Is pharma ready for AI? Beyond the hype in clinical trials

AI could cut clinical trial timelines by up to 50% – so why isn’t pharma moving faster? 

In this episode Sebastian speaks with Annette Østergaard, a recognised leader in the pharmaceutical industry, specialising in clinical development and digital innovation. Drawing from years of hands-on work in global clinical trials, Annette shares a grounded perspective on where AI can bring real value – and where the industry still faces significant barriers. 

The discussion covers the complexity of clinical trial processes, the critical role of data quality and availability, and why regulatory, ethical and organisational challenges cannot be overlooked. Annette also explains why the real transformation is not only about technology, but about people, mindset and the ability to rethink how clinical operations are designed end-to-end. 

Key topics: 

• Why AI has the potential to significantly accelerate clinical trials 
• What stands in the way of effective AI implementation in pharma 
• Why data quality, structure and accessibility are critical foundations 
• How regulations, privacy and ethics shape AI adoption in life sciences 
• Why an end-to-end approach is necessary instead of isolated solutions 
• How AI can improve predictability, reduce costs and increase success rates 
• Why people, skills and organisational mindset are the biggest challenges 
• How pharma companies need to rethink collaboration and operating models 

⏱ Timestamps: 

 00:00 – Introduction 

01:41 – Is pharma truly ready for AI? 

04:28 – Key challenges: data, regulation and complexity 

10:40 – Why mindset and skills are the biggest barrier 

12:07 – How AI could cut clinical trial timelines by up to 50% 

15:53 – Can AI become the breakthrough for pharma? 

20:35 – Key benefits and challenges of AI in clinical trials 

🎙 Guest: 

Annette Østergaard, Pharmaceutical Industry Leader (Clinical Development & Digital Innovation)
LinkedIn: https://www.linkedin.com/in/anetteoestergaard/ 

🎙Host: 

 Sebastian Dzieniak 

LinkedIn: https://www.linkedin.com/in/sebastian-dzieniak/ 

Transcript

SPEAKER_02 0:00

We found out that we could cut something between 40 to 50%. That is this constant risk of failing very late. If you want to have benefits from AI and machine learning, you need to have your data readily available.

SPEAKER_00 0:16

Welcome to Beyond Delivery by Holistical Connect, where tech meets true business value.

SPEAKER_01 0:25

Welcome to another episode of Beyond Delivery. Today we're asking

Introduction

SPEAKER_01 0:30

a big question: Is the pharma industry truly ready for AI? To help us navigate this landscape, I'm joined by Annette Oestergaard, an experienced leader in clinical development and digital innovation. Annette has spent her career driving global clinical trials operations, digital transformation across the life sciences sector. She brings a rare combination of scientific expertise, leadership experience, and hands-on understanding of the realties of clinical RD. Together we will explore what AI can really solve, what stands in its way, and whether Pharma is prepared for this next era of development. Hello Annette. Thank you so much for joining me today.

SPEAKER_02 1:18

Hello, Sebastian, and thank you very much for inviting me to have a talk about what is happening with AI in pharma these days. And I'm looking forward to our conversation.

SPEAKER_01 1:33

Thanks so much again. It's an honor to have you here. Let's start very broad. Is pharma ready for AI?

SPEAKER_02 1:40

Pharma

Is pharma truly ready for AI?

SPEAKER_02 1:41

is indeed ready for AI. And we will see a lot of things ongoing the next couple of years for sure.

SPEAKER_01 1:52

Okay, that's that's good to hear. It's just that AI is pretty much everywhere these days.

