In interview with Vikrant Bhan,Global Head of Analytics, Data and Integration at Nestlé, hosted by Dana Gardner, principal analyst at Interarbor Solutions we discuss the transformative potential of agentic artificial intelligence (AI) in re-imagining end-to-end business processes.
The conversation also delves into the evolution from business intelligence (BI) to AI, the advantages of using Snowflake's modern data cloud environment, and the necessity of change management and adoption for realizing AI initiatives. So please enjoy this interview between Vikrant Bhan and your host, Dana Gardner.
Dana Gardner: Welcome to the Data Cloud Podcast, Vik, we're delighted to have you with us.
Vikrant Bhan: Hi, Dana, nice to be with you all as well.
Dana Gardner: Tell us about your role at Nestlé, your background, and what excites you about the future of data analytics and AI.
Vikrant Bhan: My role at Nestlé is quite a complex one. I am managing what is called analytics, AI, data and integration globally, across the scope of the 180-plus countries that we operate in. I'm managing all of the platforms, the data assets, and our data domains at an enterprise level, as well as all of the AI analytics services and products.
As for what excites me about the future, we have been doing AI to optimize a lot of our functional capabilities over the years and obviously having a very value-driven approach to choosing which problems to solve for the business. But now with what we are seeing -- especially with agent AI -- the big opportunity within our company -- and generally across the industry -- is that it's not just about a particular capability, it's about the whole end-to-end process within Nestlé.
That cannot be optimized because of something that has been done before but completely reimagined with the help of AI. And that re-imagination has a massive implication on the operating model of our processes. And it also has a massive implication on our interactions with consumers and customers as well as internally within the company from an experience perspective.
That’s what excites me the most about what's coming in the future.
Dana Gardner: So not only process re-engineering, but cultural re-engineering, which sort of gets to the heart of making things better in holistic ways.
Vikrant Bhan: Absolutely. What we say here at Nestlé as part of our enterprise digital transformation is that while the processes will be reimagined, there are four key fundamental enablers that always need to be thought about.
As you rightly said, the people, capabilities, and culture are very important, but so is the operating model. Sometimes in the past when we have done optimizations or various kinds of analytics and AI use cases, that model gets lesser importance than the tools, data, and AI capabilities.
You have to keep in mind that data and AI are key enablers along with technology. And until all these ingredients come together, I don't think that re-imagination is going to work.
Dana Gardner: For many large, diverse, and complex businesses like Nestlé, end-to-end and ubiquitous data flows are essential, as you were saying, to enable better and faster decisions -- and deliver a real difference to the business.
After 18 years at Nestlé shaping the IT and the data strategy, Vik, how does this point in time where we are right now stand out in terms of reaping the rewards from all the past data maturity efforts? Are we actually seeing an acceleration in additional business value now?
Vikrant Bhan: I think so. One aspect that is different than the earlier part of my career is looking at our internal processes, at least at Nestlé, and hence a lot of focus on business performance, key metrics, and how that requires a lot of internal data from the company. What we are now seeing is the end-to-end view from a business architecture perspective. Not only are we looking at data that is impacting the consumer journeys and channels, but we are also focusing on the back-office processes within the company, such as operations, finance, and HR -- and just everything.
What has changed quite a lot is that data has gone from being just focused on your internal data to a lot of external data and signals, too, and this has been a journey that I've seen over the last four or five years where we have matured because managing that data is not easy.
A lot of the external data we acquire needs a lot of art and science to manage. Then also what has changed is the interplay of all of that data. In the past you would require maybe for an individual use case or a project data that is required. I always think that we were a data as a service organization.
So, what data do you need? We give it to you. I think what we have now seen is a very, very mature data product organization. So how do we actually treat data as a data product itself? That's been one major shift that I've seen. The external data and internal data combination has been another shift.
What I have also seen is the importance of master data, which used to be focused on key internal things like core product, core customer, core vendor or business partner-type definitions. With all the complexity of external data, and other data that is coming in, that master data has also become very rich, and you have to think about all kinds of relationships and that critical master data in the company has also increased.
When I think now with GenAI, Dana, what I'm seeing is something, to be honest, I didn't ever have to think about in the past. And that is how to deal with unstructured data -- especially data that is what we would term as knowledge within the company, not necessarily structured data as such.
