In CPG, AI Has to Be Infrastructure, Not a Project

AI AdoptionEnterprise AI
Written By Christina Ellwood

Episode Summary

Nitin Gupta is global product manager of AI at Mondelēz International, and the work he describes is not greenfield: integrating AI into legacy systems that still run fine, standardizing the data underneath. Most use cases go through a governance or AI council first, which he frames as general enterprise practice rather than as his employer’s. One wow moment he names is computer vision at the shelf, where LLM-based models turn a photograph into actionable insight without waiting for the data to be processed. A proof of concept is easy; a result that holds in one may not hold at scale, so production means iterations, testing and a human in the loop. Adoption is a different problem: resistance to change, leadership turnover that loses the vision, and outputs a business will not trust unless they are explainable. His structural point is that AI belongs in the DNA of every function, the way mainframes quietly run financial services, rather than sitting at the periphery.

Key takeaways

  • Vision models changed what a shelf photograph is worth. Photographing shelves is not new in CPG, but LLM-based vision models turn a photo or a scanned video into actionable insight immediately, rather than waiting for the data to be processed and returned

  • A POC result is not a production result. Something that works in a proof of concept and gives good results may not give those results at scale, so the route through is iterations, multiple rounds of testing and a human in the loop

  • The AI council is where use cases get vetted. Thousands are technically possible, so most go through the governance or council team to be checked for fit, for responsible AI and for exposure of confidential data, and a successful POC is still a separate leadership decision to go to production

  • Legacy systems are the friction, not the failure. They work, and they keep working, so the problem is integrating them with AI-driven approaches and replacing fragmented architectures with a unified one

  • Explainability is what buys trust. Data scientists build the models to be explainable rather than black boxes, business users vet the results before anyone relies on them, and impact is then measured with a control group against a test group

  • Adoption stalls on people, and on regulation. Resistance to change, gaps in leadership buy-in, a change of leader partway through after which the same vision is not carried forward, regulatory concerns, and the fear that AI will replace jobs sitting underneath all of it

  • The visible impact reached the sales rep. Reps sell better because they are able to have all the information together with them when they are actually negotiating or working in a store

  • The value is spread, and the rep is where it shows. Asked where AI creates the most immediate value in consumer packaged goods, he declines to pinpoint one thing and says it cuts across everywhere, naming ground sales reps as where the true impact is visible and listing waste reduction, forecasting, planned visits and a supply chain undisrupted by weather and events

  • AI belongs in the DNA, not at the periphery. Mainframes are the analogy: nobody puts them in headlines, and they are the invisible engine under financial services and aerospace. AI should sit under every function the same way rather than being a project

  • Data foundations decide everything downstream. If the foundations of data are not correct it is junk in and junk out, which is why he puts data first, ahead of objectives, platforms and vendors, for anyone setting up an AI practice

About Guest1

Nitin Gupta is global product manager of AI at Mondelēz International, and has led AI work across retail, telecom, industrial, financial services, manufacturing and now consumer packaged goods. He describes data as the thread through all of it, and the mandate as commercial: delivering revenue and commercial insight through AI-driven products, and not just producing proofs of concept. That has meant predictive maintenance at a manufacturing conglomerate, demand forecasting, and in every case the same second half of the job, which is scaling the team and the solution to run at enterprise level and getting the C-suite behind it. On strategy he is blunt about dependency: too much reliance on a vendor or a single platform product is a risk, and a company setting up an AI practice is better served by someone who has taken projects to scale before.

 

In this episode

00:42 Welcome and guest introduction
01:09 Financial services to retail to consumer packaged goods
01:52 Data as the thread through every domain
02:30 Predictive maintenance, demand forecasting, and influencing the C-suite
03:49 Legacy systems as the friction point
04:34 Silos, and a unified way of approaching AI at scale
05:17 Applying AI where the work could not be done at all before
05:28 Traditional AI and Gen AI, separated
05:50 The wow moment in CPG and retail
06:12 LLM-based vision models, not the foundational ones
07:00 What the AI council decides
07:11 Guardrails, and thousands of possible use cases
08:15 Moving from pilot to production to adoption at scale
08:40 The toughest thing
08:44 Why POCs are easy and production is not
09:06 Results that hold in a POC and not at scale
09:46 Adoption as an altogether different ballgame
10:08 Cross-functional teams and change champions
10:40 Resistance to change, and gaps in leadership buy-in
10:58 Trust issues, and why black box outputs do not survive
11:45 Training, and reps who sell better with the information in hand
12:14 Where AI creates the most immediate value in CPG
12:26 Cutting across everywhere, and reducing waste
13:33 What an AI first mindset means in practice
14:04 The end-to-end chain, and the mainframes that left the headlines
14:51 AI as part of the DNA, across every function
15:05 Overcoming trust gaps
15:16 AI governance and data governance, side by side
16:18 Guidance for executives early in their strategy
16:42 Data foundations, and junk in junk out
18:00 Resources for listeners
18:50 Do not shy away from failing fast
19:25 Have AI in your DNA
20:33 Learn, unlearn, relearn
20:48 Wrap-up

