Gaurav Bhatnagar: The AI Reality Check

Ram Yalamanchili (00:00)

Welcome Gaurav for our ⁓ final episode for season two. And ⁓ I think it's been a really interesting last couple of ⁓ months for us especially. And ⁓ for just as an intro, Gaurav's a colleague of mine at Tilda we work very closely on customer development, product, ⁓ everything around the agentification of this space now in in research.

So I'm quite excited to bring in Gaurav and ⁓ sort of share our thoughts on what we've seen ⁓ in the past ⁓ several months. And ⁓ also as we've brought on more guests into the season two of this podcast, I think we've had about ten different guests. I think it's a great time to sum up and summarize many of the things ⁓ which we've spoken about and heard about and are also obviously living through our own experience with our customers. So I'll start with that, Gaurav. How are you doing and ⁓

⁓ what's what's on your mind as far as what you're seeing in terms of where AI is going in research?

Gaurav Bhatnagar (00:53)

Yeah. No, thanks Ram. ⁓ it's been, I think, an exciting six months to one year, if I were to put it that way. I mean, I think I've seen more transformational change actually deliver value in the last six months to one year than I've seen in my entire career of being a historian of the clinical research industry for the last fifty. I mean, what are some of the key areas, I suppose. One area there have been

you know, crests and troughs of technology and technology in the last twenty, twenty, twenty-five years in in the clinical research industry. But particularly now, how organizations are are very aggressively and actively approaching the the application of AI within within clinical trial operations, which is the thing which which I'm super close to. So it is it is different from before. So AI as not a tool, not as a toy, not as another one good to have.

But as part of their embedded team is what what it kind of resonates very across. I think what I listened to all of the different podcast episodes, different guests, kind of that's a central theme which has emerged, right? How, how, how the work can now run it itself while while we take a different role in terms of providing oversight as or providing or prova or or kind of helping with decision making more than than anything else.

So I think that's a central, central theme. The other part which is which is I think emerging clearly is that there is a different like the people who have who have or some of the I think the guests that you have had on your podcast is where that there is a huge difference between proof and potential, right? I mean there's a like everybody, every sentence of ⁓ of the English language and probably every other language has AI as the prefix. But but it is about what is, you know, what

There there's a very clear demarcation of what's working, right? And and so you know, people who have products and poop people who have implemented talk with a very different kind of a competitive mode, very, very different operational kind of results and an end game in in in many cases. So I think those are the two really central themes which kind of come out as as as this thing, which is what I think ⁓ which it resonated very well across, I think, several different guests that you have had. Polyus talks about it, heard his couple of his presentations even.

even in certain conferences as well. George talks about that that that and and and so on and so on.

Ram Yalamanchili (03:11)

Yeah. No, I I I think that's ⁓ that that's also my perspective. I think we've been able to see our own customers sort of come come in and say, hey, like this is not automating a small part of the ⁓ you know activity or something like that. I think we've had really interesting successes in terms of saying like this whole workflow can be rethought and you can bring in this AI teammate, it'll help you on the entire end to end workflow, complete the loop on it.

And that gives you these like tremendous values from an ROI perspective, from quality timing. So one other area I I think we also spent a reasonable amount of time over the past few quarters is ⁓ the fluency angle, right? So you know we've we've certainly seen customers who love to adopt AI. They they do believe that work should run itself for the most part with ⁓ with with human in the loop.

⁓ but fluency is interesting. We spent a lot of time. I'd like to get your thoughts on how you saw it and w were your takes on it.

Gaurav Bhatnagar (04:07)

No, I think fluency is is is is absolutely critical because I think the presumption of familiarity with AI, like we using AI in very simple everyday lives from searching things and and writing emails to getting to fluency is a whole different ballgame. And the organizations which have moved on in that like or are rapidly trying to move on and creating infrastructure around fluency are gaining ground and I think will create some

sustainable advantage over that. I mean, I think the c I I was particularly intrigued by some of the comments made by Shobith, who's, you know, who's a managing director in a large system integrator. And that's kind of in some sense is like how they're thinking about, you know, how do we integrate that within our organization structure and kind of truly transform and and provide the value that our customers are looking at. So I think there is, I mean fluency is is basically going to be the next, I mean, you know,

If if if we were to put together a continuum about where AI takes us, all of us as as we move this direction, I mean fluency is the next biggest kind of ⁓ area where which which provides clear differentiation from the organizations who catch onto that and the ones who don't.