That’s become extremely important. And, finally, I also see, although, in very limited spaces where we had data gaps, how can we also start thinking about synthetic data creation? What we have done at Nestlé, Dana, with the help of modern data cloud platforms is create our enterprise data domains.
And essentially these domains are a combination of master data, internal and external structured data, and unstructured data. Because at the end of the day, AI needs all of it for getting the best for the company.
Dana Gardner: And finally, after many years of standing on the shoulders of other technologies and bringing together more and more efficiencies, we're finally at the point where we can bring all of the data, resources, and increasingly knowledge resources together to be optimized, to be accessible, and to be managed.
And that truly does mean a whole greater than the sum of the parts, and that is quite new and interesting.
Vikrant Bhan: Absolutely. And also, an element there is about reusability and interoperability because there were cases in the past where the same data might have been replicated multiple times across the ecosystem of the company in different digital initiatives.
What the enterprise data domains have allowed us at Nestlé to do is to genuinely create data products that are reusable and used by multiple products. One of the ways we measure the success of our data strategy -- apart from many other things -- is on how often we reuse the same data product, rather than creating new ones. And, while it helps the company on the total cost of ownership and efficiencies, financial operations, and various other dimensions, it really allows getting more and more in intangible and tangible value out of the same data products.
Dana Gardner: In effect, we're using the data and analytics, and now AI, to do data and analytics and AI better, faster, and cheaper, right?
Vikrant Bhan: Absolutely, absolutely.
Dana Gardner: Nestlé is one of the best-known brands in the world, but I'm going to guess that many of our listeners, readers, and watchers don't necessarily know the full extent. Can you give us a quick encapsulation of what Nestlé is, what Nestlé does, and how your organization delivers a pervasive value across all of those businesses?
Vikrant Bhan: I'm pretty sure most of your viewers would at least have had a touch point with Nestlé at some time in their lives. When you're born, for example, you are obviously into a baby food business, and if you are potentially old, you are still using some of our health sciences and life sciences products. And in the middle, you are potentially indulging in a lot of amazing food products from our company.
Nestlé is one of the biggest consumer goods companies in the world. It's the biggest food company in the world. We have more than 2,000 global brands, close to around $90 billion of revenue. And some of the top brands, which I'm sure you would have heard about, are NESCAFÉ and Nespresso. KitKat, if you love your chocolate, and as I said, 2000-plus brands globally.
And such diversity means one IT solution doesn't fit all. You need to be centralized when you can, but you also have to be very customizable and focused at the diverse edge as well.
Vikrant Bhan: Absolutely. What further adds to the complexity of such a business operating model is you’re structured by categories because you are only playing in many consumer goods categories. We are actually playing pretty much across the value chain of food and beverages businesses, and now life sciences as well. And you can imagine multi-category set up like that. The need for localization in a category like food is high, because food is very personal to people at a consumer level, meaning our business from a profit and loss responsibility is decentralized by nature.
At Nestlé, we have always understood the importance of a very strong core digital for our enterprise architecture. And for many years back -- being very foresighted -- we created what is called our enterprise resource planning (ERP standardization program. We actually run our entire business, as far as the back office is concerned, in one ERP system. It's one of the biggest SAP implementations in the world. And, on top of that, that digital code allows us to create enterprise data domains, but we can never let those enterprise data domains cover all the needs of every single market.
Just imagine in China, for example, you have very specific media and consumer journeys. You will have them very different in Latin America, where you have different channels of trade. You have mom and pop shops in India or in some of the South American countries. While the world of grocery looks very different in a US or a European marketplace.
Protecting that localization -- whether it's about recipes, consumer journeys, or channel management -- means we have to be flexible. That's why our data and analytics model, from an operating model perspective, is hybrid at Nestlé. We try to centralize what we need to centralize, and we also allow the flexibility on the edge. We have teams working very close to our 40-plus markets across the world, which are each running a profit and loss responsibility. They help with that last-mile integration of data. But at the same time, they're not duplicating the efforts of what’s standard across the entire company. And they use the digital code that is coming from our enterprise data domains. Making it all work together is the art of working in a matrix organization.
