How to Build a Responsible AI Ecosystem
In this episode of the HR Leaders Podcast, we sit down with Michiel van Duin, Chief People Technology, Data and Insights Officer at Novartis to discuss how the company is building a human-centered AI ecosystem that connects people, data, and technology.
Michiel explains how Novartis brings together HR, IT, and corporate strategy to align AI innovation with the company’s long-term workforce and business goals. He shares how the team built an AI governance framework and a dedicated AI and innovation function inside HR, ensuring responsible use of AI while maintaining trust and transparency.
From defining when AI should step in and when a “human-in-the-loop” is essential, to upskilling employees and creating the first “Ask Novartis” AI assistant, Michiel shows how Novartis is making AI practical, ethical, and human.
🎓 In this episode, Michiel discusses:
Deciding where AI ends and the human begins
Developing “Ask Novartis,” the company’s HR AI assistant
Upskilling HR with AI-first capabilities and new hybrid roles
Building AI governance and responsible-use frameworks at Novartis
Partnering across HR, IT, and strategy to align AI with business goals
Discover how to Build Human-Centred Workplaces which Thrive!
Workhuman partnered with Gallup to decode why only 33% of U.S. employees feel engaged and what leaders can do about it.
The answer? Recognition done right.
When companies build flexible work, psychological safety, and authentic appreciation, engagement soars.
The research shows recognition:
✅ Boosts retention
✅ Fuels feedback cultures
✅ Elevates wellbeing
✅ Powers upskilling in the AI era
Engagement isn’t chance, it’s design.
The future belongs to organizations that make recognition their competitive edge.
Speaker 1 0:00
Mitch, how you doing, my friend? I'm good. Thanks for coming by. Thank you for inviting me. How was the event going? So far? Good. I met
Michiel van Duin 0:19
already lots of people, and it was a real buzz. So I wasn't able to visit too many sessions yet, because also had one or two speaking engagements. But through that, I met quite some people. So
Chris Rainey 0:29
yeah, I really enjoyed amazing. What was your session about? Are you speaking on? Well, I did one on
Michiel van Duin 0:35
AI governance and responsible use of AI. Everyone is talking about what use cases, and we're not moving fast enough, but you also need to think through, what are the risks and what are the challenges, and how do you think upfront about what mitigations do we put in place? How do we make some of these decisions? How do we work with our partners in it, or legal or data privacy? So that was one session, and then the other session was a little bit about the team that we created, because we're in between the traditional HR and IT, so that's really interesting. So we were taking, yeah, we were kind of doing a bit of a session with Robin, who I think is here tomorrow, yeah, on exploring. Why did we create it? What did we learn? What did we achieve? So, yeah, it
Chris Rainey 1:18
was fun. Yeah. Well, let's jump in there, because you have such a fascinating role, and it's obviously evolved now, given what you just said, How is Novartis aligning its digital infrastructure with the Workforce Strategy to ensure that technology directly supports people and obviously the business goals?
Michiel van Duin 1:35
Yeah, I think I was, I was thinking about this, and I think at the at the strategic level. So what we've now been doing is we've been getting together with the Corporate Strategy Office, because that's also where we have our strategic AI team, our CHRO and myself and but, but the HR team, which in the artist we call people in organizations so and it our head of it, we've been coming together to look at, okay, but what is the IT strategy? What are the things that are coming How does it align with our people strategy? And where does the, where are we going to differentiate in terms of our corporate strategy, and therefore, where are we investing from an AI point of view, and then how can we partner better together? Because what we want to avoid is that the IT team, for example, based on the business demand, is developing all these beautiful solutions, but at the end of the day, people are not able to use them, or maybe people are worried about what will happen to their job, and therefore we don't get the ROI, or we create anxiety in the organization. But we also don't want the opposite, that we train everyone in AI, we upskill everyone, but then, for example, the technology wouldn't be there, or it's not moving in the same pace. So we've now starting to come together as a Trinity almost try to assess, okay, what are the topics that we need to discuss with the business about in your area, which jobs are going to change due to AI? When is that going to happen? So when is the IT infrastructure going to be built? And what other things need to be true in order so data, quality, skilling, rethinking about jobs, what other things need to be true in order for us to truly become AI first to the extent possible in healthcare. Do you
Chris Rainey 3:24
have, like, a framework in those meetings? Because I'm wondering, like, how do you decide what to focus on, what to not focus on? Like, what is the framework do you use for those conversations? If that makes
Michiel van Duin 3:35
