to be offloaded and there is new type of work that is going to come to us.
Speaker B:When people say that you gain time, they are out of their minds.
Speaker B:Time is something that we all have the same exact quantity, which is 24 hours a day.
Speaker B:It is the most equally distributed wealth on the planet.
Speaker B:Every single human being has 24 hours of it every single day.
Speaker B:And the only question is, how do you use it?
Speaker B:How do you prioritize what you do with your time?
Speaker B:Our people, as in most companies that I know of, have a very hard time prioritizing.
Speaker B:They wish they could do everything and sadly they can't because there are so many hours in the day.
Speaker B:24.
Speaker B:And so they make choices.
Speaker B:And a lot of those are hard choices.
Speaker B:The menial tasks that can be automated were never the reasons why we were hired to begin with.
Speaker B:There are things we have to go through in order to get to the higher added value tasks in our environment.
Speaker B:In pharma, we are not out of problems to solve, we are not out of diseases to treat, we're not out of viruses to cure or to therapeutic areas to nurture.
Speaker B:We don't have an understanding and a solution for every ailment out there.
Speaker B:Some of them are even beginning to scratch the surface off.
Speaker B:So the notion that we would help people with their arbitrations of priorities by taking away a little bit of the things that just prevent them from doing more and empower them to then explore more options, be more creative, innovative, thoughtful, dedicated in delivering their mission seems so obvious to me that I am in complete.
Speaker B:You know, I've already in some form of self denial on why they're even that conversation.
Speaker B:How did people come to it?
Speaker B:What world do they live in?
Speaker B:Because it's.
Speaker B:Take a moment and pause.
Speaker B:We have a lot of problems, a lot of challenges.
Speaker B:We have climate change, we have demographic change.
Speaker B:We don't have too many doctors, we don't have too many teachers, we don't have too many engineers.
Speaker B:Far from it.
Speaker B:How is it possible that you think that if you give your people more power, more intelligence, more acumen, or even a mirror to reflect on their behavior or to have novel ideas that they are going to turn their back to you to start slacking at work, to become useless, and that you're going to end up firing them or having to pay them just for breathing out of kind of a universal income.
Speaker B:It baffles the mind.
Speaker B:So I just want to say this to everyone who hears this.
Speaker B:Love your people as much as we love our children, Adam and I.
Speaker B:This is the same thing, right?
Speaker B:Apart from the fact that they are our equals and our peers, and we're all adults in this room and all smart and intelligent and out of decades of work experience.
Speaker B:Love your people, empower them, and let yourself be amazed with what they are going to do with the tools you give them.
Speaker B:With AI in particular, you can definitely use the assistant to cut down on some of the tasks.
Speaker B:You can absolutely use the mentor and the coach to help guide you as a person.
Speaker B:You can definitely use the expert to elevate your understanding of domain expertise.
Speaker B:You can use the creative partner to think differently about topics and to engage in a growth journey for yourself in another way.
Speaker B:None of those things point for me obviously towards mass firing or mass work displacement.
Speaker B:Certainly some skill set will be less useful than others in years to come.
Speaker B:Suddenly, some jobs will change in nature and yes, a few jobs might just go away, as they always have.
Speaker B:A few jobs will be created.
Speaker B:Let me end up with one concrete example.
Speaker B:I train executive assistants at Moderna.
Speaker B:When we had the connectors, they asked me, can we use the connectors to Outlook to reorganize the schedules of our leaders?
Speaker B:And actually the connectors don't allow for this.
Speaker B:They're more allowed to create a knowledge base from Outlook, right?
Speaker B:To leverage all your meetings, their attachments, your emails, and pull from that knowledge to query and to gain insights or to orchestrate work.
Speaker B:So once we had come to that conclusion together, the executive assistant started thinking, oh, I have new ideas of what to ask.
Speaker B:And then they started asking things to the effect of how has my leader been perceived through social media, out of their press conferences and their earning calls in the past months.
Speaker B:And they got A trove of information out of all the image exchange plus all the web search through the connectors that both surface, internal and external knowledge.
Speaker B:And then out of this, the workshop completely transitioned to another type of surface, which was how can I make my leader more successful with the way that they are messaging themselves and their work as their assistant?
Speaker B:And I literally saw during that workshop, the executive assistants start to migrate towards chief of staff or campaign managers for their leaders.
