Ep #7 I The Biggest Lie Pharma Industry Tells Itself About AI | Dr. Neha Gupta

Every pharma company today says they're doing AI. But how much of it is actually real?

In this episode of The Best AI Show on Pharma with Ayush Mishra our guest is Dr. Neha Gupta who is a clinician, digital health product leader, and Ironman athlete and the one who has spent years building real digital products that have had to survive regulators, compliance frameworks, audits, and actual patients across multiple markets. She is one of the rare people who has seen healthcare from inside the clinic, inside the product team, and inside the pharma system.

You will hear insights on:

• Where AI is genuinely creating value inside pharma organizations today and where it quietly collapses
• Why AI confidence without calibration is one of the most dangerous problems in healthcare
• Why model accuracy and clinical trust are not the same number
• How real world evidence is being used to change what's possible in clinical trials and drug labeling
• Why 85% of AI pilots in pharma never move beyond the pilot stage
• What AI transformation in pharma actually demands from organizations beyond just technology
• The one space startup founders building for life sciences should be focused on right now
• Which global markets are leading healthcare AI innovation and why Brazil made this list
• The biggest lie the pharma industry tells itself about AI adoption

If you work in pharma, digital health, clinical operations, or life sciences technology, this conversation offers one of the most honest assessments of where AI in pharma actually stands today.

Dr. Neha Gupta's LinkedIn: https://www.linkedin.com/in/drnehagupta/ 

Ayush Mishra's LinkedIn: https://www.linkedin.com/in/ayushmishra/  

Watch this episode now on YouTube: https://youtu.be/kDG4VpM8kfo 

Watch Now on YouTube

Transcript

SPEAKER_04 0:00

Every pharma company today says they're doing AI. Where do you actually see AI creating real value inside a pharma organization or in general farmer organizations today?

SPEAKER_00 0:10

There is an inherent challenge with AI that it's it's confident, but it's not calibrated. So it it comes across as authoritative, as fluent, but it's also wrong. And healthcare has no room for errors.

SPEAKER_04 0:25

As far as AI is concerned, where does it actually survive in the ecosystem today? That it meets actual healthcare environment or pharma environment. And where does this product actually quite collapse?

SPEAKER_00 0:36

Accuracy of a model is not equal to a clinical trust. So if your if your model is 98% accurate, that will not really translate to a 90% clinical trust.

SPEAKER_04 0:45

AI transformation in pharma is not about innovation, it's about endurance.

SPEAKER_00 0:49

It's about consistency. Only 15% of pilots within pharma ever get to a real uh phase one. But nobody's really talked about about what after pilot.

SPEAKER_04 1:01

So meaning that says 85 out of all 100 projects in the pharma start with no long-term goal.

SPEAKER_00 1:08

Not start, they end with no long-term goal.

SPEAKER_04 1:11

What do you think is the biggest lie the healthcare industry or pharma industry tells itself about AI. Today we have a very interesting guest, Dr. Neha Gupta. Neha has had a pretty unique journey. She started as a clinician, moved into digital health and product building, and has spent years working across the pharma and healthcare ecosystem, building real digital products that actually have to survive the real world. So, regulators, compliance, audit, patients, everything. So she's one of those rare people who has seen healthcare from inside the clinic and inside the product team and also inside the pharma system. Also, she's an Iron Man athlete, which means she voluntarily, which I really fail to understand, she voluntarily runs swims and cycles for insane distances. So clearly she enjoys the pain much more than the rest of us. Nea, welcome to the show.

SPEAKER_00 2:35

Thanks so much, Ayush, and thanks for the warm words. And I think there is some truth to the element of wanting to have pain in life, because life in general also keeps providing pain without asking. Any of VR signing up and paying for it.

SPEAKER_04 2:49

But I mean, if anyone asks me to be an Ironman athlete, my philosophy is going to be I already have enough pain in my life.

SPEAKER_00 2:56

I mean, the only other difference, the only excitement here was going to be an additional medal at the end of the finish line.

SPEAKER_04 3:02

Oh yeah. Yeah. So let's start with the journey actually. You've been a clinician, uh product builder, and someone who looks at healthcare systems more broadly. Uh and you know, people generally stay in one lane in healthcare, which which is also quite enough sometimes. What made you step back and think, you know, I that I just don't want to work in inside the system? I just don't want to sort of work in one single lane and I want to experience the entire airbag.

SPEAKER_00 3:26

Yes, and I think um when I look back on my journey, I think it's one thing is clear, it was not a single defining moment. It's actually a collection of experiences, a collection of frustrations, a collection of insights, which which have sort of led me from starting as a clinical service provider to being a product leader to being this ecosystem builder that I find myself to be in. So part of it is experience and learning, and part of it is being intentional about impact. And um, I I want to say that when I was being a clinician, I think the intent was very much around how do you improve patient outcomes, how do you improve treatment outcomes. And slowly realizing that technology can do much more in impact than what it was doing. I started finding myself in rooms where there were great insights, there were great people, uh product people. But what was missing was the implementation. Things were getting handed off, and in this handoff, things were getting lost. So I want the intention here was to be that one person who can actually stitch this continuous theme. So who understands healthcare, who understands what patients need, who then I then you know works towards getting the product experience, becoming a product leader, develop the systems thinking to ultimately getting to a place where the impact is maximized. And this is how I find you.

SPEAKER_04 4:56

Actually, I'll I'm going to take you back to a conversation we were just having before we actually hit record button. And uh, we were talking about uh how our our journeys in the past, the colleges we went, and the you know, the the decisions we took at that time, not knowingly anything, actually shaped up who and what we are doing today. So, you know, just for the context, um, and especially in the context of the Indian culture, I want to talk about because you know, a lot of audience that we have on the podcast are generally uh from the US. Uh, you know, from the from the Indian standpoint, you know, can you sort of talk about how does this agility, you know, that you know you can learn to be a clinician and then can easily switch on the product side and then you can become an AI expert. This this demands uh a special set of skill, a special set of agility to be thinking into that direction. And when the opportunity presents itself, so you must be a clinician one day and then an opportunity presented itself. And you know, you you had two choices, whether to continue on what you know best or just to sort of start wandering and into uncharted territory. You know, what what is it in in those initial formative years of our lives, you know, or your life in Jaipur growing up, uh, that sort of defined or sort of gave you the courage to go into the uncharted territory?