In Nitin’s words

“Something which is working in a POC and giving you good results may not work and give you results at scale.”

— Nitin Gupta   (09:06)

“You cannot have black box outputs. The outputs needs to be explainable so that there is business buy-in”

— Nitin Gupta   (10:58)

“It should be part of the DNA where for enterprise, it should sit and seamlessly work across every function in the organization, delivering efficiencies and helping in being competitive in the industry.”

— Nitin Gupta   (14:51)

“The foundations of data have to be absolutely correct, otherwise it’s all going to be junk in, junk out.”

— Nitin Gupta   (16:42)

“Don’t shy away from failing fast.”

— Nitin Gupta   (18:50)

“I don’t assume things. That’s one simple thing. I don’t believe in assuming anything.”

— Nitin Gupta   (20:33)

 

Resources

Nitin Gupta and Mondelēz International

Ideas and frameworks discussed

  • Computer vision at the shelf: Shelf photography is long established in CPG and retail. What changed is that LLM-based vision models read a photograph, or a scanned video, straight into actionable insight rather than sending data off to be processed first

  • AI first mindset: AI as part of the DNA of every function in the end-to-end chain, rather than a project one or two teams are running at the periphery of the business

  • The mainframe analogy: Mainframes stopped appearing in headlines and never stopped running core financial operations. His argument is that AI should end up invisible and load bearing in the same way

  • The AI council: The governance body most use cases pass through before a POC is approved, checking fit, responsible AI and exposure of confidential data. He describes it as how any enterprise works rather than as his employer specifically. A successful POC is still a separate leadership decision to go to production

  • POC to production: His account of why the gap is the hard part: results that hold in a proof of concept may not hold at scale, so production means iterations, multiple rounds of testing, a human in the loop, and the enterprise architecture guidelines applied

  • Explainability as the trust mechanism: Models built to be explainable rather than black boxes, results vetted by business users, and impact measured against a control group and a test group

  • Data foundations first: If the foundations of data are not correct it is junk in and junk out, which is why he puts data ahead of objectives, platforms and vendors when a company is setting up an AI practice

  • Fail fast, then learn as you go: Creating proofs of concept has become easy enough to test a hypothesis and move on quickly, and the more you fail the faster you learn

  • Learn, unlearn, relearn: His answer to what his personal superpower is, and it starts from not assuming anything, on the grounds that anything can turn any way

Named on air

  • Perplexity: perplexity.ai. One of the two systems he recommends keeping close for brainstorming and trying things out

  • ChatGPT: chatgpt.com. The other, named alongside Perplexity in the same recommendation

  • AI Realized executive roundtables: airealizedsummit.com. Christina raises the roundtable discussion about the definition of scale, and the theme that adoption follows use cases integrated into existing workflows

Related AI Realized episodes and events

 

Frequently Asked Questions

  • AI proofs of concept fail to reach production because a result that holds in a pilot may not hold once it has to run at scale. Nitin Gupta of Mondelēz International puts demand forecasting as the example: it performs in the POC and then does not replicate once it has to run for real. Getting through means iterations, multiple rounds of testing and a human in the loop evaluating the results, plus the enterprise architecture, security and legal guidelines followed so the system can handle things at scale.

  • Computer vision models improve retail shelf execution by turning a shelf photograph into actionable insight without waiting for the data to be processed first. Photographing store shelves to see how consumers buy is long established in consumer packaged goods, and what changed is that LLM-based vision models, rather than the earlier foundational models, read a photograph or a scanned video directly into actionable insight. Nitin Gupta of Mondelēz International compares the pattern to autonomous vehicles combining vision, audio and sensor data for real-time insight, and credits the increase in available compute for making results possible that were difficult to get before.

