Integrating AI into life sciences supply chains presents new opportunities and challenges. While many companies recognize AI's potential, there's often uncertainty about where and how to begin implementation, especially in highly regulated environments.
We recently sat down with Jennifer Chew, Vice President of Solutions and Consulting at Bristlecone, to explore practical approaches to AI adoption in life science supply chains. Jen, whose work focuses on solving supply chain challenges for Global 2000 companies with particular emphasis on life sciences organizations, shares insights on everything from tactical starting points to strategic transformation.
This discussion breaks down the often overwhelming topic of AI implementation into actionable steps, offering guidance on use case prioritization, data management, and organizational readiness in the context of life sciences supply chain operations.
Apple Podcasts | Spotify | YouTube | Web + Others
Summary, Key Points, and Practical Takeaways
This interview has been edited for clarity and length.
Nick Capman: What's the current state of AI adoption in life sciences supply chains, particularly after COVID?
Jennifer Chew: Every organization was hit during COVID-19 in ways that no one really outside, and maybe some risk managers, had really thought through and prepared for. But life sciences had an even bigger challenge — they had to help us solve the COVID crisis while employees were at home, isolating. The stress on life sciences organizations was just tremendous — hopefully once in a generation or even less frequent than that. As we've come out of the crisis moment, organizations are still not totally settled back in. There are still studies going on around whether hybrid should be the model. Should everybody come back to work? There are a lot of CEOs out there who believe everybody should be back full-time in the office, so there's still a little bit of settling down around what the "new normal" really is going to be. We compound that complexity with AI, and we now have this new capability being presented to us, especially with ChatGPT coming on and making it so much more accessible to not just the IT organization but to everyone, whether you're a consumer or a middle manager or an intern or the CEO.
How should companies approach implementing AI in their supply chain operations?
First of all, we encourage everybody to think through a use case orientation. There's a lot of experimentation happening across your organization, sometimes whether or not the IT department wants you playing with AI or not, and it's been sort of a ground up technology. By forcing a sort of use-case mentality, it helps you prioritize as an organization where you can begin. You want to think about the complexity, and you want to think about the strategic nature of the use case. Things that are simple and tactical, or what we could refer to as intern work, are probably easy to write a chatbot for and automate, and it's a good way to learn. Things that are complex and strategic are probably not ready yet for AI. They may be your most significant bang use case, but not in 2024 — maybe after you've got a few years of learning under your belt from your organization and once you've gotten things like your data cleaned up and the right data made available, and once you've started to meet the organizational challenges. We also encourage people to think about multiple use cases within an organization so that they can build up some muscle memory and some capability around how you deploy and leverage AI, as opposed to a scattershot all over the organization.
What are some immediate, tactical ways that supply chain professionals can start using AI today?
Everybody buys stuff, right? So we're all procurement experts, even if it's in our own consumer mind. So let's start with procurement because I think any of your listeners will be able to track this one. If we think about some of the tactical activities that we do all the time in procurement, we send out an RFP, and we get a lot of data back — why not have your digital assistant review those answers, structure the responses, and give a first pass grade to that RFP? That's a pretty easy way to get started. You can also use AI to help find vendors, though I would move that up a notch — a bit more strategic. If you're just saying, "Go find me everybody," like if it's the equivalent of an old Yellow Pages lookup, then that's probably not strategic. But most strategic sourcing operations are far more in-depth than an internet search — there are conversations, RFIs, and RFPs. In logistics and distribution, calculating alternate routes or sourcing out of different distribution centers is another good use case. If you see a hurricane coming in Florida and you've got a big facility in Orlando, maybe for a few days, you want to temporarily work around that Orlando facility because they're going to be busy with other things. Those are great logistics-specific redeployment, on-the-fly activities you can do with these bots.
Why is data so important in AI implementation?
If you're not feeding it the right data set, then how, in a regulated environment, could you feel comfortable going to the FDA and saying this process is valid? You know, yeah, I have 100% confidence that the answer I am giving you or the document I am signing off on is robust. So you've got to have, especially in the regulated industries like life sciences, you have to have an outstanding and rigorous and robust data set from which to run your strategic sourcing operations, or for which to run your S&O P scenarios, or for which to run your logistics replanning and your cold storage calculations. And it's generally probably not data that's willy-nilly out there on the internet. So, I think these organizations will have to spend a lot of time thinking carefully about how to build and train the data sets on which they want to ask their McKinsey questions.
What's the vision for how AI could transform supply chain planning processes like S&OP?
The real big one that we want to be able to get to is all of the S&OP planning. That's what happens in a lot of S&OP sessions. We've got people from finance, people from sales, people from supply chain, and people from manufacturing. Very expensive, costly resources. We pull them out of the business for several days. They sit in a conference room, and they run planning. That's hugely complex and hugely strategic. If we can get to a point where those same resources already have several scenarios that have been pre-run and say, what if this happens? What if this happens? What if this happens? What if this happens? And then we're left to respond to those scenarios with specific advice and guidance. Then we can get that S&OP down from a two or three-day session to a three or four-hour session. Maybe it can finally all be virtualized because a lot of these companies are still getting together at headquarters and running that, although we are seeing hybrid S&OP. So I don't want to, like, the whole business case isn't on avoiding travel, but to go from three or four days of S&O P preparation and execution down to three or four hours of preparation and execution because we've offloaded so much of that scenario planning onto a well-trained bot — that's what will make it work because we need the answers will hopefully be better, right? We'll get a more optimized set of results and do it in less time.