Dana Gardner: I want to go back to a point you made, Vik, about trying to eliminate replication when it comes to data. There's the need to remove replication when it comes to applications, too. When you make a shift from a BI focus to an AI focus, is that about removing replication or is that a different type of journey? How would you characterize it?
Vikrant Bhan: You use the example of BI to AI, but there are countless examples. Obviously, in the past we had monolithic data warehousing technologies. Then we added a modern data cloud environment, and essentially the use of the monolithic data warehouse became less. I own the platforming strategy along with the data and AI services. We are always thinking about what is the sunsetting strategy or where we can be more efficient.
Nestlé, being quite a big company, means we have lots and lots of applications that we have within our application management platform, but we have a product management mindset on what are the things that we are conceptualizing, developing, further industrializing, and scaling. We are also looking at a very important aspect of sunsetting. But the reality is I do see, for example, BI starting to become more niche and not as expanded across the company today. We have a massive use of tools like Microsoft Power BI and other such BI tools.
But now with AI, we can talk to our data in a more AI-centric way. So you are right. I think we will start seeing that having an impact on the consumption of BI tools. There are many other such transitions. For example, there will also be an impact on robotic process automation (RPA), which was the earlier part of deterministicautomation. That is a part of my portfolio as well. And we'll see some of the AI agent capabilities to take those capabilities into semi-autonomous, if not autonomous, realm. And the list goes on. We are constantly looking at that roadmap and how to sunset what is not as relevant anymore.
Dana Gardner: While you've raised the subject of Agentic AI, how do you see that evolution manifesting when it comes to optimization of people, process, and technology? I suppose BI is very good at telling you what's already happened. Perhaps Agentic AI can tell you what to expect with high probability. And then also, how to get there and execute on it. Is that in your thinking?
Vikrant Bhan: Yes. When we began this journey, like every other company, two or three years back, we were obviously starting with a lot of what I would call AI assistance and still having the humans in the loop. That gave us a lot of learning on where we can augment within our processes and using people with capabilities that are essentially making their life easier. And you had to create some focused assistance for that. Not yet autonomous agents, but we started delving into agents and those agents were of different nature.
For example, Dana, you just talked about the talk to data kind of agent, or they might be an agent that is allowing you to write back data into your ERP system, or there's an agent that is following a step that potentially is multiple things that you have to do in different systems -- and it's orchestrating across all of that.
What we are now seeing is that because we have learned in that journey across these capabilities, which are like steps or tasks or activities within a process, we have understood what the building blocks are, which ones are working well, and which ones need a bit of work.
To be very honest, they're also learning the difference between what is advertised by the platform and product companies versus what's the reality when it hits our data, our ecosystem. And essentially, what we are starting to do now, in line of that re-imagination that I was talking at the very beginning of this call, we are starting to see how we can take something like an idea to launch a process within Nestlé.
Essentially, that means thinking about orchestrating across the different steps, different platforms, and different business workflows -- and then identifying parts of that so we can reimagine the flow. And, that is a veryinteresting exercise, Dana, because sometimes we also realize through it that maybe that step is not required anymore.
And we are going at it from an outcome-focused lens. We obviously are having outcomes aligned to a business strategy. We are focusing on the outcomes that we want to achieve and then drilling down from there into the processes that need to be changed. I wouldn't say we have cracked it, but I think clearly from a strategy perspective, that's where we want to go rather than do a plethora of lots and lots of assistance and agents. We want to focus our efforts on how we completely change the process itself.
Dana Gardner: It sounds as if you want to elevate project management to an agentic activity and then organize many different projects happening simultaneously in such a way that the people, the best knowledge and human resources, can then execute on those project elements.
But orchestrating it has always been very difficult. Is that a part of what you're seeing as for the potential for Agentic AI?
Vikrant Bhan: Yes, absolutely. We are now even looking at a portfolio of projects, which used to be use cases, to help us in specific functional capabilities and starting to think about how we cannot just prioritize based on value-amplification complexity, but also on how it’s contributing to making that an end-to-end process. I won't say it’s completely reimagined but optimized to the degree that it begins unlocking the maximum amount of value. Our big bets and our projects are now taking shape based on all of those factors.