SPEAKER_00 6:21

So I think that's a very good question. And something that you only start, or rather, something one that I only started appreciating once I left India. It's been more than, I think, 10 years since I've lived in India, and more than 20 years that I've that we were discussing that I've actually lived in Jaipur. But what Jaipur Foundation does, which I'm very proud and which I think is very characteristic of a tier two town upbringing, is the value-belief system it instills in you. That even today, when I am 8,000 kilometers away from my family, I know that I have a family behind me. I'm not alone. I have never been alone across till today. I think I've lived in more than eight cities. And five of those cities, I didn't really speak the language. And even, you know, with these drastic environments and situations, you know that we're not alone. So that is, I think, a very good um security and a foundation to begin and begin and shape an outlook towards change. The other feel other thing I think is more about an individual perception. So as as a person, I'm a very curious person, and a person whose curiosity to an extent that it's I think at some point in time it's it's it's a disadvantage. And that has been the foundation, and not foundation, but that has been the driver for all of these different experiences. And one thing that I feel like I might I might I don't agree with you fully about opportunity presenting itself. I think it's less about opportunity and luck presenting themselves, but it's it's more about how you want it to be.

SPEAKER_04 7:53

Because manifest it to be.

SPEAKER_00 7:56

Exactly. Because I think there is this very famous saying of Steve Jobs that connecting the dots. Whereas I think there is a slight change, and there is, you know, you choose the dots you want to connect. Because our life has so many experiences, and like we were discussing today, just before the call, uh just before the recording, that at some places your experiences are not intentional, but when you look back, you you choose what you want to connect, and then suddenly it makes a beautiful story. So I think it's a it's a collection of both.

SPEAKER_04 8:27

And I I think that this is this is such a you know, such such a nice starting point to my another part of the discussion as well, because that's precisely how you know you generally structure AI as well. I mean, which is connecting the dots. So when you sort of create an AI strategy, uh, you know, this is this is precisely the methodology that you follow, which is, you know, the decision that I take today, how does that change the outcome of tomorrow? Uh you know, so let's let's delve into right away the big buzzword right now, which is AI. Um, every pharma company today says they are doing AI. Uh, but from where you sit across the ecosystem in pharma, in healthcare in general, where do you actually see AI creating real value inside a pharma organization or in general pharma organizations today? Uh, you know, so where do you see it creating real real value? Not just dashboards, but you know, where do you see it creating value for actual decision making?

SPEAKER_00 9:25

Mm-hmm. And I love your disclaimer, not just dashboards, because I think that's the most hyped place.

SPEAKER_02 9:30

Oh, yeah.

SPEAKER_00 9:31

Where everybody's is talking about an AI-generated dashboard or and and there is nothing, there is no AI behind it. I think in in digital health, the areas where um, and I say digital health because to me it includes broader than pharma, and it it actually incorporates the end users who are our patients and the care team. And in digital health, there are there are three areas where AI has significant impact so far. One is, of course, your drug discovery. And we have enough examples today to show how AI can help you in targeted, focused molecule identification. Um, the one example that comes to my mind is um in silico medicine. I think they were able to bring a molecule to phase two in around about four years, which is half the time that is required usually by the traditional route. So that's a significant, significant impact, both from cost, timeline, and ultimately treatment outcome perspective.

SPEAKER_03 10:32

Yeah.

SPEAKER_00 10:33

The other element where I feel, again, this is the clinical trial space where AI has been um impactful is the patient recruitment and the AI-assisted site identification for clinical trials. Also expanding the label by helping um fulfill the critical sample size and extension of label. For instance, the Pfizer's Iberance is a great example of breast cancer drug where because uh where, with the help of real-world data, Pfizer was able to extend its label of the breast cancer medicine for male patients because the incidence of breast cancer in male patients is 1%. And by virtue of that incidence rate, it was impossible for them to have a sample size and clinical trial to show the efficacy. But this is where they partnered with Flatiron and were able to actually demonstrate the safety and efficacy and extend the label of the drug. So I think these are the places where AI has been significantly instrumental. The places where uh the other category is the admin element. So within pharma today, we uh within pharma, a lot of admin-related tasks, the internal focused tasks, can be taken care of by AI. And when you do that, you actually identify and you release some capacity for higher order thinking. You you release some capacity for more human trust-based decisions and a more uh supervision of the what the motor models are doing. So I feel like those are the areas where AI today has been doing a great job. The areas where AI still has a bit of a leg up to get to is um personalization. So we talk about patient personalization. We are it's it's not there yet. This is mostly uh a GPT wrapper on a rule-based engine, a very poor patient segmentation. And it's unfortunate because every time we are doing a conversation, whether I'm at a vendor site or I'm at the non-vendor side and buying side.

unknown 12:35

Yeah.

SPEAKER_00 12:36

I know, I know that this is this is a lie that is very carefully gift-wrapped, and everybody is accepting it because there is no other way to it right now. And and the other, and and one of the reasons there is no other way for it is because our regulatory and the compliance frameworks have not caught up to that. So we are obviously not able to deliver that kind of personal personal the kind of personalization that AI has the capability. And I don't want to say it's most it's only because regulatory and compliance, but it's also because at some there is an inherent challenge with AI that it's it's confidence, but it's not calibrated. So it it comes across as authoritative, as fluent, but it's also wrong. And healthcare has no room for errors. So I think there's it's it's a very interesting problem space to be operating in.