Ram Yalamanchili (05:16)

Yeah, yeah. No, I a hundred percent agree. I think ⁓ bringing along your ⁓ your organization on the AI journey has to be you know, you have to think about it from an AI fluency perspective. And I think like Krishna from PPD also kinda resonated with this with this, right? He was talking about how it's ⁓ fundamentally a human and organizational challenge like AI transformation. And I think we couldn't agree more with that angle. So ⁓ move

Gaurav Bhatnagar (05:39)

And I think even within

that, I think there is there is the this aspect which so I was recently also speaking at a panel with a lot of stalwarts from from the biopharma, biotech industry, who are all kind of deeply immersed in their digital innovation. And I think it is about how do you tee up those conversations leadership down. So all those organizations where the senior leadership is invested in kind of bringing those those that change and fluency within their organizations are kind of seeing a very different

⁓ s kind of return right away, like not have to wait for a long time. They're they're seeing a lot of return right away. Yeah.

Ram Yalamanchili (06:15)

Yeah, so that's like the economics of it, right? So there is the you know, I think the promise of AI in a organization to help them do better, then the question of ROI comes in, then implementation of AI comes in, but then it's not just a tech problem because there is an AI fluency associated with it. So I sort of say, okay, like ⁓ so if the end goal is to bring this kind of transformation using AI to make your organization much more comparative.

Gaurav Bhatnagar (06:32)

Right.

Ram Yalamanchili (06:41)

much better at what you do then you're starting with AI fluency and the right technology. ⁓ so both have to come hand in hand. And I think we've had a few guests also talk about the economics of all of this, right? Like basically how this is all going to be justifiable internally in the organization as far as ROI. And also I think because we work with large Clonops organizations, ⁓ you know organizations with ⁓ large teams who manage Clonops, whether it be pharma or CROs,

there's a question of like how do the economics also ⁓ change in these kind of organizations because now AI is coming and there's there's some transformation involved in that. maybe I yeah, maybe I could start with what do you think? I mean there's there's certainly some change which we expect, but curious somehow what and what you're seeing.

Gaurav Bhatnagar (07:21)

Yeah, I think my own view is see, one of the things is that the organizations thus far in terms of buying technology, everything that has come prior to AI, has been on many ⁓ in in in on the most part has been like on the CapEx side of the column, right? This is this is providing them infinite operational leverage. So what they have, like the crests and troughs of drug development, of the bumpiness of a failure of a particular clinical trial or a molecule and going

This provides them that in finite leverage to do that. Economics has to work out. So I mean, I think our guests, like you know, Paula Brown Stafford, the CEO of Illusion, like as as as she spoke about that, it has to lower your ability to do unit costs. So you will are able to do a lot more using a lot less. And I think that's kind of already panning out if we think about it, right? When you have the infrastructure that is just super efficient.

An infrastructure which produces very high quality consistently, you're not worried about the ebbs and flows of that. And it kind of is metered to, I mean, again, token costs are kind of going all over the place and probably they will stabilize in a period of time. But but I think the clear value is how it is, you know, as I as in one of the discussions I was asked that, you know, what is the comparator here? The comparator is what your current set of cost base is, right? So AI pro provides a very distinctive advantage even right at this point in time.

very very significant way. And that's going to be increasingly higher as as kinda I think time goes by.

