Accelerating Speed to Value through Agentic Systems and Intelligent Automation in Life Sciences - with Robert Wenier of AstraZeneca

Welcome everyone to the AI and business podcast. I'm Matthew D'Mello, editorial director here at Emerge AI Research. Today's guest is Wenier, global head of cloud and infrastructure at AstraZeneca. Robert joins us on today's show to examine how enterprise infrastructure is evolving from service oriented and microservice architectures toward emerging service agentic architectures and why this shift changes how organizations move from data to business outcomes. Throughout our conversation, he explains how AI is pushing technology beyond a supporting role and into a position where it directly shapes competitive advantage and decision speed.

The discussion also explores practical implications for enterprise leaders, including collapsing multi step workflows into direct end to end outcomes, reducing manual data logistics through AI agents, moving decision authority closer to teams and individuals, and reframing ROI around speed to value rather than incremental efficiency gains. Emerge is also very happy to announce that the AI in business podcast is moving to video. You can search for AI in business vision to value in enterprise AI on YouTube where you can see the video version of today's podcast. You can also find a link to today's episode on your preferred podcast platform. That's youtube.com/@emergeairesearch.

Again, that's youtube.com/@signemerjairesearch. Watch the full episode now and see how vision becomes value in enterprise AI. But first, are you driving AI transformation at your organization, or maybe you're guiding critical decisions on investments, strategy, or deployment. If so, the AI in business podcast wants to hear from you. Each year, Emerge AI Research features hundreds of executive thought leaders, everyone from the CIO of Goldman Sachs to the head of AI at Raytheon and AI pioneers like Joshua Bengio.

With nearly a million annual listeners, AI in business is the go to destination for enterprise leaders navigating real world AI adoption. You don't need to be an engineer or a technical expert to be on the program. If you're involved in AI implementation, decision making, or strategy within your company, this is your opportunity to share your insights with a global audience of your peers. If you believe you can help other leaders move the needle on AI ROI, visit emerge.com and fill out our thought leader submission form. That's emerge.com and click on be an expert.

You can also click the link in the description of today's show on your preferred podcast platform. That's emerge.com/expertone. Again, that's emerj.com/expertone. Without further ado, here's our conversation with Robert.

So Robert, welcome to the program. Thank you, Dan. Yeah. Glad to glad to be able to unpack this topic with you. As I've mentioned, sort of the big picture of infrastructure is more and more the rage in terms of our enterprise.

Audience implies more and more things, and you're at the helm of a rather large organization dealing with infrastructure infrastructure on a day to day. I wanted to kind of open with, some of the permutations that this takes within an enterprise environment and sort of ask how business goals shape infrastructure design. It's it's not gonna be the same for every department, every business. Give us maybe your walk through of what you think leaders need to understand about that.

So giving leaders a kind of this perspective of how it has evolved over time. If I go back, you know, a couple of decades, it really started with service oriented architectures, and this was the history of the three tier stacks. You have a front end of a system and then you have the data being processed and then you have some kind of storage volume. And the whole idea is that you're building systems back then that perform a logical action. Now this logical action was an end to end to means, it was just a step.

And normally that date, you know, you put in data and you got an output, and then you put that output into another system that would put another output. And somewhere along the lines at the end of the day, you would actually have the answer to the question you're ultimately looking for. This was kind of a beginning of, you know, the ability to get to that end state. It evolved in time into what we call service based architectures, which is the emergence of event trigger type of architecture. This is where you get into microservices and serverless and containers, where instead of it being in that traditional monolithic design, it starts to be, you know, different fragments performing, you know, a small but very finite, you know, that has a beginning and an end to end process to deliver on a a capability and that that ultimately is getting you to your answer a little bit faster than a service oriented architecture did.

And at the same time, leaner, cheaper, a little bit more environmentally friendly, those kind of factors. Now we have, you know, the emergence of a brand new paradigm on top of that now, which is starting to be coined service agentic architectures where, you know, you have the infrastructure is taking on a major change now is you're looking a little bit less at that step by step evolution and mostly getting to what is the end state we're trying to get to. How quickly can I get to that? I have a question, but it's not a, you know, it's it's not a step question. It's the very end of it is how do I get to the end of my strategic ask?