Dana Gardner: Fascinating. Of course, bedrock to all of that and foundational is to use as much data resources as possible. We mentioned earlier that knowledge and unstructured data are a big part of that.
I'm wondering how Snowflake is helping you to bolster that foundation in order to enable some of these higher-order agent values over time?
Vikrant Bhan: When I mentioned our use of a modern data cloud environment, Snowflake is that for us at Nestlé. And when I was talking about the 15 enterprise data domains that we have been working on, that Snowflake data cloud is what we are using.
Snowflake has certainly helped us achieve the ability to start piecing together all of this data, which used to be in lots and lots of systems of record and systems of engagement. We can now bring them together as data products for consumption in multiple use cases.
Those use case patterns have obviously been very successful, and have been scaling machine learning use cases, scaling BI use cases, and scaling some of the capabilities you would typically associate with AI assistance and agents.
We are still trying to do some groundwork when it comes to the unstructured data field because we understand the importance of it. And obviously when we are working on a particular use case, like focusing on source-to-pay, we are focusing then on the contracts within our company. Or if we are looking at our factory operators, then targeting all the standard operating procedures and various documentation that is related to the different tasks.
We are going at it use case by use case, but there is still a lot of work to do to just clean up the data using clean rooms and things like that to bring it all together. As you can imagine in a very big organization like Nestlé, with our decentralization, we did not have all of these components of unstructured data necessarily well-organized or well-defined or have metadata around it.
Dana Gardner: You also mentioned earlier, Vik, that talking to the data is important and that people can begin to interact with data as a conversation. Natural language suits the human rather than the machine, if you will. Is Snowflake an important element of making that possible?
Vikrant Bhan: Absolutely. We have obviously been using Snowflake a lot for the patterns that I mentioned, but when it comes to the agent side, one pattern that we are seeing is talk to data. Obviously, we don't want to have everybody going and creating reports in the company. But they spend a lot of time on it, and we want to focus on getting the insights, and certainly we are using some of the capabilities that are coming from Snowflake. For example, Cortex AI, which is a product that obviously Snowflake is investing quite a lot in.
Working closely with our key account and architecture groups that are provided by Snowflake, we are starting to use those capabilities that are coming on the ground running, and then we are starting to use them in those features like talk to data.
Dana Gardner: And how about that other concept we touched on, being able to use analytics and AI to bettermanage and process and optimize automation. Is Snowflake a factor in that, being able to point the technology at the use of and cultivation of processes in terms of analytics?
Vikrant Bhan: Yes. There is a lot of metadata that we are generating, and we talk about the active metadata. We have many different measures for how we see the success of our data products. And I think just like you would want to see a successful business or a successful function or a successful anything, you need to have metrics to measure that. And I think we also need the measures for a successful data product.
We are using a lot of active metadata from the systems to see what the consumption patterns are, how people are querying things for financial operations, for seeing the type of queries people are asking and hence how can we tune them. So, there are lots and lots of capabilities that we can use from the data about the data to make our data products even faster.
We actually have a maturity score within the company that we are measuring across five dimensions. And obviously the data about the data is helping us come to that maturity score as well.
Dana Gardner: I should think in an organization that's hybrid, as you described, with that with a high degree of centralization and the need to go out to the edge -- in various markets, different geographies and jurisdictions and legal environments -- that openness and standards whenever appropriate are important.
Is there anything about the way Snowflake manages openness when it comes to data, repositories, and cloud structures that allows an organization like yours to better accomplish the goal of a hybrid efficiency?
Vikrant Bhan: Obviously we don't replicate the data outside of Snowflake into fringe environments. So, we have set architectural patterns from our enterprise architecture, as well as our centralized platform teams. And one of the things that we are starting to do within our edge infrastructure is -- rather than people exporting data or potentially sending data to some other environment -- we enhance the data domains.
That also comes back to capability of building and training the teams at the edge, which is not easy because you have to deal with 40 different entities. But some of our operators in these markets have very, very amazing skills. I would say even more mature than maybe some of the people in the central team.