SPEAKER_02 2:14

Because in pharma it takes a long time to develop products. There's a long way from early discovery to early development to doing clinical trials, which can take years and before you get everything ready to do submissions to the authorities and have your product in the market. We know this is something that has taken between 8 to 15 years, more closely to 15 years than to eight years. So it's something that has taken a long time. It is very expensive, and there is this constant risk of failing very late in development and especially clinical trials. I think we've all seen how news are coming out. Now there was phase three results, and it did not meet the target of where you want it to be. So overall it's said to be a failed trial, and that late in the process, which is already years down the uh the road of getting to the market, it is extremely expensive to fail products. So there is a big challenge for pharma that I think we will have to overcome, and this is where AI and robotics, machine learning, uh all the different new technologies, uh, advances within biology, physics, and computational capacity and power will have a big impact.

SPEAKER_01 3:55

By AI, we mean it's not just AI, it's the machine learning, automations, right? All the technology around it as well. Yeah, so that's that's quite an interesting uh perspective. And because you did mention uh some of the challenges that come around it as well. And perhaps you could elaborate a little bit on that. Um, what are the key challenges that you see in the implementation of AI?

Key challenges: data, regulation and complexity

SPEAKER_02 4:28

I would primarily speak about my own area, which is clinical development and running clinical trials. With AI, you need very good data that is available and structured and so on, but also when it comes to clinical trials, we need to be very clear on um regulations and ethics and uh legal requirements within handling of data. So, of course, all the privacy, the uh GDPR and uh and and the likes also outside of Europe has an impact on how structured you need to be around where to use data and and how to apply it, also outside the specific consent you have for a specific trial. But I do think that's something that we can come around, but it's just that's one thing we have to be aware of that may be different from other other data that can be used for AI. So there's the data privacy, there's the regulations, there's of course the notorious bias that you can have in any algorithm. So you have to be aware of that. So when you're going to use data where you have used AI in the first place, you also need to have some transparency to how did you generate the data, where did they occur from originally, and and how is this? So you need to have a traceability of everything that you're doing, which might be slightly different than from other places. But again, this is something you have to be aware of. It's a challenge, but it's not something you can't overcome. It is something that you can work around, but you just have to be aware of it when you're working in this area. Where I think another challenge when it comes to the clinical landscape is how we set together uh the setup of clinical trials, where we have the pharma industry also called the sponsors uh of the clinical trials, but you also do have big variation of service providers, which you can uh set up in different models. You can have, I mean, partly data and ownership on the sponsor size, or you can have the CROs, the contract research organizations doing everything with their systems and their processes and so on, or you can have a combination of these. This is something that makes the complexity a bit higher because if you want to have benefits from AI and machine learning, you need to have your data readily available, and you also need to be able to show it in a structured way so that you can use that to monitor all the process from end one end to the other. So I think that's that's a level of complexity where you need to have a very uh top-down insight and top-down approach to changing the strategy of what you're doing and not having all these different I do some, you do some, because then none of it becomes really state of the art. So you have to be very conscious about how you're setting up your entire operational model and how you are then structuring around that. And then the last bit is the overall mentality, the the um the mindset of people working in this area, because it has been a very, you can say, um, manual and people-involved process because there has been these requirements of going out to the different hospitals, checking that the patient actually exists and the data is there, and it's the true data that is being put into the systems in the end. Uh, so there's been a lot of um hands-on work, which is of course built into many of the models, also for the service providers as well as for the pharma industry. And that's labor-intensive, which is an obvious case for AI, but it will it will again change uh how we're working also in this industry with that. And that, of course, requires structural changes, uh, organizing ourselves in different ways, making sure that we combine competences from, I mean, IT, savvy people, people who know how uh the AI is built, how it's working, but also people on the other side that know how to operate in a world where we have much more AI put in without having those people being, you know, too lean back and and lazy and thinking the data is there. We need to work in a different way because we still need humans to you can say to verify and check that the output is not hallucinating or or is not uh occurring with the before-mentioned biases. So um, yeah. So it will be changes in the in the entire industry, how we structure ourselves, how we organize ourselves, how we are setting up outsourcing of clinical tribes, how we collaborate across with different partners.