  • An enterprise AI council vets proposed use cases and puts guardrails around them before leadership decides which ones proceed, on a field where thousands of use cases are technically possible. Nitin Gupta of Mondelēz International describes the test as what fits the ecosystem, what maximizes benefit, and what avoids exposing confidential data to any public system, with responsible AI underneath all three. Most use cases are presented to the governance or council team, leadership decides whether to move forward with a POC, and a successful POC is then a second and separate decision about production.

  • An AI first mindset means AI sits in the DNA of every function rather than being a project one or two teams are running. Nitin Gupta of Mondelēz International draws the analogy to mainframes: nobody writes headlines about them, and they are the invisible engine under core financial operations and aerospace. The test he applies is whether AI is driving the business from its core or sitting at the periphery, and an AI first company is one where it works seamlessly across the end-to-end chain and shows up as efficiency and competitiveness.

  • You build trust in AI model outputs by making the models explainable and having the business verify the results before anyone relies on them. Nitin Gupta of Mondelēz International describes two teams, AI governance and data governance, sitting alongside each other, has data scientists build for explainability rather than black box outputs, and has business users check that results make sense. Impact is then measured with a control group against a test group, so the case for trusting the system rests on a comparison rather than on an assurance.

  • A company should start its AI strategy with the data foundations, because if those are not correct everything downstream is junk in and junk out. Nitin Gupta of Mondelēz International then adds three things: embed AI projects in existing business processes instead of running them in silos, set a clear objective rather than building Gen AI for its own sake, and avoid over-dependence on any one vendor or platform product. His fourth is to have someone who has taken AI projects to scale before, on the grounds that the cost of failure is high enough to be worth the hire.

  • AI adoption inside a large organization is blocked by resistance to change, gaps in leadership buy-in, and outputs nobody can explain. Nitin Gupta of Mondelēz International adds a specific and underrated one: a change of leader partway through, after which the same vision is not carried forward. Underneath those sits the fear that AI will replace jobs, which he answers as augmentation rather than replacement, and the practical response across all of it is cross-functional teams, change champions and training, with the evidence being sales reps who negotiate better because the information is in front of them.

 

  • [00:42] Christina Ellwood: Welcome to AI Realized, the podcast for enterprise executives leading AI deployments. From tackling security, data, and organizational challenges to navigating organizational transformation, AI deployment offers unique opportunities to redesign organizations from the inside out. I'm Christina Ellwood, your host for today's episode, and today we're talking with Nitin Gupta, the global product manager of AI from Mondelēz International. Nitin has led AI initiatives across multiple industries, from financial services to retail to consumer packaged goods. Today, he's helping a Fortune 200 company navigate the reality of deploying AI responsibly, moving POCs into production, and setting up governance councils, and building an AI-first mindset. Nitin, welcome to the show.

    [01:31] Nitin Gupta: Thanks, Christina

    [01:33] Christina Ellwood: Prior to your AI role at Mondelēz International, you've also worked at telecom, financial services, retail, and now consumer packaged goods. What unifies your journey, and how do you approach AI strategy?

    [01:48] Nitin Gupta: Sure. Thanks. Thanks, Christina, for the overview about my background. So yes, I have worked in different domains, especially retail, telecom, industrial, financial services, manufacturing, and now in CPG. So what connects, I would say, is data. The core of my mandate has been how do you deliver more revenue, more commercial i- insights for the company using AI-driven products. And what has fundamentally connected every step in my journey is it was not just create creation of proof of concept, it was how do you scale AI teams, how do you take it to run at enterprise level and scale the solutions, right? From predictive maintenance work in, at a manufacturing conglomerate to driving more demand forecast and forecasting for the public goods that are there. And leading to a tangible business impact, influencing C-suite, building scalable solutions, I would say were the core which were connected across all my background so far.

    [02:50] Christina Ellwood: Scaling was a big topic at our most recent round- executive roundtable. In fact, there was quite a brisk discussion about the definition of scale, and one of the themes there was that in order to scale and have actual adoption, you need to have a use case that is integrated into existing workflows rather than establishing an entirely new area for, for the application, even though AI holds so much promise to allow us to do things that before now we haven't been able to do at all. Has that been your experience, that tackling an existing workflow and making it more efficient or more powerful, like in the case of your work in getting snacks on the shelf, if you will, for Mondelēz. Is it your experience that taking established workflows and applying it there is a more successful way to achieve adoption?

