The next episode of the Data Cloud Podcast features 2026 vertical industry predictions with experts from Snowflake, Rosemary DeAragon, Global Head of Retail and Travel; Rinesh Patel, Global Head of Financial Services, and Tim Long, Global Head of Manufacturing.
Our guests explore how AI will transform retail, financial, services, and global manufacturing in 2026. Together they break down the forces reshaping consumer behavior, enterprise operations, and competitive dynamics across these sectors. Please enjoy here this discussion between these three vertical industry experts at Snowflake and your host, Dana Gardner, Principal Analyst at Interarbor Solutions.
[Listen to the discussion or watch it.]
Dana Gardner: Welcome to the Data Cloud Podcast everyone. We're delighted to have some vertical industry experts here with us to make some insightful AI predictions for 2026. Welcome to you all.
Rosemary, looking ahead to 2026, and how specific industries will be impacted by AI, how will agents specifically begin to automate shopping for consumers, especially retailers of all stripes as they compete for attention across all sorts of domains and across all sorts of media?
Rosemary DeAragon: We see it happening already today. The way that the internet is being used is fundamentally changing. In the past, a lot of people would search using a traditional query. Typically, that would be a few words, not as verbose as it is today, and people would search through the search index: the first search result page, the second page, the third page.
Now with AI summarization, a lot of people aren't even visiting those websites anymore. They're getting the answer right at the top. They're getting a recommended product right at the top, and it's already changing the game. So AI has already changed how products are being discovered and how people are shopping for different products.
To your question of how agents are going to change shopping, I think hugely, especially in the world of consumable products that you would typically use an auto replenishment feature for. That's where you'll see a little bit more trust from a B2C standpoint in shopping agents, which we haven't yet seen. Here at the end of 2025, we haven't yet seen large-scale consumers trusting AI to transact and make purchases on their behalf.
But I think in 2026, you can expect to see that more and more, especially for “boring” items like toilet paper or for products and brands that you know and love, and you know that you need to replenish them in a certain cadence. A lot of those are ripe for agents to transact. There are other types of products, luxury products, et cetera, that are less prone to agentic automation and shopping. But certainly, we'll see a rise in that from a consumable standpoint.
A lot of companies, and not just end users, but in the B2B space, have purchasing “agents” already, sometimes entire organizations and teams, set up to help people with procurement. Is that then something that can be largely automated?
Rosemary DeAragon: Totally. I think the future within a retailer is a multi-agentic future. You know, Snowflake's MCP server and the use of Model Context Protocol (MCP) servers is going to increase dramatically.
In the future, you're going to see enterprises having agents, for example, multiple agents within supply chain -- multiple agents within merchandising -- buying, procuring via multiple agents, even within customer 360 advertising and all of that. We are going to be in a multi-agentic future.
Retailers that have a good grasp of their data, and the ones that are doing deep semantic modeling on their data -- where AI can actually understand what that data means -- that is going to be super important for them to ensure that they're not left behind in this world of AI.
Dana Gardner: And some of the more sophisticated individual shoppers, as well as individual businesses on the buy side, can soon use all sorts of ways to improve pricing and delivery times and create incentives for certain events. And so, it can become rather complex.
But in doing that, the savings and the value around the new productivity, and the just-in-time nature of consumption, improves a lot. Maybe, therefore, we're going to democratize that sophistication, to bring from what the bleeding edge does down to the base.
Rosemary DeAragon: I think so. I also think that there are traditional machine learning methods that have been used for several decades in the world of dynamic pricing, smart pricing.
You can use linear aggression models for things like that. Obviously, there are also cases where traditional machine learning models are being replaced with generative AI. So that would be like you were talking about – just-in-time merchandising. We're seeing that at a fast pace, especially in fashion and apparel where you have these fast fashion trends that are coming about in competition with the sustainability trend as well.
But you have that fast fashion trend and you have generative AI taking over a lot of the more traditional machine learning use cases where you can then understand human sentiment from social media data hoses. You can understand what the trends are as they're occurring and actually manufacture things just-in-time.
It’s actually a big threat to brands because you're able to manufacture so quickly and produce these fast fashion goods so quickly, especially with the rise of social commerce, TikTok shopping, and all of these e-commerce methods of delivery.
I think traditional brands need to understand their product data and tie the product data to human sentiment, as opposed to the traditional product catalog for a red sweater.