SPEAKER_04 13:30

And I I I agree with you, and especially with the personalization, because even in our day-to-day, and you would receive multiple emails where you can easily say, oh, this email was written by GPT, and this is actually a person that wrote it. It's so easy, right? So you know, some somehow Chat GPT has got a personality of its own, which is so, you know, you can tell this is a GPT email. I mean, so I have actually completely stopped using Chat GPT to write my personal emails because at the end of the day, it doesn't feel like I wrote it, uh, even if the message is the same. So I I agree with you. And I I want to take you back to a very important use case that you mentioned. And if I'm being honest, this is the first time that someone has identified as an interesting use case, which is the real-world example that you took for identifying the right patients or identifying the right sites. Uh, you know, and you know, I would I would like to say thank you for it because that's also a bit of a validation for us. We recently launched a product called Alpha Dev uh while we were at SCO. And that product, so we partnered with Komodo Health, which is one of the majors in real-world data in the US. And uh we created uh a platform on top of Komodo Data's Health that helps, you just upload the protocol, you just type in what your uh IE criteria are, or even just specific uh you know, random specifics. It then converts them into your IE criteria, and then it you goes, takes all that IE criteria to your Komodo uh data set and then brings back some very interesting insights on where your right patients are exactly. And then we have a proprietary data set on site intelligence. We have uh intelligence on 1.2 million sites globally. So Komodo's data tells you where the patients are and where's which sites are they available at, and then our site intelligence runs on top of it. And the idea is exactly what, and I to be honest, I didn't know about the Pfizer example. I should have but uh so this is this is a very nice validation, and uh, you know, I think uh that's also uh something which I want to sort of expand further on with you. When we show it to a lot of our clients, eight out of ten clients are surprised that oh, real-world evidence can be used like this as well. So you saying it so confidently, oh, this is something that should be done versus the conversation now we are having. And we we are speaking across the spectrum. We are speaking to CROs, uh, obviously sponsors, we're even speaking to the recruitment agencies. And, you know, as I said, eight out of ten guys tell us, oh, we never thought real-world evidence is to be used like this. What do you have to say about that?

SPEAKER_00 16:16

I mean, how how how how is there so difference, you know, in sort of understanding the Yeah, no, but I think Naish that's that's the uh that's something that we briefly touched upon, is one of the biggest challenges of AI adoption. Today, the AI adoption, more often than not, is considered as a procurement acquisition decision. Which company, which tool, which model, who. But ultimately, acquisition is different from adoption, and which is what you're highlighting, right? So at some level, adoption is inherent on organizational change program. As an organization, which is the and the building blocks of an organization are your teams and your culture, they need to change and adapt in a way that there is balance of innovation, there is balance of people who are saying that this is the constraint, these are guard drills, and allowing impact. Because what happens more often than not is that people are very siloed in what they're doing. They're very siloed in their day-to-day activities and and then they hear these buzzwords about AI. But because they're so siloed, they're not really able to appreciate the the impact potential. And all they're able to see is what comes on their desk and what comes on their calendars.

SPEAKER_04 17:28

So you and obviously, you know, you you are you're well the person in your organization, I would assume, whose job is to actually solve exactly that. You know, how how can more and more people sort of contribute to create that digital impact? And I'm, you know, I I I would love to have that gossip from you. I know you are not going to give it to me, but uh w which is which is that team, you know, that you guys sort of, you know, it's the most difficult to sort of get together on a product. Is it is it the clinical guys, is it the commercial guys? You know, I'd love any gossip.

SPEAKER_00 18:08

Yes, I think this is um this is the chat, this is the um the bane of existence, if you want to call it that way. That the the days, you know, like when you when anybody asks me how your job is like it's a bunch of good days, bad days. So this actually is not a good day, a problem statement. But I don't think there is any shortcut to it because in some sense this is how change looks like you always have early adopters, but those are not your those are not the people that you have to convince. The people you have to convince are your naysayers and the the lagards. Um for for healthcare in general, they are mostly found in the regulatory and the compliance teams. And by the impact of the services of healthcare, this is a highly regulated industry. It's a it's a it's an it's an industry that has severe compliance needs. I think the other industry that comes close to healthcare is finance. Those are the only two industries which operate under such strict compliance needs. I I having read a little bit about innovation and finance, I think finance innovation is a it's it's light speed ahead of where digital healthcare is. And and I think it's it's it's a a lot part of that is because they are compliant. They need they have compliance, but they don't have regulations. In in healthcare, you have compliance, so you have all of those QMS, you have all of those ISO standards, but you also have MDRs and MDDs and FDAs and and and D Gars. So I think it's it's just like a double hard trajectory to approach.

SPEAKER_04 19:48

Uh so I mean, you have built these products before. As far as AI is concerned, where does it actually survive in the ecosystem today where it meets actual healthcare environment or pharma environment? And where does this product actually quietly collapse? I mean, so do you think it it collapses from uh from because of a lack of uh context or technology, or is it collapse on the compliance side, does it, or probably it just collapses on the bureaucracy side or the organizational culture side? Where does an AI product actually collapse today?

SPEAKER_00 20:22

So I will answer it in the kind, I mean, it's it's not one reason for all different kinds of products. So for instance, if your AI product uh is um supporting clinical decision making, so it's actually a tool for your care team. Yeah, there the possible reason for its collapse is not integration to workflows, where it's actually not helping in their current workflow, but it is an additional task, and and the care teams are already very burdened and they're already very jaded by the technology. So that's the that's the core reason a clinical decision support tool would fail. The second reason this category tools fail is the trust. And you know, it's it's also very interesting to know that accuracy of a model is not equal to a clinical trust. So if your if your model is 98% accurate, that will not really translate to a 90% clinical trust. And I hear this argument a lot of time that my doctor made a mistake, my doctor also misdiagnosed. Sure, but that human trust cannot compete with an algorithmic trust. And I I don't think, or at least I'm yet to know, and if you know, and if any of your audience knows, to tell me how can you actually compare a 98% accurate AI model to a clinical accuracy, how do you compute clinical accuracy? Is there a way to compute clinical accuracy? So I think those are the two elements where the clinical decision support tools fail. The tools which are facing the patients fail more often than not by not actually understanding that patients what patients need is a behavioral support. So if you're giving Information that's great, but you also have to understand that the patients need to be seen, the patient needs to, they want to be heard. What is that behavioral element of the patient's journey in disease management that needs support? So I think that's the core element where the patient support programs fail.

SPEAKER_04 22:17

That's a very interesting take, actually. Uh and and I and I agree, right? I mean, um, and we we as a as a native AI company has faced that a lot of time, um, especially even with Alpha Dev. When we started building it out, we want to build the most fancy, most complex machine learning model ever made. And then what we ended up making, because fortunately we had some very interesting, and when they saw the product, they said no one is going to use it because for them this is just another work. And uh it needs to integrate in their workflow, so it needs to be extremely easy. It doesn't, it shouldn't look like uh additional work. So uh but have you have you uh found a way to sort of build trust with AI in your organization? I mean, are are there any best practices that someone can sort of implement if they want to create more trust for AI products in their organization?