    [03:49] Nitin Gupta: I would say every company or every enterprise would be having a lot of legacy systems, and they are complex systems, right? So that is one major friction point I would say that a lot of companies do have that, okay, how do you ensure that all these systems, things have been sitting and, of course, they are working fine even till date. So it's not that legacy systems are not working. But how do you interact, or how do you ensure that the integrations of these legacy systems happen with the new age AI-driven, you know, ways? And having unified architectures or fixing these fragmented architectures so that the deployments happen seamlessly, they usually slow down how solution integrations happen. And I think that is probably one big problem which the entire industry is facing, that, okay, how do you ensure that these silos of legacy systems, they go away, and then you have a unified way of approaching AI at scale. And we are also working to, to solve some of these things like standardizing our data, having the right foundations for having governance in place, and of course, everything should be backed up by responsible AI. So ethical and legal con- legal angles are also taken into consideration, especially when you are integrating with the legacy systems charter.

    [05:08] Christina Ellwood: So it sounds like you, you... in this integration with legacy systems, that's really an ex- using AI as a new approach to existing work.

    [05:17] Nitin Gupta: Mm-hmm.

    [05:17] Christina Ellwood: Have you had a situation where you were applying AI in an area that previously couldn't be done without the use of AI, and can you contrast that experience?

    [05:28] Nitin Gupta: For me, AI has been existing for quite some time. It's not a new thing, okay? It's... Now we are separating it with that, okay, one is a traditional AI, which is foundational AI, like machine learning, forecasting, NLP, and the new, the Gen AI world, which is there, which is more on, on a generative AI component. So I would say our use cases were already there where we were seeing value. With Gen AI coming into picture, I would say it's even more, it has expedited our delivery. So something which was... I'll take an example here, and this is probably one of the wow moments as well, especially in CPG industry or in retail industry, I would say that, okay, earlier we used to take photographs of shelves, right? Of, of how consumers are utilizing or consuming or buying products at a store. Now with vision models, right? Now I'm not talking about the foundational AI models, I'm talking about the new LLM-based vision models which are there. How do you... You take a photograph, right, and you even scan a video. It's converts it into actionable insights so quickly. That's a very unique use case which I've seen that, okay, you don't have to wait for the data to, to process and then give out information. Self autonomous driving cars, they also use similar complex multi-model systems where vision, audio, and sensor data are combining, and you get insights in real time. So I think with the advent of more compute power scaling, we are able to achieve better results which were earlier a little bit difficult, you know, to get.

    [07:00] Christina Ellwood: What role does the AI council at Mondelēz play in how the company makes decisions about using AI and how it affects your work?

    [07:11] Nitin Gupta: So, and this is for in general, like any enterprise company, I would say, not just speaking about Mondelēz. AI councils usually help in ensuring that right guardrails are put in place and because there can be like thousands of AI use cases which can be deployed. But everything has to be backed up by responsible AI, how, what things fit into our ecosystem, which would maximize our benefit and minimize any kind of impact or exposure of our confidential data onto any kind of public system. So I would say majority of the use cases, they go through the governance team or the AI council team, and then they are vetted over there. Presentations do happen, and then the leadership takes a decision whether they should be moving forward with or doing a POC. And then of course, if the POC is successful, then should that go into production? Because every POC may not lead to same benefits when they actually go into production. So hence decisioning, leadership decisioning, leadership buy-in is very important, and that's what happens as part of the council.

    [08:15] Christina Ellwood: Certainly getting that buy-in is critical to being able to go to scale with a... or even just to go to deployment from a pilot. When moving from pilot to production to adoption at scale, what has worked for you in that flow, in moving from project to pilot to production to scale?

    [08:40] Nitin Gupta: Okay. So I would say that's the toughest thing. Creating P- the POCs is easy. The reality is things when they actually have to move in production and work at scale, right? You need to ensure that things are, uh, happening the right way, like the models don't hallucinate. You get the results or you're able to replicate the results. Even forecasting, demand forecasting, if I take an example. Something which is working in a POC and giving you good results may not work and give you results at scale. So doing the-- And it-- So it has to be backed up by a lot of iterations, right? Multiple rounds of testing, human in the loop, right? To evaluate the results, to ensure that what is expected of it is actually comes out even from a production perspective. So that's-- those are-- And then of course, security and legal aspects I already talked about. So basically, moving from a POC to production, we have to ensure that all the guidelines which are set up at an enterprise level from an architecture perspective are followed so that the system is able to work and handle things at scale. Now, when I talk about adoption, that's an altogether different ballgame, right? Where we need to-- Again, the change management is very important, especially having the right set of people in place, having the r-right leadership buy-in so that the adoption of the product happen. There are lots of organization challenges which have to be over-overcome, especially when you're talking about adoption of an AI product. So working with cross-functional teams, be it in business, be it in tech, and having change champions is very important from an adoption perspective. And it might take time, right? Because of different issues and areas, so hence it's necessary to take care of all that.