sense, yeah. I mean, a lot of it is almost looking at the traditional, like od frameworks, like a star model, or does those type of things with, but then apply it with, like, what is happening in AI, and where's the disruption going to be? So, yeah, so you can really think, what is the business strategy? Because the problem with AI is also that everyone's to do something. That's what I mean. That's why I was wondering how to decide, but that's why we brought the corporate office in say, sorry, the corporate strategy offices to say, but where are we as a company going to differentiate ourselves? It was also very helpful because, for example, we're not an HR company, we're a healthcare company. So if I want to do something on AI in HR, a Plano, I'm not going to build a proprietary solution. I'm go to work day or service now, or, you know, I see what is out there. Yeah, there's loads of vendors here. Of course, yeah, would gladly help us. But if we want to do something special in R and D, which is the crown jewel of the company, or maybe in manufacturing, maybe we do want to build a proprietary solution. Yeah, because that is critical to the success of the company. It's a competitive differentiator. The data is extremely sensitive. So so that kind of looking at it from the corporate strategy lens really helps us to understand where would we make which decisions, and then looking at again. At the it like, which investments are we making, how fast are we going? But it's also helpful to look at, and that's why I said it's almost like OD, like your readiness, like what needs to be true as Rob, our Chief People Officer, Chief people and organization officer, likes to say, what else needs to be true. And then you go to, like, data, you go to process, you go to capability and maybe mindset, just looking at a framework like that, yeah, to see where are we and then determine what you need to do when, yeah,
Chris Rainey 5:33
most of our listeners are obviously HR executives and their teams. What advice would you give to them of how best to partner with IT and the business,
Michiel van Duin 5:42
I would say, when you partner with it, it's incredibly important to frame the business problem as clear as possible, and keep ownership of The problem that you're trying to solve and because it it teams are incredibly capable and want to add a lot of value, but they don't necessarily understand the business process as well as you. So if you delegate everything to it, you're not going to get the value that you're looking for. So you really need to, you need to keep ownership of what is the outcome that we're looking for. But then you need to also trust your IT team to bring you the best solution. Because, I mean, that's certainly a challenge. We run it. We get inundated with vendors, and a lot of them have fantastic solutions. But in the end, I also will need to trust my IT team to say, if they say that work day can do something, for example, or service now, maybe that is good enough, or maybe it is, as long as I understand what my outcome is, that I'm working towards the business one we're solving. I think that's where you get most out of the partnership, recognizing and that's also when we set up my team. We worked very closely with it. What are the unique capabilities of it? What are the new capabilities of my team and the rest of HR? And then, how do we then work together rather than try to get in each other's way? Makes sense, but it requires ongoing because every now and again, someone crosses over, and then you have to have a conversation. So it's hard work. You need to keep working at it. But that's, I think, how you get most out of both worlds
Chris Rainey 7:24
I've seen in the past. A lot of times, HR gets brought into the picture very late on, yes, when a lot of decisions have already been made and it becomes a real challenge. Yes. So having this partnership up front, yes is super important.
Michiel van Duin 7:37
Yeah, exactly and, and the opposite can also happen exactly that I think I've gotten, yeah, way too late. You're like, hey, business and HR, get very excited about the possibilities, but then we forget. You know, what does that mean? But I think the other thing that I mentioned it but that it's really important to think through, okay, the technology is one thing. The people upskilling is another thing. But again, what about your data? Quality, access to data, that's, I think, where the hard work is.
Chris Rainey 8:08
Where do you start on that? Because people, we hear, a few people have mentioned that during my interviews today. Where do you start?
Michiel van Duin 8:16
I think, first, of course, you start, I think, with your use case. Like, what is the problem we're solving for, or what is the thing that we try to do? Yeah, then you from there, you work back, what type of data would I need? Okay, and then that's the part that you're prioritizing. So I'll give you an example. One thing we're working on is something I think IBM is also pioneering it with ask, HR, yes, we have something we call ask P and O, and we're now working with it to convert it into ask Novartis in house build or have a vendor. We would like to work with a vendor, okay, but we might see if we do. But, but we want to go to this ask novarti So we can create a good user experience. But then we work back to say, Okay, but what are then the most common questions people ask, HR, because there's 1001 things. Yeah, but most people will ask it for payroll, time management, maybe leave performance management, those are the things that most are hiring. So if those things are not clear, if we haven't curated that knowledge, if it's not accessible, if we're not updating it, if we're not active feedback, then then this is where it's going to fall down. The experience there might be tons of other things we need to do, but maybe that's step two or step three. Yeah, if we can get decent answers on payroll related questions, like tax brackets, what do they mean? Or how many leaves do I have? Or what's the policy in this country? Yeah, again, where do I go for my check in with my manager? How do I orchestrated if we can get decent answers. Then, then, then, that's where we start, yeah. And then you start building it out to say, Okay, now, what are the next 20 questions that you get?
Chris Rainey 9:52
Issues older, right? Like, where? Where's the human in the loop? Yeah, because you don't want an AI assistant if someone has a chat. Balance of their well being. Yeah, you probably don't want that to be answered by the agent, correct. There's a balance yeah as well.