And that is what the infrastructure is starting to be designed to. Certainly, it's a journey to get there. All this stuff is massively new, but we've been on this journey for several decades now.

Yeah. So you kind of talked about it as, you know, a new paradigm kind of sitting on top of that. It almost feels in some ways like it's a a new paradigm that's taking over the the center rather than being laying over layered over the top. Right? Because it almost feels like infrastructure will fundamentally have to operate differently.

And now in early AI for the last decade has mostly been kind of playing on the surface of very ancient systems. Right? Let's pull a little data here, pull a little data there, and it's like another topping on the pizza. Right? Just keep stacking the toppings on the pizza.

It feels like we're moving towards more of a lasagna here where sort of data waking up is more at the center of sort of what we're doing infrastructure wise. Do you think about this sort of agentic paradigm you just articulated as kind of draping over the top of those older systems, or do you see it as somewhat kind of emerging and bursting through the middle here? How would you frame it analogy wise? I think

the biggest difference is historically, you know, the evolution of the technology has been in the realm of the technologists. We are achieving, technical outcomes that will ultimately drive a business outcome. With the emergence of eugenetic architecture, it's it's it's different in that it's outside of the realm of the technologist to a certain degree. Now you have business leaders looking for business outcomes being facilitated by technology. So, you know, in the look at it from that perspective is not necessarily looking for the technical outcomes anymore.

We're looking for the business outcomes. I talked about getting to that strategic deliverable faster. You know, we we had a little bit of an evolution. I talked about the the the service based architecture design. That was technology, driving technology.

You know, it was a technical outcome for technologists. The eugenic design and the a the way the AI is being consumed now, not the way it has been in the past was which was also an element of technology for technologists. This is different now. This is the business is expecting a business outcome directly from the use of technology. So now technology is becoming in and of itself a business discriminator where it used to be a business enabler.

That's a big difference.

Business discriminator versus enabler. Can we double click even more distinction between those two? I almost I'm imagining in my mind almost an infographic here Robert of like, here's what it used to be, here's what it is. I want to make this crystal clear for the listeners and viewers like, how do you draw that line? Like what what are the the the the core cruxes that separate those?

So when I talk about business enablement, I talk about the old paradigm. I say the old, it's not that old, it's it's been enforced for several decades. But it's the the paradigm where the business requests a certain type of outcome. And we know the technologists will build applications and systems and capabilities platforms to achieve a certain end. It it enables business users to input data and get a certain type of output.

Now they will take that output and either do research with it or do some follow on action tie and you've rinse and repeat this several times, and eventually you'll get a particular business outcome. With business as a technology as a discriminator for the business is we're bypassing all that now. We're actually going directly to the business outcome and using the technology to get us there. So no longer is, are we would necessarily look to the technologists to build these discrete platforms, to get us on a step by step answer to our problem. We want to actually go from the beginning to the end of the problem in a single swap.

And how fast we do that and how effectively we do that is now a point of business discrimination, not necessarily technical discrimination. I could change the entire corporate landscape.

I'm I I do not disagree whatsoever. And I guess I'm interested from your vantage point, what are the keys to making that shift? None of this stuff is overnight. I mean, you know, admittedly, AI capabilities have leapt to such a level that the paradigm you're articulating is now viable. Right?

I think five years ago, maybe there were a handful of people that would have seen things going this way, but I I don't think it was as palpable as it is today. This is a journey. It's it's gonna take a bit to to sort of realize this this vision. What do you see as kind of the core stepping stones that enterprises are gonna have to chunk their way through to get to that place of more immediate business discrimination value versus this kind of stepwise technology people build the the training wheels for us kind of a kind of a ballgame? Like, what what fundamentally has to change, to bring about what you're saying?

So in my opinion, it goes down to the way that value statements are made or what the ultimate business objectives are. It used to be that we carry those at a corporate level or for example, is a major point to the business unit level is that's the value statement that entire business unit is going to achieve. With the way that AI is being consumed nowadays, it's moving down to a more personal level or a team level is what is that individual person or what is that individual team going to deliver as a value? And I and I like to use the heuristic of, it used to be a history of everybody was a cog and a great machine. Well, no more because each individual team or even in some cases individual themselves is now the machine.