    [10:24] Christina Ellwood: On the adoption side of things, working cross-functionally in your organization, what is an example of a adoption challenge that you faced and then solved with a cross-functional team?

    [10:40] Nitin Gupta: Some of it, like I face every single day, I would say, okay, especially working in from a product side of things. See, number one is it can be resistance to change. There can be gaps in leadership buy-in, maybe a change in leadership also when people move, when leaders move from one role to another role, right? So same vision is not carried forward. So it's important that the vision is carried forward. Then you also have regulatory concerns. Change management also moves not just from a product perspective. There are-- There can be trust issues also, right? You cannot have black box outputs. The outputs needs to be explainable so that there is business buy-in, there is like people who are stakeholders responsible for taking it forward with and helping in change enablement. They are satisfied with the results. So those are some of the blockers which I feel are there and hurdles which we face and then we overcome. And then there is some, some fear, right? That okay, people say that, "Okay, AI will replace jobs." It's not that, right? So it's helping in augmenting or helping in improving the way things are happening. So we have to do the right set of trainings also. So training also plays a very important role that, okay, your productivity is going to increase or improve when you have these kind of solutions back in your day-to-day work. And we have seen substantial improvement. People are-- Reps are able to sell better because they are able to have all the information in, in-- together available with them when they're actually negotiating or w- working in a store. That change in mindset is helping them understand through, through the right training and the right tools for them.

    [12:14] Christina Ellwood: In the consumer product goods area, where do you see AI as creating the most immediate value today? Is it in the supply chain, in the consumer insights, or in the product innovation, or in some other area?

    [12:26] Nitin Gupta: I would say it's touching all across. It's not-- Like, I cannot pinpoint one single thing. It is cutting across el- uh, everywhere. The most important impact, like I said, when our ground sales reps are able to use things, uh, and we are able to assist them, that's where the true impact is visible. And we are seeing, like, lots of impact in terms of that, okay, how do you reduce waste? How do you have the right forecasting, right? How do you ensure that visits are planned and the supply chain is not disrupted due to weather, due to different events which happen? So it's cutting across, I would say, and everywhere there are, there are touchpoints where we are making use of AI.

    [13:05] Christina Ellwood: So it sounds like at the end of the day, you're able to ensure not only is your product on the shelf and it's fresh, but that the, the store and the, and your salesperson are all on the same page about what is needed w- and when. Is that a fair summary?

    [13:21] Nitin Gupta: Absolutely. And what actions they need to take. So the, some of my products are based on that, where providing the right insights to the sales reps is very important, and that has to be data backed. It really helps them and...

    [13:33] Christina Ellwood: Yeah. So you've called for an AI first mindset. What does that mean in practice for a global enterprise?

    [13:40] Nitin Gupta: By AI first mindset, what I mean is we need to be thinking of AI not as a side project or not as just as one project. In my opinion, it should sit like an, like a DNA, right, in, in the work that we do. And it's not, it should not happen that, okay, one team or two teams are working in, in creating a product. By AI first mindset, it should sit in every aspect of the entire end-to-end chain when you're working in a company. It should be driving business and right from its core, and it should not just sit at a periphery. Now to that example, I'll-- I can correlate it to maybe like mainframes. You don't hear about mainframes in the headlines anymore, right? But mainframes is something which is the core, right, which is helping in all our core financial institutions' operations. That's the invisible engine which is sitting beneath this industry, be it in, in financial services, be it in, in, in, in our aerospace industry. Likewise, AI also should not be seen that, okay, it's a project and then, okay, we are implementing AI projects. It should be part of the DNA where for enterprise, it should sit and seamlessly work across every function in the organization, delivering efficiencies and helping in being competitive in the industry.

    [15:05] Christina Ellwood: Well, trust is such an important part of that equation. Sure. What are some of the things that you have done that have helped to overcome trust gaps?

    [15:16] Nitin Gupta: From a trust perspective, like I said, I mean, having responsible AI framework and having the governance in place is very important. We have an AI governance, we have a data governance. Both teams are there, okay, who ensure that the trust is there. Our data scientists ensure that the models are explainable. They don't sit like a black box, okay? So any results that are coming out or given out, they have to be, like, explainable. And then, of course, when the results are vetted by a business so that everything which is coming out from the models, they just don't remain like a black box. It has to be, uh, the URT has to be pi-passed by the business users that, yes, these results are making sense. And that's how we are improving our trust. And then, of course, once we see the results, we do an A/B testing kind of a thing, right, where we compare the results and we have a control group and a test group, and we see how much impact we are able to, like, bring the two. And that clearly improves the trust in the systems.