Michiel van Duin 10:10
And this is, I mean, a bit to your earlier point about how do you work well with it, and where, this is, again, where from a piano point of view or HR point of view, you need to keep control over the outcomes. Okay, I want the solution to answer these and these type of questions. Well, if this topic comes up, then it's a human. It needs to go to a human. And yeah, we I don't know how well the product is in production, but the other day, I listened to a very interesting voice demo from ServiceNow. And then exactly that happened in the demo, like I got an issue with my manager after applying for leave, but then the bot kind of recognized the situation. I said, I can't help you with this, but I know who can, and so that's the type of experience that you can create. But again, as an HR function, you'll need to draw the line, okay, this is okay for AI to answer. Here, we want a human experience, human in a loop. Plus, I think the human in the loop is also on, acting on the feedback, yeah, like making sure the knowledge is there. Then people give it feedback. This was helpful, not helpful. And then you can retrain the board update your knowledge. That's also, I think, where you need to keep the human
Chris Rainey 11:23
in the loop. Are you thinking about embedding that into your existing communication channels, like Slack or teams, so that it's one experience for people? That's the issue, right? It's like everyone, every company here, has their own agent, and you have multiple vendors internally. Yeah. So how do we make that one experience for everyone
Michiel van Duin 11:43
correct? And again, this is again, where you work then closely with it, because we might have all opinions, and we might talk to many different vendors, but then you give your requirement to it. And in this case, it's not only for HR, it's also for finance. For it, we all work with employee questions, yeah, but then it will help us assess the platform maturity capability. This is the advantage of doing it with Microsoft copilot in teams. This is the advantage of ServiceNow. Okay, which way do we go ServiceNow, and then we re evaluate and find the best solution. How
Speaker 1 12:14
is this? Have you had to build new capability within your own team? Now with this shift in terms of the skills and capability within your own team, yeah. What does that? What does that look like?
Michiel van Duin 12:25
Yeah. So we had to bring in, I call it AI and innovation director, but we had to call it, bring in an AI Director to help us evaluate use cases, to work with it, to come up with our upskilling strategy. We brought someone in for upskilling HR people on AI. And then I think where, what we're not done yet, but what we're thinking of is, do we also want to bring process experts into our team, but with an AI first capability or mindset? Because we do have process experts in the company, yeah, but have they already kind of said, Okay, can I do it? Then, how do I redesign the process completely, and then where does the human fit in? Yeah, so that capability we don't have yet, but I'm thinking we should try to probably build it in.
Chris Rainey 13:11
Yeah, that's going to be really interesting, because they're going to become, you know, AI orchestrators, yeah, in the organization, right? Yeah. And that's kind of a new role that's kind of emerging in a business, correct the role of someone orchestrating AI, you know, totally, yeah.
Michiel van Duin 13:28
And I see also a lot of in our own team, but I also see a lot of people in HR, they need to kind of adapt. Like, I mean, I'm using recruitment as a favorite example, but, and we haven't gone that road. But is it, if you're a recruiter, do you like to do interviewing? Or do you now need to think about and maybe interviewing you would still do? I'm not saying that we automate everything. Of course, there is the need for the human connection, yeah. Or do you need to think through? Is part of your job now? To think through how AI can be successful, to take some of your tasks?
Chris Rainey 14:00
Yeah, you could spend more time interviewing, yeah, because you remove a lot of the repairs Exactly. So actually scheduling, I think that's the misconception. Is like AI is going to make it less human. In many ways. It makes it more human because you remove someone a process and tasks that AI can do exactly.
Michiel van Duin 14:18
But then have you thought exactly? And have you then thought through, but how do I then make sure that AI is successful? Otherwise you start complaining, oh, the technology is not mature. Yeah. So now I still need to do the admin piece, because it well, have you spent time curating the knowledge? Have you spent time training it? Have you spent time giving it context that's now part of your job? And then the other job is, yeah, the human connection that working with the candidate, creating a good candidate experience, finding the right talent for the organization, just just to pick one example, but yeah,
Chris Rainey 14:49
before I let you go, because I know you have lunch. Okay, as you look ahead, what are you most excited about?
Michiel van Duin 14:58
I'm most. Excited about creating a better experience for people, making things simpler, meeting people where they are so not having to wait for an HR person to answer a question, but to really when you have a need or a query, whether you're a line manager, whether you're an employee, whether you want to look at your talent data, when you want to look at, you know, hiring someone or onboard when you need it, you have access to this information. That's what I'm really excited about, and for the opportunity, then, for us in piano to redefine where we add value, to really see what is it that we what we as humans are good at and can do and do better with the support of AI? That's the part I'm really excited about. I also know that it's not going to happen overnight, yeah, but it certainly
Chris Rainey 15:45
excites me amazing. Well, listen, I appreciate you taking the time out. Thank you so much. Enjoy the rest of the day, and I'll see you soon. All right, yeah, we'll see you around. Thanks. Thanks.
More from the HR Leaders Podcast
Michiel van Duin, Chief People Technology, Data and Insights Officer at Novartis.