Looking ahead to 2025, what should companies be doing now to prepare?
Massive amounts of data projects are in the near and medium term horizon, because doing any of these AI calculations on top of bad data just gets you bad answers faster. So, I think the data and analytics piece is just skyrocketing. And so if you don't already have a project to optimize your data, to clean up your data, to determine whether or not you even got the right data to answer the types of questions that you want to be able to ask your digital supply chain assistant then you're behind. And so I think data will be king in the next 6-12-18 months. The kernel of the large language models that you referred to earlier is broadly trained through this OpenAI or through other large organizations, and that's good. Still, everybody's got access to all that data. So, how do you stand out as a large life sciences company? You have to append your own unique data. Now there's industry-specific data, the rules and regulations, and then there's your organization's unique data, so that sort of Bullseye from most strategic, the smallest amount, but the most important, all the way out to the internet, open AI data. It takes that combination of company, industry and common data to get us to the right answers when we're running our super duper S&OP process one day. But that core of data becomes your organization's gold mine.
How can companies assess if they're ready to implement AI more strategically?
Well, I would say everybody needs to be planning for it, whether or not they're putting a massive team in place and tackling a dozen different use cases across different parts of the organization or whether or not they're starting with one. But almost every Global 2000 that we work with is doing something in AI today, and so I'd say, if they haven't started thinking about it, they're behind. But I also would bet that there are people in the organization who have been experimenting on the side, even if they don't know about it. The data shows that I think only 2% of Global 2000 companies had locked off all access to ChatGPT, all AI. Everybody else has some degree of limited access, even if it's to an enterprise version that's not allowing things to get thrown out onto the general internet. So, I think most large companies have at least started to tiptoe into the water. Now, whether or not they build a strategic roadmap, that's probably the early adopters who've sat down and put a method to the madness and said, "Here's how we're going to start. We're going to build here. We'll learn from this, and then we'll slowly expand."
What's your final advice for companies looking to start their AI journey?
Don't be afraid to get started. Everybody will stub their toes here at the beginning, and that's part of the learning process. But I would discourage anyone from just waiting and see what shakes out because AI is here, and AI is getting adopted. It's just a matter of how fast your company will be able to take advantage of this new capability. I hear a few companies say, "I'll just wait for SAP to figure it out" — that's too late. We'll say is too little, too late. You will be out-competed. That's not to say that SAP will not bring some very valuable capabilities and build-in stuff that you don't have to worry about. Some basic transactional processing may not be some of your first use cases, but I would always recommend that you build some organizational capacity to learn about AI and not just wait for SAP to do it for you.
Jen's key takeaways:
Start with Simple Use Cases: Begin with tactical, straightforward applications like RFP review automation or basic logistics calculations. Build organizational confidence and capabilities before moving to complex strategic implementations.
Data Quality is Foundation: Clean, robust data is essential, especially in regulated environments. Life sciences companies must have rigorous datasets that can withstand FDA validation. Without proper data, AI implementations will only produce "bad answers faster."
Think Departmentally, Not Company-Wide: Focus AI implementation within specific departments rather than scattered across the organization. This builds "muscle memory" and organizational capability more effectively than a dispersed approach.
Maintain Human Oversight: Keep humans in the loop for complex and strategic decisions. This is crucial for regulated industries and will remain important for the foreseeable future.
Transform Core Processes Gradually: Major transformations, like reducing 3-day S&OP sessions to 3-4 hours, won't happen immediately. Plan for a multi-year journey as organizational capabilities and data quality improve.
Build Unique Organizational Assets: While general AI models are available to everyone, competitive advantage comes from combining industry-specific data, organizational knowledge, and common data. Your organization's unique data becomes your "gold mine."
Don't Wait for Enterprise Vendors: Waiting for solutions from vendors like SAP will put you behind competitors. While vendor solutions will be valuable, organizations need to build their own AI capabilities and understanding.
Expect and Accept Early Mistakes: Everyone will "stub their toes" at the beginning. This is part of the learning process. Don't let fear of mistakes prevent you from starting the AI journey.
Develop a Strategic Roadmap: Early adopters succeed by methodically planning their AI implementation, learning from initial projects, and expanding strategically rather than haphazardly experimenting.
Recognize the Timeline: While 2024 might not be ready for the most complex AI implementations, organizations should be actively preparing through data cleanup, organizational development, and tactical implementations.
The FDA Group helps life science organizations rapidly access the industry's best consultants, contractors, and candidates.
Need expert help planning and executing compliance projects or connecting with a quality and compliance professional? Get in touch with us. If we haven’t worked together yet, watch our explainer video below or head to our company introduction page to see if we could help you execute your projects now or in the future.
Our service areas:
Quality Assurance | Regulatory Affairs | Clinical Operations | Commissioning, Qualification, and Validation | Chemistry, Manufacturing, and Controls (CMC) | Pharmacovigilance | Expert Witness
Our engagement models:
Consulting Projects | Staff Augmentation | FTE Recruitment | Functional Service Program
Our podcast:
Apple | Spotify | YouTube | Web + Others
Share this post