And there's expected to be a business deliverable, not a technical deliverable, but a business deliverable that comes out of that individual unit. That's a big sea change. Now how quickly we can go there depends on a few different factors. The first is the companies and the business units and the individuals themselves have to have access to the tools and technologies that will enable them to reach that end, and they have to understand how to use them effectively. The second one is depending on which type of toolset they have, there are certain dependencies that they're going to have to address.

If you are looking to build your own proprietary models or just, you know, basic way of saying is we wanna energize all of our the data that our company has created over decades and be able to start gleaning direct value out of that. Well, that data is probably dirty. That data probably has all kinds of problems with the way it was stored, the way it was kept. You know, a lot of that has to be cleaned up in order to derive the maximum value out of it. That's a pretty big lift.

In the old days, we used to call that data cleansing. Well, this is at a much bigger scale than that. So we need AI's help even to do that. The other side of this is with the advancement into capabilities such as co piloting, a Gentex general knowledge models. So we're actually leveraging the things that are being designed per se outside of our company.

We have to know how to do that effectively, and we have to make sure that they're going to deliver the value statements, individual value statements, like I mentioned for that unit that we're asking for.

What what is the main driver that's permitting sort of, the machine to go from the department or, you know, a certain team or something down to more potentially even the individual? Is it is it simply the more wildly accessible, nature of the technology? We're not we're no longer having to hard code all this AI stuff. We can kind of conjure forth insight and action and code from these systems. Is that the primary driver of shrinking that sort of unit of production from, the team to to the person, making people less of a cog?

What for you is undergirding that shift the most?

I think it's, you know, the ability of a lot of the artificial intelligence to do the heavy lifting, the legwork on a lot of things. I I use the example of an accountant who uses Excel to manage numbers on that. You know, what is normally happens is those numbers are providing from another source somewhere, a database for example, a report for example, they input that into Excel, they manage it in different ways, create pivot tables, all kinds of done analysis of it, and then the results of that is then fed to another system. That might be the general ledger system or something like that, and then there's an analysis level of that. All those moving parts I just described are logistics.

We're taking data from one place, we're doing some manipulation of it, and then we're moving data to another place. That is I don't wanna call it a menial task because it's hard. I've been I've been there. I've done that kind of stuff in my part of my life. But at the end of the day, if I'm seeing revenue coming in, say, example, to a bank account, what I wanna do is some analysis on that right then and there.

I don't wanna output that into Excel and output that into a general legislation and etcetera etcetera. I wanna see it right then and there, and I can build artificial agents, intelligent agents to do that for me. And say, I wanted to know right at the source of the data. This is my bank account. Tell me how much of this is based on interest, how much is paid to revenue, how many things like that.

So and and that again, I guess that gets rid of because there are entire positions that are primarily that necessary shoveling to put things into formats that then we can make decisions. Right? There's entire positions that and and I guess what you're saying is those are more cog like roles by their nature. While if you can look at the torn data and ask the questions that are gonna support a business decision, now you can be that unit of value because you're empowered to do the decisioning, you're not just doing data cleansing, wrangling, formatting, whatever. Am am I roughly nutshelling you here?

I I think that's an accurate way of saying it is we're moving folks, and and they and in a lot of cases, they need to be upskilled, and there's nothing wrong with that, is we're moving folks into the capacity of making decisions. And when I say decisions, I don't mean tactical decisions, like whether or not we turn the the valve open or we return the valve close. Tactical decision is we are only running at a 11% margin rate right now. I think we should be at a 13% margin rate. What do we have to do to achieve that?

And, you know, we may leverage several different systems and artificial intelligence agents, other factors like that to achieve that, but it moves these folks into making those types of decisions. And the requirement would be to upscale them into strategists, into visionaries, into people that are making decisions based on the information presented by the agents doing the logistics work.

Yeah. So okay. So, again, that's that's putting a a pretty good pin on, like, the fundamental shift in the nature of work. I imagine there'll be some people in some organizations more more sort of willing to ride this change and and, take advantage of those opportunities than others. What do you see as kind of I mean, the the the obvious immediate top of mind here is like, okay, Robert.