Ram Yalamanchili (08:46)

Yeah. No, Paula definitely makes a great point. I think she she's in the midst of that service layer transformation, right? I think a lot of the CROs, even for us, I think we've we've start to see these AI efficiency rebates being demanded from sponsors, from organizations which which are traditionally performing the services. And I don't see that trend coming down, right? I think it seems like this trend will continue to accelerate. I think you know

Buyers of these kind of large pharma services or or any type of pharma services will probably assume that you're gonna be more and more efficient with AI. And hands on a unit cost basis, they would like to capture some of that benefits through through a rebate, through an efficiency rebate, which to me is like very logical. and ⁓ and then the question is like how do you then go back to everything we just spoke about, which is the AI fluency, the technology, the investment, to be able to like prepare yourself for those kind of a

eventual ⁓ realities. So one thing I was curious about is you know I think there is a lot of at least last year I felt like there's a lot of talk about AI can do this, AI can do that. There's a lot of hype I would say around saying like, you know, there's a product we can we can sort of demo this thing. It's very easy to sort of ⁓ adopt it. But the reality is not that, right? We've we've made great progress.

in areas, but there's certainly some ⁓ some learnings here in terms of what what the what we've learned. w what do you think, like you know, in in terms of where where we are today, how that's evolving and where you think we'll go in the next ⁓ coming quarters.

Gaurav Bhatnagar (10:11)

Yeah, so I think where we are is right now we are in a stage in the last year, particularly when I say about when I say I use the royal view of the whole industry. I mean there's also a not a lot of noise. There's a you know everybody's saying, you know, AI this, AI that, and there is there's a lot of noise. And that's where I kind of like to double click. See, it's it's easy to tell a compelling story of, you know, a sci-fi type of a story. It's probably a little harder to build a prototype.

It's it's or or or or a or a demo. But I think a scalable product, that's what I think everybody should look at when they're thinking about, you know, what what do we do that? A scalable product which can work in an environment. And what what I think, you know, the conviction with which you know, one of your guests, Polyus is speaks, speaks to, right? He can he speaks with the with the conviction of that, you know, that their entire organization has kind of become, you know, is is is AI first in many ways. It's AI native in many ways. They have

They've reorganized their training. They're the people they hire. They have, and I think they have seen the benefits. I mean, their benefits are significantly. They have seen benefits in terms of every single dimension. The quality is way higher than a human dimension of quality can have any the best person you have ever hired to do TMF or study startup care. The productivity is 10x and more. So I think those similarly, if we think about so I think those aspects are significantly.

important right now to double click on. Like, you know, is there, are there products? And there is, and again, so to cull through some of those noise, asking the exact, you know, be able to do a quick pilot, be able to see that how does it work really in my environment? Because those are very different things than to be able to show something in a very, very controlled environment or or to show a demo or to talk about some of those things. And that's where, you know, I think the the people who are kind of taking bets towards that,

Or trying trying those things I think are are moving really, really far ahead and much

Ram Yalamanchili (12:03)

What's working in that regard? Like for example, I definitely think there's a lot of interest to try or bring some kind of AI into organizations. But we also know some organizations are leaps and bounds ahead of others. motivations are the same. Maybe at the at the C suite the motivation is the same. But in practice there is there's a difference. curious, like what w where do you think the the gaps are? Where where where are some of these things?

differentiating.

Gaurav Bhatnagar (12:27)

So I think the major gap is that there is ⁓ major gap, particularly in larger organizations today. I think what what we're seeing is is the is is is a little bit of ⁓ inability to connect the core operations with the leadership vision. So where both of those things have been kind of bought into or created, I think that makes that that's the fastest way because you get a set of early adopters, you get some some some ⁓

wins under the belt and then you know they can't those teams can't welt, can't wait to virally infect everybody else and the rest of the organization and kind of it it does. It it kind of flows through very, very, very quickly. Where there is where where so I think that's one lesson that that you know even we have learned through our through our experience in the last year as well. I think the other key lesson is that wherever there is a lot more you are looking or

Use the challenges of the barriers are higher. If you are viewing this technology with the lens that has existed for the last 30, 40, 50 years, right? If you are not ready to be pragmatic, quicker, understand how to analyze, how to assess the the evaluate the these technologies on where should I place a bet? How do I try this out? How do I see if it will work and what will I do? Wherever those frameworks or that thinking from a

Leadership downwards is not there, that's where the barriers are. I mean, if you're pushing it through an age-old vendor qualification process, or you have QA auditors who don't fully appreciate what this technology is really about. Like it's like I say that, right? It should be compared to what the alternative is, right? I mean, when you're comparing when when we when we think about going through some of these things, compare it to what's the alternative otherwise you would have had. Your alternative, if it's a person, what is the auditability, traceability of the pieces of information? So