Speaker B:And in that experience, it was not very long ago, because connectors have only been there for a few weeks, I suddenly saw the future of work take shape, that someone whose most frequent task was to organize your calendar was now starting to think that maybe they can help you thrive as your campaign manager.
Speaker B:Because as a leader, you're a politician.
Speaker B:And I say with a capital P, you have to understand the perspect perspectives of your shareholders, of your employees, of the media, of the public, of the patient.
Speaker B:And you have to make sense of all of this and guide the company in the right direction.
Speaker B:And now the people supporting and helping them are elevating into a whole new dimension of what their role could be.
Speaker B:And this is for me, the perfect illustration of we never get too much automation.
Speaker B:We sometimes have not enough imagination.
Speaker B:And if you let the people whose role is in full screen and evolution explore what the tool can do and what they can do with this, and do this with their decades of experience, at some point are going to amaze you with the way that their mind is thinking about this.
Speaker A:I love the story.
Speaker A:Yeah, I love the story so much.
Speaker A:I mean, it's an example of something I think about a lot, that if the main thing you think of is there's some work we do now, let's have AI do it, like scheduling.
Speaker A:And sure, that'd be great to make scheduling easier.
Speaker A:Every AI tool I've used right now confuses me.
Speaker A:It doesn't work well on scheduling, but that's fine.
Speaker A:But they thought of a new thing.
Speaker A: are locking your company into: Speaker A: You're saying that in: Speaker A: at's a new thing we can do in: Speaker A: an you're doing new things in: Speaker B:And that's a combination of both.
Speaker B:You're exactly right.
Speaker B:I would say, because you need to do the other thing, too.
Speaker B: o the effect of let's package: Speaker B: And then let's embrace: Speaker B:Yes.
Speaker B: The part of: Speaker B:And here we are in 20 years, back to the beginning of our conversation, living the lives of science fiction novels, doing artificial intelligence for nanoparticles that cure cancer.
Speaker B: is with the work processes of: Speaker B:Those 6,000 heroes that we have at Moderna with the power of AI, they are now superheroes.
Speaker A:Well, that is a beautiful place to end.
Speaker A:As with so many feedforward conversations, I could truly talk to you forever, Bryce, and I'm going to want you back soon.
Speaker A:There's so much more I want to ask, but I feel like this was a lot for one podcast.
Speaker A:It was amazing.
Speaker A:I want to ask about what's the future of organizations.
Speaker A:I want to ask about more about this human innovation interface, but we'll do that later.
Speaker A:Thank you so much, Bryce Chalamel, and you are active in the discord, so if people want to reach out to you, I know you're very responsive.
Speaker A:You love having conversations with folks who are thinking about the same things.
Speaker B:Of course.
Speaker B:And thank you so much for having me.
Speaker B:I'm so grateful for the chance to be a member of FeedForward.
Speaker B:I have a passion for what everyone in this organization does, and I can't wait to have more frequent and more profound interactions with all of you.
Speaker B:And let's go ahead and be optimistic.
Speaker B:We need this.
Speaker B:It's a design component of who we are.
Speaker B:And start paving the way to the future for people who are not as lucky as we are and who are maybe gripped by fear or by uncertainty and it's its own Darwinian advantage.
Speaker B:But let's have a positive look on the future.
Speaker B:Those conversations, they are the way that we do our own little sandboxing of ideas and share concepts and share limitations.
Speaker B:And so I'm so grateful for you too, Adam, for having the conversation and thank you for welcoming me into it.
Speaker A:Yeah, well, I so love talking to you every single time.
Speaker B:All right.
Speaker A:Au revoir.
Speaker A:Bryce is a fairly active member of our Discord community.
Speaker A:You will see him at most of our live meetings.
Speaker A:And Bryce told me he's quite eager to talk with other members.
Speaker A:So please feel free to reach out.
Speaker A:If you want a personal introduction, just let me know, or Jessica or Maddie, and we'll be sure to put you in touch.
Speaker A:And also let us know if you have things you think we should share on the podcast, if you'd like to talk about, or are there other subjects you think would be worth sharing.
Speaker A:As always, I'm Adam Davidson with the FeedForward podcast.
Speaker A:I look forward to seeing you on the Discord.