SPEAKER_00 23:15

So I don't think again, I don't I don't have an answer that fits all situations, but what has helped is actually publications. So what has helped is if I'm being able to draw analogy from published evidence in repetitive. And also um what has helped is um having other industry AI supporters talk, have a having a panel discussion, you know, percolating the ideas around. So those are the two elements that have helped. But I again my expectation with this is that it is in the same pace as an organization change, where it takes one step at a time. And this is where my my endurance journey helps because you know many a times where you're at the 32, 33 kilometer mark, or when you're you finish your swimming, you finish your cycling, and you still have a full marathon remaining, you're not looking at it as of 20, 42 kilometers left after having done, I don't even know what, four kilometer swimming, nine, one eighty kilometer cycling. You're looking at one kilometer at a time. And I think that's that's a very key translational learning that when when problems are big, for instance, this trust is a massive um win. You it can you cannot achieve it overnight. You have to achieve one step at a time, one person at a time, one moment at a time.

SPEAKER_04 24:35

Wow, this I mean, this code should go in our thumbnail, by the way. AI transformation in pharma is not about innovation, it's about endurance.

SPEAKER_00 24:44

It's about consistency. I think it's about consistency. In fact, I feel a lot of life is about consistency. And people keep asking me, well, what's your motivation? I don't think there is a motivation. I think it's about determination and consistency. Because motivation is that, you know, it's it's like the flavor of the month. It comes someday, it doesn't come the other day. And you can't keep waiting for motivation. But when your alarm rings in the morning and you have a schedule and you have a race in six weeks, you know you have to practice.

SPEAKER_04 25:12

All right. That's that's that's going to be new a lot new for a lot of Instagram goers.

SPEAKER_00 25:20

True. I mean, and and unfortunately, we live in a world of instant gratification, instant response, instant satisfaction. And this is where I'm addition of the day. Exactly, modification of the day. And here we're talking about making a change in the organization, which has been around for 20 years and has more than 10,000 people. It's not an overnight thing.

SPEAKER_04 25:42

I need to play for the longer game.

SPEAKER_00 25:44

And you have to play the longer game. You you only win in the longer, longer game. You have instant gratification in shorter ones, but the win is only a longer game.

SPEAKER_04 25:53

So when when you're in the game and you know, you and I I'm sure you talk to a lot of uh, you know, vendors and software developers and model developers who must pitch to you uh for various reasons. Uh when you hear those pitch, what are those green flags that you generally look at? Uh and you know, what are those green flags that make you say, okay, these people know what they're doing?

SPEAKER_00 26:19

So I'm a big fan of uh people presenting failures, people presenting gaps. Because I think pitches in general are a very marketing, uh has have a marketing approach. And where you're in, where when you're working with technology as AI, you want you're not looking for, again, you're not looking for an explosion. You're looking for honesty because I know that you know we are not there yet, and we need we will together get there. So if you come to me saying that this is uh we've done this with banking option, and yeah, exactly, you know, we've conquered it. And and then I know that this is not a vendor I want to work with because we we I know you've not conquered it. I want to know that you can conquer it, and that's that's that's my you know, that's that's my selection criteria. So for instance, I mean I was reading this article the other day. It seems that only 15% of pilots within Pharma ever get to a real um phase one.

SPEAKER_04 27:20

Deployment, yeah.

SPEAKER_00 27:21

And and and to me, this was this was again going back to our very short-term gratitude scenario where there are massive kickoffs, there are massive stakeholder alignments, and everybody is talking about a great project that is in pilot. But nobody's really talked about what after pilot. So 15% is a is is is a very eyebrow-raising number.

SPEAKER_04 27:42

And I'm looking for a vendor much bigger number than it should be.

SPEAKER_00 27:47

It should be a much bigger number than this, right? Because 15 is really, yeah. And and I think in in my vendor, what I'm looking for is the ability that, okay, if this fails, you will tell me how we can change it. You will tell me how we can work together. Because don't come and tell me everything is golden. It's not.

SPEAKER_04 28:05

So, I mean, some meaning that says 85 out of all 100 projects in the pharma start with no long-term goal.

SPEAKER_00 28:14

Not start, they end with no long-term goal. Yeah. Everything starts with a great long-term goal, lots of project plans. But it's only after a pilot that you realize that, oh, this is not going anywhere.

SPEAKER_04 28:26

Wow. Founders building AI for pharma actually understand how how decisions actually get made inside these organizations. So I mean, A, and I think this is I mean, this is such a unique point that you raised is if there's any founder who's not talking about failures, you know, that's a red flag. Uh, what should these founders understand about how uh decisions are being made inside organizations like yours or in you know in global pharma in general?

SPEAKER_00 28:53

I I think the two elements, right? One is I would paraphrase that saying that it's not just about talking failures, it's about talking learnings and how do you navigate and mitigate failures. Because what pharma is looking for is the partnership with the team and the capabilities. The other element, I feel like the decision making in pharma is is a very long-done process. So it's it's unfortunate, but it's a reality that decisions don't get made in six weeks. Decisions require a fair number of stakeholders who need to be aligned, who need to be brought together for the project, and who need to be on board with the with the particular um product, particular process, particular vendor. And ultimately it's it's it's it's a process that required I mean the selection process will always have a process around it. So I mean it's unfortunate because I've also worked with startups where we have very lean, mean teams, and you're not always able to keep up with these long cycles of the uh of a pharma company, not being able to keep up with these multiple requirements of different kinds of decks, different kinds of reports, fill in these 20-page long documents, procurement documents which have been created for traditional IT systems. But it's again, it's I go I keep going back to the organizational change framework because this is all part of that slow-moving Titanic, which is, you know, which will which will take time. So today, sales in pharma is not a six-week, it's 12, 18 months process. And unfortunately, again, this is a bit of a um a bummer, but stakeholders change in the in the meantime. So I think overall there is nothing I I don't have a rosy want to paint here. It's not. The only only silver lining here is that once you're in, you're in. Because then removal in itself is the similarly long process. So, you know, like think of that as your finish line.

SPEAKER_04 30:57

Wow, that's uh I I've heard this kind of uh framing for the first time.