What does a red sweater evoke? Is it fall? What kind of red is it? Is it Valentine's red or is it more like a fall red? And a lot of those types of additional products, data enriched product data attributes, traditionally wereoverlooked and maybe you didn't have the manual capacity to be able to add that onto your product catalog.
Now with generative AI features, you're able to augment your product catalog and with semantic modeling be able to understand exactly what the products’ item data and how it ties to human sentiment. That is super important in this new age of discoverability within AI. When we were talking about just-in-time merchandising, it's going to be about how do you decipher human trends through social media, data, content reviews, et cetera, and then translate that quickly into the manufacturing.
Dana Gardner: As the sophistication increases and more organizations can disrupt brands, for example, wouldn't it perhaps start blurring the lines and distinctions between digital and brick and mortar? Does it really matter where and how people intercept this information and then make a decision to purchase?
It seems like this might give the brick-and-mortar companies, if they get into agentic AI and make that sophistication part of their processes, a new distinction. Maybe they'll be able to hold onto their turf better.
Rosemary DeAragon: Absolutely. It's a great point, and one that I love talking about because traditionally, back in the day, you had brick-and-mortar department stores where retailers revolutionized the ability for consumer products and brands to reach the customer by putting all of those brands next to each other in a physical location.
Then you had the dawn of the internet, which caused an endless aisle where you have an endless amount of assortment and an endless amount of products in your “online catalog,” and where you could shop for all of these different items. And that was happening within the confines of a website. So, you still had a digital brick-and-mortar feel by landing on a website and seeing and browsing through all these products.
Now, with AI and large language models (LLMs), you've vastly expanded the digital four walls. You can speak to a LLM and discover products that you didn't even know existed.
For example, if you're building something, if you're a DIYer in woodworking, and you need a certain kind of hinge, you don't even know if this particular hinge exists, but you can describe the problem to a LLM and it can then crawl -- because they're trained on the world's data on the internet -- and tell you whether that that product exists or not, and then obviously be able to transact.
So, really, the initial disruption occurred in the 2000s with the dawn of the internet. But it's occurring anew now by expanding the product catalog again. It's beyond an endless aisle. It's beyond the digital four walls to be a true endless aisle of products and catalog data.
That's a really big threat to retailers because, if you think about it, the whole value of a retailer is to bring together interesting products that are similar so that customers can compare and contrast as they're browsing through the physical shop -- or through an online website.
Now that value is called into question, right? Because you can break through that by just having an AI conversation to find the exact item you're looking for. You don't necessarily need to browse through similar items. The AI ranks the question of what value the retailer is bringing to the table.
Dana Gardner: It's exciting because when people can use AI agents as an advocate for them as shoppers, they can be more intelligent, get the data they need, and it therefore puts more pressure on the retailers to step up because we could have dueling agents on the buy and the sell sides.
Rosemary DeAragon: Right! It's both exciting and also a little anxiety-ridden because those agents are going to be scouring through objective data, whereas retailers used to reach customers through emotional means via the look and feel of the website, the font, the logo, and the packaging, right?
All of that is out the door when the agents are the ones doing the research. Unless in the prompt you're saying, “Okay, make sure that the packaging looks sustainable using compostable materials.” The agent is using the hard data, the objective facts, to be able to find the best product for the customers’ demands.
As of right now, for example, OpenAI hasn't yet brought in the advertising model into ChatGPT. You are pretty much getting the true recommendation of a product without the influence of retail media, which again, is very anxiety-ridden for retailers that typically depend on that sort of monetization system to be able to push the products to the end customer. So, again, a lot of that emotional draw is taken out of it and replaced with an objective look at your product data and using that as the main source of truth for recommending the product to the end customer.
Dana Gardner: Now, if we're doing more machine-to-machine commerce – with the algorithms, LLMs, and AIagents all beavering away to make the utmost of this transaction between a buyer and seller -- in many cases it's going to hinge on the quality of the data. At the end of the day, that’s what delivers the best outcome.
But the ways in which we're getting the data are changing and being disrupted rapidly. It seems to me that all sorts of data -- structured, unstructured, and personal sentiment -- may be coming off of your smart watch, your social feeds, whatever. So we have to bring all of that together.
So tell me about this new era of all sorts of interesting data from all sorts of places being brought into this buying and selling equation.
Rosemary DeAragon: Absolutely. Another super-exciting topic for me is talking about how customer 360 as a topic is fundamentally changing. It's exciting for me because for many years, customer 360 – the customer data platforms -- how do you reach the customer? How do you build a well of data? It was all getting more sophisticated, sure.