We are always using them as lighthouse markets to try these things out and then see how we can learn from that and potentially spread that across different markets that we have. And essentially everybody contributes to the same ecosystem. But keeping in mind that there are some guardrails, we call it within Nestlé, and I'm sure it's called in the industry, something similar, freedom in the box.
The box has to be very well defined, and the box has to be defined through some standards and some guardrails, some guidelines, some automation scripts and various other capabilities. But at the end of the day, it's more a governance topic and not just a platform topic.
Dana Gardner: In our discussion, it's become evident to me that you're very advanced in how you're using analytics and are not just creating a vision around agentic AI, but a true path that you're walking. That gives you, I think, insight into the relationship between people and their culture and what agentic and AI can do.
So, can you share with us what you may have learned about what's the right mix or the approach to enabling the best of what people and AI can do? What is the cultural path? Because this is quite unprecedented, entirely new in human evolution where the opportunity to combine the best of digital and the best of organic come together.
Vikrant Bhan: A lot of the focus when you talk about AI goes on the technology and the delivery models, and on how to deliver the architecture of AI. But, honestly, what is the most important thing? And most people would relate to this.
You can have the best agent, AI tool, machine learning model, or application to solve a particular business problem. But unless and until you have the right people who are going to use it with you on the journey, it's never going to happen. So, there are various steps that we have embarked on and that is our functional capabilities.
We are focused on using functional excellence within Nestlé. So, our products, which we build, are owned by our business counterparts who take a proxy across all of our markets. And there, one of the big roles of theirs, is to make sure they are doing a lot of learning capability, building programs when it comes to change management and adoption.
And our business entities, when they are deploying these capabilities have a very clear focus on change management and adoption because we can see that there are tools that we have created that have been extremely successful in one market, but in the other market, it doesn't work. It can't be about the tool because the tool obviously is working for many markets. It really comes down to that change management and adoption capability. So that's very important.
But the other thing, which I think is starting to become quite interesting, is in AI. I think the complexity that people had to deal with the different workflows, as an example, or learning new tools -- because imagine what used to happen. You had systems of record, systems of engagement, systems of data, multiple reports, multiple applications that people had to be taught on. And at the end there is this one guy in the market or one end user who has to learn all of this.
And what is now happening, and this is what they see in their consumer day-to-day life, too, is when they are operating some of the Gen AI tools, as a consumer, we are trying to bring that experience to our enterprise as well. Maybe through chat-based interfaces, we are creating the experience layer for all of these products, so they don't have to essentially learn a workflow in a tool X or a tool Y, or a simulation capability in a particular report or an application.
But at the end of the day, the experience layer is a common one. The complexity of the agents and complexity of the assistance and orchestration and data products is hidden from them. That's the other advantage that we are now seeing with this technology, which is going to make it hopefully easier for that change management and adoption as well.
Dana Gardner: Wow. It sounds as if we have a new equation for managing and dealing with complexity, a combination we haven't quite figured. But, certainly bringing the best of what Agentic AI and what human creativity can do will allow complexity to continue -- but not become a barrier, but provide a path to even better complexity, if there's such a thing.
We always think of complexity as a bad thing because it's so difficult, but perhaps this combination, as you're describing it, will allow complexity to become a good thing.
Vikrant Bhan: Yes, and maybe eventually it'll also help us imagine different operating models, which today is not the case because you're limited by the way things work in a process, right? If you start re-imagining a lot of the steps in the process, then you might start thinking about what roles do you need in a particular process, and what are the humans in the loop in which process? And that actually will, maybe, remove some of the complexity from that equation and help people focus on decision intelligence and where we need their minds to operate and work.
We aren't there, Dana, just to be clear, but we obviously want to get there. Just to be clear.
Dana Gardner: Yes, that's probably another good topic for another day. But, yes, I appreciate what we've been able to get into and understand the level of accomplishment that you've made at Nestlé. I'm sure that's going to help other people think about their direction and approach as well.
So, I thank you so much for joining our latest Data Cloud Podcast, Vikrant Bhan, Group Head of Analytics, Data, and Integration at Nestlé. We really do appreciate your sharing your thoughts, experience, and expertise with us all.
Vikrant Bhan: Thank you, Dana. Very happy to do so.
Comments ( 0 )