SPEAKER_01 10:16

So looking ahead, you you you've mentioned multiple challenges and benefits that go um with the implementation of AI. Some of these are data quality, regulation, cultural readiness. If you could fix one of those challenges or barriers, which one would it be and why?

Why mindset and skills are the biggest barrier

SPEAKER_02 10:40

I think the most important part is preparing people for having more AI into our industry. It is an educational part, it is involving new competences compared to the more classical competences uh being used in running clinical trials, because we need a different skill set, we need new ways of thinking, and we need to have an end-to-end focus because in my I mean past positions, I get approached, I mean, several times a day by different providers who have a new idea of this bits and parts, uh, where they found a solution, and I'm sure they found a solution for that part. But on my side, being a senior person in an organization where I need to match it up with what else we have and make sure that all these different systems and offerings actually link together. So we need to have an end-to-end focus on processes, systems, um tools and capabilities that can come in and actually help us expediting things.

How AI could cut clinical trial timelines by up to 50%

SPEAKER_02 12:07

I can give an example of what we did previously in with my team. Uh, we were actually taking an end-to-end approach where we were looking at every step in in running the clinical trials, and then we were looking at how we can shorten its, you can say, shop delivery, so that it's not overall creating delays. And we occur uh we um we did go this step by step and looked at how the dependencies of all the deliverables before you can start something or finish it. So when we looked at this, we found out that we could cut something between 40 to 50 percent. If we could shorten both processes but also speed up time by you can say content creation with AI, can shorten things from months to days. And then you also need to think about I mean, how do you review things, how do you approve things, and so on, but you can really shorten things. So I think there's a big win in terms of speed, but there's also a win in terms of reduced cost. But I also think you can have a much higher predictability so that you can you can model and uh assess your clinical trial protocols upfront. You can uh you can you can make it much more consistent what you're doing. You can allow yourself to get much more input so that the clinical trial design is optimized before you even start. So I think there's there's a lot of administrative work that can be automized. There is a lot of data points that you can build in already up front so you get something that is much more predictable. So hopefully you get a higher success rate in the end, but also speed up things.

SPEAKER_01 14:11

That's amazing. Um yeah, it's a very interesting experiment. Just out of curiosity, was that like a six sigma uh approach, the the process you analyzed?

SPEAKER_02 14:22

No, it wasn't specifically a six sigma. We we did we did deep dive into the different areas and then we assessed what was even possible where we were, what would be the investment? Uh would we likely be allowed to do the investment at the time, and and also just looked at what what would it take. Uh, in reality, the question was how can we speed up things without expanding the number of people? So, what would it take for things smarter and faster with less manual power, you can say, because this is this is what some of the challenges that we have, because it can be difficult to get workforce, it can be not necessarily faster and cheaper to have more workforce to do the same. So we need to find smarter ways, and and that was uh a process we went through.

SPEAKER_01 15:24

On that note, you recently shared a report with me from Cup Gemini, and an interesting insight there was that the pharma industry is struggling with productivity since 1950, and it's constantly declining, and the the pharma industry definitely needs a breakthrough. Do you feel that AI could potentially be that breakthrough moment for pharma?

Can AI become the breakthrough for pharma?