SPEAKER_00 31:02

No, it's true because we have vendors who have disappeared, and it's for us it's uh it's the same amount of hope to get them out of the system.

SPEAKER_04 31:10

I always thought, you know, you you need to really perform if you have to keep in.

SPEAKER_00 31:17

Growing up again, this is going back to Jeopardy and the competitive exams, right? I've always found that cracking the exam was the most difficult part. And you had the exam and you're in the school or in the college or in the institute, then it's their problem.

SPEAKER_04 31:36

Wow. So on that note, uh what is so I mean, uh, you know, I'm obviously I don't want to sort of get into the specifics, but um, you know when when a when a vendor sort of reaches out to you um and there there is a guy who told you, okay, I have this amazing platform ready to deploy it from day one. You know, all the buttons are working, it you just need to press the button, right? And then there is another partner that comes in that says, Okay, I have the I have the most fantastic team who can build a product for you, right? What is the you know, which which is the pitch that sort of gets excited or I mean that excites you the most?

SPEAKER_00 32:24

I think a competition of both, where there is already evidence that there is an existing product, there is an existing platform, there is an existing technology, and the capability that this is actually continuously being iterated, or there is a capability to change it, modify it, adapt it depending on the use case. So I think a combination of that. But also I feel like today, the world where we still sit today, it's it's it's a very personal relationship-based sales. So I mean, I get a lot of emails, and I'm sure a lot of people on the buy side get a lot of emails. It's not possible, it's not feasible to reply to all, and that's unfortunate because at some level, you know, we are not able to sift and identify who's a good. I mean, my I my ideal way of approaching this would be that I answer all the emails, I do demo calls with all of them. So I I'm making an informed decision. But it's not feasible with the other things that are already going on in my day. And then what happens is that, you know, like how we find a Docker in India by, you know, do you recommend somebody?

unknown 33:28

Yeah, yeah.

SPEAKER_00 33:29

It's like that, that oh, I have this problem. Sit uh problem, I either know somebody already because there's ultimately a very small space, or then I'll ask somebody in pharma who's already done this, that you know, who did you do? What do you who who else did you evaluate? So ultimately it becomes a little bit of that. So, and and the reason I bring this up is because that people who are starting fresh in in the technology space on the provider side. There is there is a benefit of partnering or going to conferences, networks, networking events to actually find that individual who is in that space, bandwidth, to look at your product and become your champion.

SPEAKER_04 34:10

Wow. Yeah, I mean, because uh there there are uh and and I I know a lot of very good founders who are trying to build something from India uh for the global pharma space. Uh and most of them have amazing pedigree, you know, uh best institutes, best ex-organizations, you know, someone's from, let's say, an IQV and so on and so forth. Uh, but yeah, and and they would have the best minds, best product, best services. But you know, three years down the line, they're now thinking to shut shops. And I think precisely because of this, is because they're not able to penetrate uh or that they're not able to sort of penetrate that relationship inner ring. Yeah. Uh we just spoke about. And uh so basis you, the recommendation is to participate in as many conferences as as you can.

SPEAKER_00 34:57

Yes, you know, like find your hook. You need to find that person because after a while it's it's it's ult it's unfortunate, but that no it gets uh lost in the noise.

SPEAKER_04 35:07

Yeah, yeah. No, I I get that. I get that. And uh, you know, this is something I'm I'm quite proud about my team. Uh I mean we we we we understood this pretty well pretty early in our game that you can't do everything sitting in India. And uh, you know, we I I remember uh two or three years in the business, I started talking to a couple of my own uh clients and say, okay, can you recommend someone whom you would buy from? Right. And he's we kept on reaching, and then we we made a pretty amazing US team. And I think that's one of the best decisions we ever took. So I I I second, I second what every every single word that you said. Okay, uh, you know, let's let me take the conversation a little on the product side. Uh, we discussed about the people, about the companies that are building it on the product, and uh you probably see a lot of AI demos and products. Let me ask you this as someone who has seen a lot of them. Is there any AI product in healthcare where you genuinely thought, okay, like this is impressive? And not just a good demo, but I mean something where you felt that this could actually change how healthcare works. And I mean, I don't want to take names. I mean, if you if you can, perfect.

SPEAKER_00 36:28

If you can't, I mean No, I think to be honest, it's I I can't think of a name in particular, but I do know that I'm a little bit far removed from areas where AI is doing magic in healthcare. So for instance, today the the AI penetration, AI adoption, AI impact is much, much felt in the whole imaging diagnostic field, mostly in on college, where you have multimodal AI as a possibility building up. I am that's not my focus area. So I don't really look at solutions who are in the market and doing this, but I read about them. And I think that, I mean, I can't think of a name right now, but those are the areas where there is definitely a massive um impact by AI image-assisted diagnosis, AI image-assisted treatment monitoring, and the ability to actually also get the software medical device statuses up to class three on images. So I think that's a that's a fair amount of impact that AI is generating in that space.

SPEAKER_04 37:29

So you see you see value uh in that model, in those products that AI is building up.

SPEAKER_00 37:34

Oh, absolutely. Absolutely. So for instance, I mean, in any of those uh skin cancer conditions, the election becomes so much more powerful. And even if it is not a clear diagnosis, but it is more a prognostic in the in the more prognostic space, even then it's very powerful. And it's a great tool for a clinician.

SPEAKER_04 37:55

So uh let's talk a little bit about what you actually sort of do. I mean, which is building health products and uh, you know, I would assume building uh digital products for the use of the patients and the end users, as you rightly say. Uh when you strategize of building a product like that, you know, how does how is that product different? Or how does the strategy or the framework of that product different from what someone is going to make in drug discovery? Because here you are engaging with real patients, and then you know that's that's also a sort of a layer of complexity on it.