I'm following you. It seems to me like if you're a company that keeps people doing cog stuff when it's actually not necessary anymore, at some point, your overhead will be more and your speed will be so slow that it will not be tenable to run your business that way because other other competitors are going to be adopting and changing and and moving faster. And so, okay. It makes sense to do this. That that seems like somewhat of an obvious conclusion.

But when you think about what's at stake in terms of making this shift away from the previous paradigm and into the one that you've just articulated for us, what do you see as the the upsides, downsides of of why enterprises, you know, ought to be moving in this direction?

I think if you look about on business strategies, and this actually goes back, you know, quite a ways, in more than fifty years, is, you know, profitability used to be based on efficiency. How efficiently can we do what we need to do and maximize our profit by reducing our our operational costs and other factors like that. I think it's switched with technology. My firm belief is technology is its only reason for being, to a large extent, in my personal opinion, is speed to market, is it's all about speed. We can do it manually, and it may actually work, but that may not be what we're trying to do.

So what's happening right now is all of these technologies are being employed or are about increasing the speed to market. And speed to market is not just a euphemism for, you know, getting subject to the to the market quickly. It could be for just getting tasks done to value, achieving value as quickly as possible. That's what's driving a lot of this. So if you can relegate things that are menial, it's mostly logistics to these type of artificial intelligence and agents, and you can get to making decisions that will create value faster, that's the driving force is how do you outspeed, outrun if you will, you know, your competitors.

Yeah. I mean, you know, some people say, you know, technology is sort of taking the number of human inputs. You're reducing the number of human inputs to get to what the output is or or or whatever the case may be. Like, it's it's it's leverage. And in your case, you're you're talking about speed.

Now speed, some people would say, oh, that sounds like efficiency. I can get more of this stuff done within one hour. But I think what you're saying is it is speed to a decision. It is speed to a new product. It is speed to a new strategy.

It is speed to adapting to the market. It is speed to the things that drive your top line, your bottom line, your risk, everything. It's not just shaving cost. So, and I think our audience is very congenial with that notion. Certainly, they've they've had it hammered into them over the years here that this stuff is not just a glorified RPA for all of eternity.

This is actually, new new ways of doing new things. For you, in terms of what's at stake, it seems like, okay. Well, the companies that can do that will, like you said, outspeed or outpace, their rivals and be able to satisfy customers and fulfill their operations and kind of do the things they need to do and actually stay in business while others will sort of fall behind. When I look at legacy enterprise, big complicated machines that have been built over, you know, a hundred something years, I I sort of ask myself, what's the tipping point where a paradigm will be embraced? You're to kind of articulate a paradigm for us and I think about what's the typical one that makes a paradigm be embraced.

The cynical part of me would say, okay. Well, basically you need either startups or slightly faster moving rivals to be eating your lunch in a way that hits your p and l and maybe then, you know, you take away these cog jobs or maybe then you and then and I'm not saying I'm I'm an eternal pessimist in that regard. I'm just saying I think the pessimistic sort of if it ain't broke, don't fix it, little bit of a defensive attitude towards change within a very, very large company would be that. What do you think is going to catalyze this shift away from a bajillion cogs and a few machines to a lot of very productive machines? Like, what do you think is going to make that needle move?

I

think the competitive pressures are going to intensify significantly. It's not there right now because the ability to fully industrialize the capability hasn't been realized yet. I think that's just around the corner here in a lot of areas. But I also will say is it's evolving so very rapidly that we're trying to trying to put a stick in the sand and say, okay, we've reached this point is a little bit nebulous. I've looked at just to changes in the last nine months and I feel like I have whiplash a little bit.

But one of the things I would say is, I talked about profitability, be the ability to advance products to market, and begin to drive more value out of every single thing turned back to the next product that is in the pipeline. The competitive pressures are intense there and may drive a lot of those changes. One thing that we see in pharmaceutical industry is SG and A burden is a problem. The more SG and A that we spend, the less we're spending on research and development. And that's that can be a problem.