Where that so so being pragmatic and really adapting and evolving your your your own processes to think through those pieces, I think that's where people c organizations which are doing that are able to move really, really fast and are able to get those results that they're looking for. The organizations which are s which are still evaluating it with the with the

I wouldn't use the word skepticism. I think everybody should be skeptical to the because you know we are in a really regulated and a really important critical where we generate patient data through through these clinical trials. But I would use where who are not pragmatic to truly understand, okay, what's this compared to? How do I think about these things? Who are not able to bring bring in those decision-making processes internally and are using somebody else's, you know, or using press information or things which they're hearing in on social media.

to to to to kind of inform themselves rather of first principle thinking are having more challenges.

Ram Yalamanchili (15:10)

And I I like what you just said ⁓ in the first point, which is the the virality which actually builds within the organizations. I I actually think this is one of the exciting parts of what we are where we are, vantage point-wise. the the learning, at least for me, is there is healthy skepticism. there's been a lot of promise under delivery on products. And when products do work, then your customers become your advocates.

And they're out there talking about your products, they're presenting about your products, because it really does operationally change the whole game for them. And I think Polyus from Perceive is probably a great example, and there's others as well from our customer base. but I also think which is sort of under pressure, at least for me, ⁓ before I got into this ⁓ where we are ⁓ in the last year, is in large organizations, once you actually deliver and you win the trust of a particular side of the team or ⁓ you know.

part of organization, how fast diffusion is starting to happen in these AI deployments, right? we've ⁓ we've sort of seen that and I think that's another interesting use case or or or case study here, which is a learning for us, in terms of how once you build the trust, I think there's sort of like a a floodgate effect ⁓ in terms of how fast things are ⁓ then moving.

Gaurav Bhatnagar (16:19)

Yeah, no, absolutely. I think the the the rapid I mean I mean I suppose let me let me kind of make a slightly slightly different point. One of the one of the things where particularly the operational AI or the agentic operational AI is dramatically different if you ha if there is a product and the product works or not, the feedback is immediate, like almost instantly.

Right. There is so your your ability. I mean, I think there should be nobody who's doing clinical trial operations anyway, should be hesitant in trying a product. Because if it doesn't work, you you know, you can move on. Right. I mean, you should move on. Right. Right. If it isn't better than the alternative, which is your which is which is basically being being able to do that in in in in some other manual way, you f can easily move on. This is not like, you know, let me tell you where the sites are and you'll learn after four years of

Of of trying to select the sites and seeing the recruitment and seeing if patient recruit. It does not like that. It is not about a a pontification of the future in which kind of generates within a certain volume. Here you will have the feedback as an instance. It's like if you put a put put put put somebody to do a certain job, if they either they do it or not. The feedback is in your face right there. And and that I think is is I suppose, you know, I think that should break the barrier of anybody wanting to try it.

Ram Yalamanchili (17:37)

Yeah, that's a good point. the feedback cycles are much faster. And I think they're more easier to understand as well in the AI world. Because ⁓ I think we've all gotten accustomed to a certain level and quality of work. and the moment something doesn't meet that, you you you immediately know. And and and that's that's right. Yeah. Well, I appreciate you sharing some of the thoughts here.

⁓ I think you know, I encourage ⁓ our viewers to go listen about these stories from the thought leaders, from the customers, from the CEOs of ⁓ major ⁓ pharma or biopharma or CROs, right? We've got a mix of them on our podcast. And I think it's a testament to sort of how fast this industry is moving and how much ⁓ how much of this is actually becoming very real in the as we speak. So it's a very exciting time.

I think Gaurav and I have been talking about this and working towards this ⁓ sort of a a vision for many years now. And ⁓ I I couldn't I couldn't be happier in terms of like where where we are in terms of like the the market itself moving towards this vision, right? So ⁓ so thanks, Gaurav, and have a good one then.

Gaurav Bhatnagar (18:38)

Thank you.