SPEAKER_00 38:32

So for us, actually, the the I mean my focus has mostly been once the product is out in the commercial space. And hence the focus is around the end users, patients, or the care team members, or the caregivers with the patients are included. And one of the key learnings so far has been the whole behavioral support and the clinical workflow integration. So for if if the if the if this if the product has a care team user, then it has to be integrated in the workflows, or then it is providing a massive clinical value to the user that they would actually use and adopt another tool. And my focus in building most of these products products is keeping that in mind. So we do a fair amount on user journeys, we do a fair amount of testing of the product and the clinical workflows, how does that actually sit? For the patient, what I've been learning and what I've learned over the years is that information is good, giving them diary, giving them symptom checker, and all of that is good. But ultimately, they do all of this only when they feel that they are either being heard or seen by their care team. So the digital adoption is very much dependent on that tool being launched by the digital by their care team members, either directly or we allow a chat functionality, it could be an asynchronous chat, but some way where the patients have this feeling that they're being visible to their care team. So, with that in mind, we do a lot of understanding of user unmet needs. We try and understand what are the users actually looking for, where are the gaps in today, what they feel, how can we emotionally support them? And aside from what is the code reason that has been identified for the solution.

SPEAKER_04 40:27

So, how how do you sort of educate your patients? So, I mean, let me let me paint a picture. So if you have been watching the World Cup, which congratulations to us, we won last night. Uh so you know, during the match, Chat GPT is running a huge campaign. Um, so and then so there is there is an ad like a D2C brand where you know there's a women's cricket team that's playing, and then uh it they're just putting up a prompt on GPT saying, you know, how do how do we improve our cricketing performance of this women's team? And the GPT is answered, right? So, you know, it was so I was so pleasantly surprised by seeing the campaign because you know, Chat GPT is now a very tier two D2C product in India, as they're sort of presenting, right? So they've taken it to sort of that level. Now, comparing that to you know something what you do. I mean, you also operate in a similar zone where your job is to make sure that the patient adoption for uh products and applications like this is higher and higher, right? Taking inspiration from something what GPT is doing at that large of a scale, uh, you know, where do you think is pharma, you know, in that comparison of running campaigns to educate their users on adopting these AI apps? So how how how is pharma keeping up in the race of making sure that their users get used to AI as compared to a GPT?

SPEAKER_00 41:53

Oh, I think we are far behind. I don't think we can. Because again, it depends on which world you're operating, which market you're operating in. So India is a little bit less regulated compared to Europe, UK, US. Here, we cannot directly promote any solution. As in V, as in a pharmaceutical company, I cannot directly promote a solution. And it has to be um unbranded, it has to be therapy agnostic. So there are a lot of these constraints. So we cannot, I don't see that at least in the next five years, a pharma company is able to do a direct D2C campaign for a medicine that is not a generic drug.

SPEAKER_04 42:37

Right. You said next five years. Uh then what parts of pharma operation, and not not just yours, like the entire ecosystem, what parts of that of a pharma operation do you see changing the most in next five years? Like, do you see the biggest impact being created on kinker trials or regulatory drug discovery, commercial? Which part do you think is uh is going to change?

SPEAKER_00 43:01

Uh quickly the the the pre-commercialization phase of an asset and the admin tasks within the internal operations of the pharma. And third, I would say is the insights piece, the competitive and the market insights. All three of these pieces have are much more advanced for AI technology innovation and adoption compared to the direct end user space.

SPEAKER_04 43:29

And and in in this, you know, in uh as a follow-up to your uh your answer, any startup founder, you know, watching this uh this podcast today, uh which use case he or she should be focusing on most? And I I ask from the backdrop of there is a lot of use case, especially on the market intelligence side, as I would assume, that would be commoditized by, let's say, a Claude in three years. So, you know, in three years. Years there are going to be teams inside pharma companies creating their own dashboard just by a prompt on clot. They don't need any external partner for that, right? So how does a farmer or how does a startup founder in this space create a mode for himself or herself that is not going to be killed by Cloud?

SPEAKER_00 44:21

I think the answer is going to be somewhere around the compliance side of things because Cloud, OpenAI, as in ChatGPT, Gemini, Copilot, they're not compliance screenlighted. And the the room to play here is to build that compliance-friendly solution or a compliance-friendly usage that sits on top of these LLM tools.

SPEAKER_04 44:49

Wow. Maybe I should cut this from the podcast, keep this insight only for myself. Sir, you know, uh and I'm I'm sort of forced to ask this because you you have also uh you know sat on the VC side of the table, uh, you know, some point in your career. Um when when you see uh you know sort of these startups coming up today, uh and when you see the pitches, you know, what is that one thing that you know you are looking in in a startup pitch that's building AI for life sciences today? What is that one thing that you would look more aggressively? And if that's not there, that's then that's not the company you're investing.

SPEAKER_00 45:36

So I think it depends where it's a series, which series are we talking about?

SPEAKER_04 45:40

So I would say a seed.

SPEAKER_00 45:42

So for a series, say, this would very much be two elements for me. One would be the ability to have a product market fit and demonstrate that through numbers. And the second would be the team that they have behind it. So those would be the two big elements, you know, like for me, because I think what team generally works best?

SPEAKER_04 46:04

So, I mean, let's say there is this guy who who's fresh out of an IIT uh and who has some amazing tech capabilities. You're already smiling, so I understand. And then there is this guy who has spent 10 years working with big pharma, you know, at let's say an IQ via or you know, a company like that. And who understands that space well, who has the relationships well. So, you know, what is the kind of team that you would sort of put your thoughts behind on the stage?

SPEAKER_00 46:30

For a CDC, I think um for a CDC, if those are the two people in the team, then I would very much like to see the person who's experienced who has the network to be in the CEO sort of or CCO sort of a role, or at least a role which is nearer that domain. And the person who comes with a technical background to be leading or being in the technical team. And of course, you know, like now, I mean, we we are fortunate enough to have people who might have both. Like, for instance, somebody like me, you know, who comes from a very specific industry experience has over the period gained um relevant network as well, built relevant network. So that kind of an experience in the advisory group would be very, very you know influential in making that. Because this gives you an understanding that this is a well-rounded team where the skills are in the right place. People know what they have to focus on. And again, having worked on across VCs, across companies, I think what I feel like when you're starting up a company when in the pre-seed or the series A stage, you need you need a very horizontal layer. So you need smart people, you need people who can continuously do, they can be the office managers, they can be your customer engagement. But when you're at CDC, you need a bit more of that vertical line of T as well.

SPEAKER_04 47:58

A bit more structure there.

SPEAKER_00 48:00

You need you need both the people who can do across, but that that layer is now much more senior, and you have the individual vertical layer individual in the teams, which need to be assessed and which is contributing to your uh competence of the team.