Now sometimes SG and A spend, it's extremely profitable if you're looking at from a point of marketability and commercialization. But other types, it can be a major impetus, or I'll say impediment. Yes. Thank you. To to actually producing that that value.

So, you know, a lot of companies, mine included, are looking at that as JNN burden and say, we we have to go faster. We have to be, you know and and again, it's not efficient. We have to speed to value faster. We can't go through all these laborers processes in order to get to the end state. We need to streamline that and get there even faster so we can reinvest those funds in other opportunities that are also going to be streamlined and also move to market faster.

A lot of them are saying, well, you know, we just add AI and AgenTex to our our capabilities, and that's gonna keep us profitable for years to come. I don't think they're reading the tea leaves correctly.

Yeah. What is a more accurate reading of the tea leaves from your vantage point?

So I think, you know, and this is the big sea change in a way, you know, from an infrastructure perspective that that I can talk about in that, is our traditional way of achieving the outcome, not the technical outcome, but the business outcome is changing very rapidly. There's the emergence of this service agentic architecture in which agents, artificial intelligence is serving as a central component, the center of gravity of performing some kind of a logic assessment, or a strategic assessment. We used to build these large systems and we connect into SaaS components to do that, large HR systems, for example, I won't call any out, but the the ones that manage personnel records or maybe payment records and other factors like this, it's that same Excel paradigm I was telling you about. Data is being pulled from one location, managed in another, and then pushed into a third location and so on and so forth. If you have agents that understand the data, that can interpret it, that can answer questions about it, can, you know, make modifications, and you don't need a large SaaS platform to do that, that may be the direction of travel.

Yeah. Yeah. That may be the trajectory we're presently on. You brought up an an interesting point. You said, hey.

You know, I really think the competitive pressures are gonna drive this. Again, I would completely occur. I think completely proactive self disruption seems to be outlandishly rare within legacy enterprise. And that's not to their discredit. It is what it is.

I'm just stating things. You know, looking at enterprise AI for eleven years, that's my take. And you brought up the competitive pressures are kind of speeding up and you you kind of went quickly over and I just want to double click on something like, yeah, you know, the ability to kind of industrialize this stuff from scratch or something akin to that. Like, it's not quite there yet, but I think it's getting there. Did you mean by that, like, the ability for a smaller firm to be able to kind of replicate the the innards and core functions of larger pharma in in more of a a fast pace?

Or what did you mean industrialize in in a particular case we're talking about?

Industrialize is looking at it at a much larger scale, and that's achieving the end to end business objectives using the technology. I think what's being looked at right now in a large scale is achieving a technical end using the technology, but that's not really where we're going with this. We're going to achieve that business end. For example and I did watch a YouTube video you recently where I watched a young developer create an agent and live on one of the platforms. And it's very simplistic, but it only performed a certain process.

It was a technical end. I wanted to send a text message in the event of such and such. I'm talking about industrializing it at a scale of delivering an entire product from conception, design, manufacturing, all the way to the end to even commercialization of the factors like that. What this means is anything that's required of technology that we we commonly use day to day, application servers, databases, storage volumes, things of this nature, if that's abstracted away from us and turned over to an agent that it knows where to go, we say we wanna know this information, we wanna make decisions based on this information, you have to package it up for me and give it to me so I can make a decision based on it. Does that mean that we really need to know what database technology it is?

I would argue not necessarily. For a lot of the managed services consumed in cloud environments nowadays, the underlying silica, you know, there are certain people that say, want it to be a NVIDIA chipset. But sometimes on the managed services, you know, I use AWS for example because they manufacture their own silica. Do you know? Did it give you what you wanted?

It gave you what you wanted. Did you really even care? It might have been cheaper or faster. I don't know. But that's the direction of travel is, am I as concerned about what database technology it was?

Maybe not.

I yeah. I would say big picture, generally speaking, the enterprise AI buying cycle that we see, the people who really care about tech x or y generally are not the people cutting checks. The people cutting checks are are the people that are responsible for how it smacks the p and l and how it smacks like market share and other things like that. And and they they really don't care if it's if it's a very reliable process where 2,000 monkeys, you know, throw darts at a wall. It it's it's it's it's Google Cloud.