But it was all in the same box, and once you understood it, you really understood it. And now, what's exciting is that customers are sharing much more intimately with LLMs than ever before.
Traditionally, in the past, you would stitch all these third-party data sets together. You would try to figure out the identity resolution of a household. Or you would try to enrich the data by saying, “Okay, this customer has this partner, has a dog, lives in an apartment.” You’d get that based on all of these different third-party data sources.
Now that is being completely flipped on its head because customers are telling you, “I have a pimple on a Friday morning and I'm going on a date with these three people and it's cold outside and this is what's in my wardrobe.”
Never in the history of the internet have we had customers so willing to share these extremely personal bits of information without the retailer being the one that has to stitch all of that data together from third-party partners.
It is such an exciting time because now you're having customers willing to divulge that information in exchange for the value of getting the right recommendation on the product. Now that is true customer 360 and we're finally landed in a place where we have that interaction digitally with the customer.
And we have trust as well. How do retailers take advantage of that to understand the customer in depth and be able to serve them better? And at the end of the day, it's better for all customers, especially if they're willing to divulge that information in exchange for the value.
Your question was also around use of the unstructured data. That might include the voice data from customer care agents, when you're speaking on the phone in real time. You're able to bring that data in as part of the customer data profile. Whereas before it was much more difficult or much more costly.
Dana Gardner: And even analyze the sentiment of that conversation?
Rosemary DeAragon: Absolutely. Yes. And now, with more and more video data, also in the unstructured world, you can use video as well as audio data. You could also understand based on how someone speaks and the slang that they're using, their age, probably of where they're from, or their particular regional location.
We're in such a world where you get a rich look at the customer, and they're benefiting from that by getting really good product recommendations -- as opposed to not really being sure. This is more of a probabilistic determinant of what you want to see. Now you're actually able to get down to the specifics of what these customers actually want.
Dana Gardner: Rosemary, your purview also includes travel, but we've so far been focused on retail and products. But this all applies to services, too, right?
Rosemary DeAragon: Absolutely. Anyone who's listening who is planning a trip, right? In the past, you would go to Google and look up “San Francisco to Tokyo.” But now you can bring in personal details, such as, “I have a toddler and I want to go to private dining. And I want avoid these routes, and I want the cheapest flight here, but maybe I want business class there.”
And providing that level of detail comes in exchange for a very personal response. And that whole thing is changing. I think next year we're going to see that be so much more in commonplace.
And then how does that disrupt the entire journey from an airline standpoint or a hotel standpoint? They often also use more emotional tactics to reach their customers. Now we're looking at the hard data to figure out the optimal path to figuring out a customized plan.
Dana Gardner: And we're talking about massive portions of the economy, right? So consumer spending is 70 percent of the economy, and the services add onto that. We're talking about a multi-trillion-dollar global market here.
So when you look at the economics, it sounds like we've been talking about agents as almost an intermediary third-party that can extract revenue for services. And the value goes in both directions, buyer and seller.
Isn't that all rather disruptive for the economics here, too?
Rosemary DeAragon: Yes, absolutely. Just thinking about those numbers, it's insane. And it's always going to land somewhere in the middle. I think you'll have people who say everything's going to be agentic.
But, in my opinion, until we get the main device manufacturers such as Apple that have a true agent -- and also true AI integration -- in the device platform, until we meet that moment, consumers will not be used to offloading transactions to an AI until the devices that they're using has that infused into the device journey.
Some say that Apple is late to the game. Some say it's all in good time. But given where they are today, we still haven't yet unleashed the power of AI and the personal device within the Apple ecosystem.
Because, if you think about it, Apple knows everyone you're talking to. They know your calendar, they know where you work, which social media asks. I mean, it knows everything about you.
You think ChatGPT is good, but imagine not even having to feed it that much information because your device already knows when your nephew's birthday is coming up and what they got last year. If you can offload the mental load of having to purchase a birthday gift for them, or knowing that you have a trip coming up, and offloading the mental load of tracking the cheapest flight.
All these things, right? I think once it reaches the personal device, that's when customers from a B2C standpoint are going to be much more comfortable offloading agentic workloads and more transactions to an AI.
Dana Gardner: Rinesh, you're up next. From your vantage point, how do you foresee financial services specifically being impacted by AI's advances and the prospects for greater visibility into ROI over the next year?