    [16:18] Christina Ellwood: So for, for executives that are early in their journey, what is your guidance for establishing their strategy? You're a strategist by nature. What do you recommend to executives who are early in their strategy journey?

    [16:31] Nitin Gupta: So people who are early in their journey, right, or also, or, or it may be the founders or established companies also who are setting up their AI practice. I would say the first and foremost thing is ensuring that f- the f- data foundations are built okay. The foundations of data have to be absolutely correct, otherwise it's all going to be junk in, junk out. AI projects should not work in silos, like I said. It should not be working as a periphery, it should be embedded into your existing business processes. And having clear objectives, right? Don't just have a Gen AI project for the sake of having it. They are powerful, and how do you let or let it, like, integrate with other systems, very, very important. And have an AI strategist, I would say. You should not have too much over-dependence on vendors. You should not to- have too much dependence on, on a platform product, right? If someone is knowledgeable, someone is from that background and has worked on projects on scale, I would say the person can fill in and help in having the right AI strategy. So if founders can take in these kind of aspects, or it would like, it's going to help avoid any kind of financial impact because the cost of failure for AI projects is significantly high as well. So if 5% projects are also successful, the revenue impact is going to be tremendous. So having the right strategy in place for that really matters.

    [18:00] Christina Ellwood: What resources do you recommend for listeners who want to learn more about you and about how to be successful in their AI deployments?

    [18:09] Nitin Gupta: For me, I can be found on LinkedIn, I would say. So LinkedIn is my most important and most active network where I am. Resources, I think there are tons of resources. Of course, on LinkedIn you can follow a lot of people, a lot of top voices who are there, who are writing about AI, and you can definitely learn through that. And I would say have Perplexity, ChatGPT. These can be your very close systems to brainstorm and see and try out different aspects. Those can be the closest ones. You want more deeper, then there are lots and lots of folks who are providing specialized talks on these areas. You know, your, your podcast is one of those, right? So they can definitely leverage these things. One more point which I can state, say to my listeners over here is Don't shy away from failing fast. I would say embrace that you need to fail fast and you learn from your approach, and that's how you know you can move forward faster. These days, creating proof of concepts is, has become very easy. So just test your hypothesis and move forward quickly. And then of course, it's learn as you go, right? So the more you fail, the more you're able to learn faster.

    [19:19] Christina Ellwood: Is there one thing you would like people to take away from the show today?

    [19:25] Nitin Gupta: One takeaway, have AI in your DNA. That's what I would say. Everything, be it your personal life, be it your professional life, even your family, right? And my six-year-old is able to make use of AI. She ask it so many questions that, okay, some of the questions even I am learning that, okay, the answers are much different than what I had thought of. So I would say embed AI into your life in every, in, in every possible aspect. You will gain from it.

    [19:53] Christina Ellwood: Okay. Very good. So in the AI revolution, what's your defining edge in guiding AI leaders?

    [19:59] Nitin Gupta: There can be two sets of things, right? People who are moving fast, so who are early adopters. So if you are, if you're one of those, then you can have an upper edge through that. And then there are companies, there has been a study on that, okay, so that companies who want to wait and watch. I would say there's no right and wrong answer to it. Uh, you can benefit on both aspects of it. It's how you improvise what you have learned from either of the two and have the competitive advantage. That's what, you know, matters ultimately.

    [20:29] Christina Ellwood: What's your personal superpower when leading AI initiatives?

    [20:33] Nitin Gupta: I don't assume things. That's one simple thing. I don't believe in assuming anything. And anything can turn any way, and you need to learn, unlearn, and relearn. So these are like the foundations that I now follow.

    [20:48] Christina Ellwood: I see. Oh, that's a great insight. Thank you, Nitin Gupta, the global product manager of AI for Mondelēz International, for joining me today on AI Realized.

    [20:59] Nitin Gupta: Thank you so much, Christina. It was pleasure talking to you.

 
Listen to other episodes
 
 
 
Join AI Realized
 
 
 
Subscribe to the newsletter
 
 
 
Christina Ellwood
Previous

Never Surrender Agency to the Agent
Next

Shadow AI Is a Permission Problem, Not a Tool Problem