SPEAKER_04 48:20

So uh, you know, there is there is again another sort of a kind of a I I won't say challenge, but obviously this is a uh a discussion that I hear a lot with founders is is about the data. I mean, if you're building AI, you you need to have access to some sort of data. Otherwise, it's just a a wrapper or you know. So one thing that I mean everyone sees a lot in healthcare and especially with AI projects is that data technically exists uh at multiple places. I mean, free data, proprietary data, you know, you have VMR data, King Cal trial data, claims, internal pharma data as well if you're working with a client. But everything is very fragmented. So think la you know, pharma companies, big big pharma or even medium-sized farmers will eventually or are planning to build a unified AI data environments, or will they keep on relying on external data ecosystems for still a considerable amount of time?

SPEAKER_00 49:16

I think the intent is definite, I think the data awareness is definitely there. I mean, today everybody knows that the these interoperable data sets are not useful, or they're only as there is the utilities only a little bit. And you need structured, good quality, reliable data to be able to build. So I think that we've done a whole amount of education and and and awareness around the requirement for good quality, understandable, interpretable data. In the future, it'll have to be a bit of both, Ayush. I think one, because there is so much data that you can simply not just delete. And so you will need solutions who are able to crack that interoperability with those data sets and unleash the power there. At the same time, the intent moving forward is to actually build data sets which are much more unified and are and are already talking to each other. So it's not just pharma, I think it's across there's the the there is this need that what we build will get built somewhere in between of talking to the older data sets, but is still talking to each other. And identifying how do you find a way to talk to these data sets, older ones which are not okay.

SPEAKER_04 50:38

So so meaning, and you're let uh let me know if I've uncovered this right. So what you're saying is if you're a startup founder building something for global life sciences today in this space, and you know, you uh either you should be building something uh for the compliance layer on top of the LLMs, let's say, or you should be offering services that unifies data for pharmacopies. I mean, is is that a are there are these two use cases that you know are still bulletproof?

SPEAKER_00 51:12

I mean, there are good there are good places to put your money in, yes.

SPEAKER_04 51:16

Call it. Call it. Yeah, I think put them online. And this is uh, you know, this is quite interesting. Let's talk about the work for workforce side. You know, pharma companies employ huge numbers of people, um, and even pharma services companies, you know, either on the sides of analytics, medical writing, clinical, regulatory ops. If AI really evolves over the next decade, as everyone says, and as we were talking before the episodes about that, you know, that magic explosion, if that magic explosion really happens, uh are there roles inside pharma that should be slightly nervous about the future? I know that's a scandalous question to ask.

SPEAKER_00 52:03

But I mean, why pharma? I feel like the entry-level roles are nervous across the board.

SPEAKER_02 52:08

Across the board, yeah.

SPEAKER_00 52:09

Especially in I think the um the analyst, the um associate, and the I mean the Today. I mean, my team is very lean today. And the reason it works is because a lot of our research analytical work is not done by a human being. And that in itself is because of AI and LLM modules. So I think so.

SPEAKER_04 52:43

The impact has already started.

SPEAKER_00 52:44

It's already there. I mean, you hear so many people mean made retundant across all servers, across the big four, across the top three. So, I mean, even the advisory space, this is already happening. And at some, and it's unfortunate because I was discussing with a friend the other day that today we are able to sit on top of these LLM models because we actually did do the grunge work at some point in time. Now, the level people, they are not given the opportunity to do a grunge work. So, how will they ever get to a place where they are actually able to sit on top and use these tools with an experience that and maturity that we didn't know which prompt to write? Exactly. And how do they know it? What is uh hypnosis and what is actually true from the model?

SPEAKER_04 53:28

But isn't that too unfair for the younger generation now?

SPEAKER_00 53:33

I don't know. I mean, I find this word unfair a bit difficult to answer because what is who who what I mean, I don't think life in general is fair. I think life is unfair to begin with. And what I think is is unfortunate is they were not warned. I think people who are just coming out of colleges did not get a warning. Whereas the people who are now in school are already aware of this and they will train accordingly and they will train, you know, adapting this. So this is this this this is the generation which is actually being thrown in the deep end. So that's that's that's definitely not right for them. But then I think every generation has its own share of deep end. So I think, you know, like as a millennial, we had our own deep end deep end. This generation, this is their deep end. Like my parents' generation, they'd never had internet. So for them, learning the whole use of smartphone, etc., I think it's it's a bit, I agree with the sentiment here that these entry-level jobs are difficult. It's um there are so many redundancies at this particular area and this particular layer.

SPEAKER_03 54:41

Yeah.

SPEAKER_00 54:42

And to be honest, I mean, having worked with human beings at entry-level roles versus having worked with these Clauds and Geminis and Copilot, you already know which one you prefer.

SPEAKER_04 54:57

I hate to agree with you, but I do.

SPEAKER_00 55:09

I mean, this it's it's it's unfortunate. Yes, exactly. It's hassle-free. And and yeah, I I wouldn't call it unfair, I think it's unfortunate.

SPEAKER_04 55:20

It is, it is. And uh, you know, um uh I think even even the generation sort of uh the the Gen Z as sort of we can who call them, uh I think uh, you know even they cannot operate without AI. I mean, uh obviously AI is sort of eating up a lot of what they should be doing. That's also a fact, but even they themselves can't be operating without AI because that's that has become such an integral part of who they are. I mean, uh so the the old good days of uh mental simulation, you know, that's that's in the past. And uh yeah, as I agree. I mean, that's that's unfortunate. Um you know, let's let's let's stop there before we sort of start getting trolled by agents. I have some very strong comments on on these things which I keep on saying in my teams internally, and I'm not I'm not very likable that case.

SPEAKER_00 56:18

But let's let me No, that's true, and I feel like like for instance, you know, like the I why I don't I have strong reservations for the word unfit. So for instance, take dentistry as a vocation, right? Up until today, dentistry is not uh there are too many dentists in India. The the kind of uptake of dentists of a new dentist in the market is limited. There is so much of potential that gets unvalidated.

SPEAKER_03 56:44

Yes.

SPEAKER_00 56:45

AI, that's one vocation AI can never replace. So it's now it's suddenly fair to the dentist. Is that is that how we would want to, you know, like articulate that?

SPEAKER_02 56:55

Yes.