It's that. It's whatever. What's the cost? What's the output? How reliable is it?

How much can I count on it? I I think you're absolutely right there. I I wanna maybe wrap on this little quick point because you're you're painting a paradigm that's appealing and really does feel like where the wind is blowing. We've only got a couple minutes, but I wonder if there's a distinction from you on this. As you look out in big companies and you work within some very large companies in in your career here, and you say, here is what I think is gonna be the core difference between kind of winners and losers in moving to this speed to value sort of paradigm, breaking out of kind of the the cog land and the technical deliverables and being way faster to to the point where they can survive this AI age at least in the next leg of it.

What for you is a differentiator between winners and losers in in that in that break there?

I think the the biggest one is going to be those companies that don't feel like they're being held down by the traditional vendors in this particular space. I'm speaking very much from a senior, you know, a technology executive. The business doesn't care about that. The business wants the outcome to be achieved as quickly as possible, and they expect the technologists to be able to do that. And if somebody else is doing it faster, they have to ask the question, why?

What are my technologists doing that is slowing me down? That that doesn't make any sense to me. I I see that as kind of being the the the race of the future. And when I say the future, I mean now. Starting now and and going on.

The race of, you know, tomorrow. Right? I mean, it's like right here.

It's the adaptability that's just gotta be there. And, you know, one of the things that used to be done with the, I would say, the traditional IT establishment is you manage IT through licenses. You manage them through strategic partnerships that were aligned, know, the old saying nobody ever got fired for buying IBM kind of thing. Those days are gone. Now you go to where the market is driving you to and to effectively achieve your business end, if that means you go to, you establish agents, I'll call them orchestrator agents and those agents go off and do whatever it needs to do, interface whatever they need to do, to give you the answer you need to make a decision that's gonna increase your profitability.

I shouldn't have to sign a license agreement for that. That doesn't make any sense anymore.

Yeah. Well, hopefully for those of you tuned in here, some of this kinda really sinks in a little bit. I think that the the paradigm you're articulating, I think is a really good way to phrase, the the the cog and the machine analogy. It's a good way to phrase sort of some of the phase shifts that are coming. And I suspect that what you've just laid out as a differentiator will in fact very much be what separates many of the winners and losers.

I'm wary of where we are on time here, Robert. I know we've got to wrap up, but I'm really glad we got to catch up today. And thanks so much for being on the show.

Thanks for having me, Dan. I appreciate it.

Wrapping up today's episode, I think there were at least three critical takeaways for enterprise CIOs, CTOs, chief data officers, and senior business leaders responsible for AI strategy to take from our conversation today with Robert Wenier, global head of cloud and infrastructure at AstraZeneca. First, enterprise AI is moving beyond stepwise technology led workflows toward outcome driven architectures. AgenTic systems shift the focus from building platforms to achieving business results directly, making speed to value a critical measure of success. Second, AI is changing where value is created inside organizations. As data logistics and coordination work are automated, teams and individuals are increasingly expected to deliver business outcomes rather than technical outputs, reshaping operating models, accountability, and skill requirements.

Finally, competitive advantage in AI will depend on reducing decision latency. Enterprises that design data and AI systems to shorten the path from insight to action will outperform those that remain constrained by legacy processes, tooling, and incremental optimization mindsets. Interested in putting your AI product in front of household names in the Fortune 500? Connect directly with enterprise leaders at market leading companies. Emerge can position your brand where enterprise decision makers turn for insight, research, and guidance.

Visit emerge.com/sponsor for more information. Again, that's emerj.com/sponsor. If you enjoyed or benefited from the insights of today's episode, consider leaving us a review on Apple Podcasts, and let us know what you learned, found helpful, or just like most about the show. Also, don't forget to follow us on x, formerly known as Twitter at Emerge, and that's spelled, again, e m e r j, as well as our LinkedIn page. I'm your host, at least for today, Matthew D'Amelo, editorial director here at Emerge AI Research.

On behalf of Daniel Fagella, our CEO and head of research, as well as the rest of the team here at Emerge, thanks so much for joining us today, and we'll catch you next time on the AI in business podcast.