Rinesh Patel: It's a great question and one that's certainly topical at this moment in time, Dana. Look, I think there's a broad and universal acceptance that AI is clearly a powerful technology and that's something that we don't need to prove anymore. That said, what we do need to prove is the ROI.
So certainly the biggest shift that we're seeing in financial services is this change in mindset toward the commercial outcomes rather than the technical wins. And I expect next year financial services organizations to really be focused on measuring business impact of every dollar spent on their AI investments.
Dana Gardner: Is there something about the metrics, key performance indicators, and the ability to analyze in greater detail that's going to help solidify that understanding of the ROI in real, brass tack terms?
Rinesh Patel: There is going to continue to be a shift in metrics with a greater focus on ROI value and business outcomes. Organizations will be measured through the lens of metrics like experiences, hyper-personalization, real time insights, and retention.
Expect best-in-class analytics for ship managers and their client engagement activities, as well as for growth metrics like revenue outcomes tailored toward things like product recommendations and a variety of others like risk and digital adoption. That's where we're going to see consistency around the metrics as industries start adopting AI more pervasively.
Dana Gardner: And this increased productivity and value that you're measuring, you can do it focused on external factors like customer retention, the bottom line, the impact on greater addressable markets, share of wallet, and some of those key business metrics.
But there's also what AI brings internally to operations. Do you foresee the ability to measure that in terms of how AI impacts the bottom line via internal productivity?
Rinesh Patel: Yes, absolutely. Think about where there's going to be a focus, it's downstream in solving some of those areas that I just mentioned, specifically lines of business and supporting the business cases.
But absolutely, it's also going to be focused upstream, unlocking the data management efficiency gains across the entire data life cycle. AI is fundamentally going to be infused across that data life cycle, driving those productivity gains across the enterprise and across the organization.
Dana Gardner: And, of course, financial services organizations are often a bellwether for other verticals. They're an early adopter and an aggressive adopter of technology, especially when it can impact their analytics productivity.
And so I would think that 2026 is an important year in that financial services might be able to establish some common rules and expectations around AI's productivity that will permeate into other parts of the economy.
Rinesh Patel: I certainly see that when you look at any of the recent surveys out there. It certainly validates that hypothesis that in the financial services industry itself its customers are very much leading the way in terms of adopting and implementing AI and uncovering some of the best practices.
Those are things that other industries will certainly be paying attention to in the hope that they can learn from and evolve from. I certainly see that. And there will be some social benefits to the work that's taking place in financial services across some of the broader industries that my colleagues, Rose and Tim, are working through.
Dana Gardner: And while financial services is an early adopter, it's also somewhat unique in that it has to be very conscious of risk, security, and governance. And so, how is AI risk management evolving in 2026? It might also be a bellwether for other industries.
Rinesh Patel: Yes. This is a highly regulated market, and trust is important and more than essential. If you think about AI agents and LLMs as they become embedded in critical financial operations there is already a greater focus on mitigating risks with a heightened focus on responsible AI.
When I think about some of those AI risks, they're going to evolve from hallucination, safety, and ethics, to also factoring in how they collide with data risks. Things like data residency, operational resiliency, and so forth. It’sclear that AI, data, LLMs, and agents can't be disaggregated. You need a unified approach to data governance. And so, AI evaluations are going to be essential to support risk management across financial services.
We’re already seeing signs of this industry taking some thoughtful steps in terms of putting those processes and practices in place.
Dana Gardner: Is there anything in particular about data management and governance that you think 2026 will advance? That is to say, if you take care of your data lifecycle, you're in a better position to forestall any issues and risks. How important is the data management phase of this overall risk management process?
Rinesh Patel: The data management is going to be essential. There is, per the phase, no AI strategy without a data strategy.
As an industry, we've already evolved to accept that it's no longer about AI first, but data architecture incorporated with AI as part of the strategy. That's the first thing that we're certainly course-correcting. The second is, as I mentioned a few moments ago, a unified approach to data governance and AI evaluation is now becoming essential.
The ROI just won't be there when siloed data erodes the trust in AI systems. And the third area is very much focused on the evolution of governance broadly from traditional data governance to modern governance and usage governance. So I think governance is going to be essential both across the data space, but also the LLM and the AI space. And I think organizations are certainly focusing on it in that way.