SPEAKER_02 15:53

I do think that and AI is not just one single thing. There are many different tools, so to say, built on language models, machine learning, and and and so on that need to be implemented, they need to be adapted to what is needed, and there are many applications that is needed across because there are different focus from early discovery to I mean the more operational parts where it also is required. So I think saying AI can can make a breakthrough, yes. But AI is many different things in this sentence here, because there are there are many things that need to be done. It's it's doing data analysis, it's doing content creation, so it's not just one tool that can do it all. So it's not a want that can just we can swing a wand and then it's a miracle and then it's done. The uh Cap Gemini report also said that 82% of the surveyed uh executive believe that we have a completely different uh implementation five years from now. And I I agree with that, but I think five years from now we will still be in process because it takes time to get it all kind of into and implemented so that we have all the different tools that we need to really make a difference in in uh in the farm industry. And I think with with that being said as well, uh there will be a big differentiation between a big farmer that allow themselves to make huge investments and really going all in on this, and then you have smaller biotech companies that cannot afford to spend the investments in uh uh AI and technology to that same extent. So so there will be a differentiation in in the companies. I just read Eli Lilly and um Navidia has entered a big agreement in one billion US dollars here in the beginning of 2026 to really really build a new you can say lab for having full-on focus on getting AI into their discovery development and and and and what they need, other companies would not be able to afford that same focus to it. So it will make a differentiation on how that will look. We can have a talk again in five years and see uh where we are. But definitely I think things will speed up the next five years. And already now it is also almost three-quarters of the companies that have been surveyed, I would say, but it but I guess those numbers are maybe to the high end, but but more than 70% of companies either have a strategy and is implementing or is building a strategy for how to do this for their company right now. And I think this is something that we will just see increase because if you're in this industry, you need to have your data strategy and your tech strategy fully implemented with your business strategy and your organizational strategy. That's a need.

SPEAKER_01 19:49

Yes, uh exactly. Coming back to the NVIDIA report last year, life sciences companies, 63% of them were actively using AI. Whereas in this year, there's uh and this number grew to 70%. So it's gradually going up and up. So it's quite interesting where the direction that we're going to. Annette, thank you so much for your insights. Last question: if you could briefly sum up key takeaways from today, so the the benefits and the challenges of implementing AI in within the pharma um industry or sector, I'd be most grateful. Thank you.

SPEAKER_02 20:34

Thank you.

Key benefits and challenges of AI in clinical trials

SPEAKER_02 20:35

Summarizing, I mean, the key benefits are increased speed, increased quality, and the data predictability. I mean, using the data to really predict things so that you get a better uh uh idea of the outcome already upfront. And then the fourth benefit in my mind is reducing the cost, and that's uh bound up to uh to both the I mean increased speed, less resources, but also if you have less failed project, uh you also do save costs. So um so that's that's really where I see the the key benefits are.

SPEAKER_01 21:25

The challenges, if you could maybe give a brief overview.

SPEAKER_02 21:29

The challenges is uh the investments that need to be made. So uh it's both in terms of the the big investment in developing uh technology, but it's also the investment it takes to get people educated, re educated, uh, and and transforming the mindset and the culture in an organization that requires investment for many companies. Of course, we do have. Have some of these regulatory and legal and ethical considerations, but I think those are things that need to be overcome by having awareness of how the regulations are. And lastly, we of course also need to overcome the challenge of the built-in biases that can be in some of the algorithm and so on. But this is something we need to constantly verify and validate and ensure that we have the right traceability of what the different models and technology and tools are doing. So I think there's a lot of work ahead of us. We can't just lean back and say, oh yeah, let's see what happened in five years. No. Now is where things are moving for sure. And that's also what you can see in any reports coming out right now, surveying people from the industry and so on. You can see lots of things are happening at the moment. So looking forward to talk with you again in five years or in three years, because I think that will be definitely a different uh perspective we are having by then, and new learnings as well as we go along. So thank you very much for inviting me to this talk. I think it has uh been interesting to um yeah, to just look a bit into the key benefits and challenges, and uh for sure this is a continuous process.

SPEAKER_01 23:40

That's a wrap for today's episode of Beyond Delivery. A huge thank you to Annette Ostergaard for sharing her expertise and for giving us a clear and honest look at where Pharma stands on its journey toward AI-driven clinical development. If there is one takeaway, it is that AI has an enormous potential, but the readiness is not just about technology, it's about data, people, collaboration, and the willingness to rethink how the trials run. Thank you for listening and join us the next time as we continue exploring how technology is reshaping the future of life sciences. Thanks so much.

SPEAKER_00 24:30

Thanks for listening. What's one thing from the episode that made you think? Let us know in the comments and subscribe to the show. It really helps more people discover it. See you next time.