SPEAKER_00 56:56

But no, I think, you know, like at some levels, it will the the presence of AI is going to be unfair to certain industries. It's gonna be fairer to industries like dentistry, which are a bit more hand-based, they're a bit more artistic, it's a bit more um uh what is the word for it? You know, it it's a bit more des dexterous field. Yeah, yeah. Dentists and mediocre dentists would then be forced to actually um perfect their craft because now they will get valued for the perfection. So that's I think a f silver lining of what LLMs are doing in the world today.

SPEAKER_04 57:34

And I agree with you. Uh I mean, uh it has been I mean, quite fortunate for me, the the whole AI thing that happened. Because as a founder, you know, I I used to feel a lot limited because you know there are only a limited number of things you can sort of uh delegate. And, you know, then the other speech other people can do exactly how you want them to do, right? But now because of AI, I'm able to pursue at least 10 times the ideas that I would do five years back. And uh I mean uh fortunately because I know the right prompts to put, uh I've gone through that that drill. So yeah, I mean, it's very fortunate for me, 100%. And uh if any kid today understands that part and you know finds uh his, her, or her way of going through the grunge, still, uh I mean, that kid is going to be the leader of the tomorrow.

SPEAKER_00 58:27

I absolutely agree with that.

SPEAKER_04 58:29

Uh let me let me take this uh you know to to a different topic. And I need to understand from you, and especially because you've also been on the VC side, uh, from a global global AI adoption in pharma, you know, where where do you see you know some of the best things happening? So which region I'm I'm talking about? So do you see Europe, uh, let's say, or the UK, the US, who, which which countries do you see, which regions do you see as the emerging markets where you think the the environment is conducive, the compliance is nice, the talent is nice, the investment is there. Where do you see the most impact happening or the most interesting things happening?

SPEAKER_00 59:08

Oh, I think it's India. It's it's I absolutely. I think India is a resting space for AI innovation in in healthcare. The other two markets where I feel there is a lot more interesting stuff happening is US and Brazil. So those are the three markets. I think there is there is a there is a very mature adoptive audience around digital technologies. There is um enough uh capabilities to build something which is meaningful, impactful. And um, all said and done, I think the the way FTA also responds to innovations and the innovative frameworks is actually quite promotive.

SPEAKER_04 59:52

Oh, I I never thought you would say Brazil. Well, can you talk a little more on Brazil? Why why would you pick Brazil? Uh I I understand India, uh, I understand US.

SPEAKER_00 1:00:05

Um I think it's a it's it's a very mature, uh digitally mature um population. And they do so much and so much interesting stuff with digital innovation and uh within that whole South America landscape. I mean they're they're a little bit on the they don't have the technical talent of India or the the the funding of US. So they don't get highlighted as much, but it's a very um it's a it's a very digitally safe suave ecosystem.

SPEAKER_04 1:00:38

Oh wow I I mean it this is the first time I'm hearing Brazil actually.

SPEAKER_02 1:00:44

Well interesting.

SPEAKER_00 1:00:45

Well you know, find some data against it.

SPEAKER_04 1:00:48

I would be very cool to be talking about I mean we app I mean we uh keep on looking for talent everywhere and uh for for obvious reasons you know it's it it gets difficult and difficult to find good talent. I mean, uh even in India for tech. I mean, because you know, I don't uh probably the similar points that we discussed, you know, it's not easy to find the right talent uh or right young talent fresh out of college in India uh today. So there is a very limited pool who are really good talent, but then because of the startups and the ecosystem just blasting away, you know, for every talent, they have 10, 12 offers. They're the right talent, right? And then there is a large amount of pool which sort of probably no one wants or just want because they're desperate about it. So we also sort of forced to look outside, and uh, we keep on sort of hiring people. Uh, we I just set up a center in Manila in the Philippines, and I never thought I'd find really good data science scientists in Manila. Uh, which which I do, which I did, you know, I'm I'm so proud about. Uh so Brazil is one thing which you know I would definitely now now look home.

SPEAKER_01 1:01:56

Yeah, yeah, let me know.

SPEAKER_04 1:01:58

Perfect. So, you know, we are we are the end of this discussion. Now, you know, before I let you go, let me ask you this. So if you could sort of magically invent one AI capability for pharma, uh what would you build uh that doesn't exist today?

SPEAKER_00 1:02:18

Oh one magic capability I think would be to do um Oh it's very hard. Are we talking about internal solution or external solution?

SPEAKER_04 1:02:40

Let me paraphrase, let me ask you this, and you know, I'm I'm a I'm a sucker for for gossip, so uh and that that should be a great question to end this conversation as well. What do you think is the biggest lie the healthcare industry or pharma industry tells itself about AI?

SPEAKER_00 1:03:00

Oh that um that you know like that um uh we are the you know like the adoption of AI is great. We are using AI.

SPEAKER_04 1:03:14

Using AI, that's the biggest lie per farmer pharma industry is telling itself.

SPEAKER_00 1:03:18

Yes, I don't think that AI penetration is as much as it's it we can call it as an adoption. There is some AI, but I don't think it's it's AI-led. I mean pharma industry is not AI AI led. In fact, I I don't think healthcare is AI-led. It's a very much traditional led where it's AI supported. Or rather, the okay, let me rephrase that. It's the the vision is that healthcare industry is AI supported. I don't think it can ever be AI-led. But today we are not even AI supported. We are in some places where there is isolated AI plugs in, and that's about it.

SPEAKER_02 1:03:58

Wow.

unknown 1:03:59

Okay.

SPEAKER_01 1:03:59

So there's a lot of speed for innovation, yes.

SPEAKER_04 1:04:03

That's obviously from where you're looking at whichever angle you look at it. I think uh that also means a lot of opportunities, but uh that also means uh, you know, there's a there's a long way to go. Uh Neha, this this was a really fascinating conversation. Thank you for sharing such honest insights about AI and healthcare in general. Uh and thank you everyone for listening to the best AI show in Pharma. And I think uh Neha gave us some of the best uh playbook rules on anyone who's trying to build uh for pharma services build sitting in India or in Brazil. Uh thank you everyone for listening. We'll see you in the next episodes. Bye bye.

SPEAKER_00 1:04:43

Thanks, Ayesh, for having me. Bye.