Dana Gardner: We've certainly seen a very rapid advance and dynamic landscape when it comes to AI and technology itself, but there are other very fast moving aspects to this point in time. Regulators are moving, changing regimes, or changing in terms of oversight.
And then we just don't know about what risks may or may not appear. The so-called unknown unknowns. From what you mentioned about unification, that being agile in fleet and how one can adapt and adjust would be an important factor.
Is there anything about financial services organizations that you predict will be required in terms of being able to pivot as needed in order to keep up with a changing world?
Rinesh Patel: That need to pivot and to be agile is going to be coming from a variety of stakeholders. I fully expect regulators to continue to signal tighter oversight. And for a global bank, that will mean global regulators overall from the US to EMEA to a APJ that they will have to engage with and make sure they are working with.
But it also will be the boards. Boards of organizations will demand comprehensive risk frameworks that treat AIdeployment as seriously as any other mission critical system. So, it'll be coming from both the internal stakeholders and the external stakeholders like regulators.
It’s going to force organizations to take a more strategic, unified approach to understand data governance and AI evaluation more seriously and practically.
Dana Gardner: The one thing that we can count on as being consistent is more change.
Rinesh Patel: Absolutely.
Dana Gardner: Let's move on to how the adoption patterns and the culture of using AI could change over the next year or so. How will agentic AI manifest itself, particularly in financial services? And why is that an important leap in how productivity will deliver results?
Rinesh Patel: Providers of data typically have been providers of structured data that's now evolved to both structured and unstructured data, and consumers have typically been humans and applications that now very quickly will become agents. And the way information is distributed from providers to consumers is going to be very, very important to enable that true agentic experience.
By agentic experience, I mean the ability to surface up insights to business leaders through natural language capabilities that allow enterprise intelligence to be pervasive to every user effectively and at their fingertips. And what you're going to see is effectively a natural language interface, an agent workflow orchestrator, that enables an enterprise to truly extract meaningful value from the data that it has and the insights that it holds within.
Dana Gardner: It sounds as if you believe that agentic AI will be an accelerant or a catalyst to the more of that democratization, that more people will be able to take advantage of, interact with, and perhaps optimize the AI value chain.
Rinesh Patel: Yes. And we're already seeing some of this pan out. We’ll see an agentic experience quickly become reality, if not the norm for most businesses. I expect financial services to start introducing AI agents into core business processes -- from risk monitoring, to surveillance, to customer reviews, and to portfolio operations.
These systems will take on the multi-step work traditionally handled by teams of analysts to create a very different set of challenges for leaders to navigate. The first will be how do you measure productivity in a world where humans and AI collaborate?
The second is how do you govern agents that make decisions and take action? So there's going to be a major change in the operating model of organizations as they straddle across these two.
Dana Gardner: It sounds like an advance and a whole new capability when it comes to process re-engineering. You have to look at this top-down and strategically to refactor how the individuals, the teams, the culture, and the technology all interact. That's a big undertaking.
Rinesh Patel: Yes, absolutely. What we're going to see is firms moving from measuring tasks handled by people to evaluating the performance of blended human-AI workflows by things like speed of detection, accuracy of decisions, consistency with policy, and overall business impact.
Dana Gardner: Tim, from your vantage point, how do you think global manufacturing specifically will be impacted by AI's maturity over the next year or so?
Tim Long: We're going to see tremendous impact fueled by the challenge that manufacturers are facing today, where there's more work to do than there is skilled labor to meet the needs.
Manufacturers across the board, not just in the US but really globally, will need for AI to play a role to help augment skilled workers, to give them new capabilities not only to help them be more efficient in their work, but also to help them level up so that their organizations can innovate faster, produce better quality products, and ultimately lead to the growth of their business.
It's an exciting time in manufacturing as the world is rebalancing where products are made. Also, the role that AI will play is going to be a tremendous transformative force.
Dana Gardner: For all those pipefitters out there, their tool belts, they're going to have to make another little slot for their AI companion. Is that it?
Tim Long: That's right. AI can really help everybody. We see it in our personal life. There's not an hour that goes by where I don't see the opportunity for AI to help me personally. And, of course, it'll help those that are doing more traditional blue collar work.
We're seeing this already with the ability to tap into large databases of historical records of how different projects were completed or how different repairs have been made in equipment. And these AI tools can answer questions much more efficiently than thumbing through a repair manual or even going into the web and do a simple search.
This is a transformative, what I would call a step-function improvement in the efficiency of our skilled workforce.
Dana Gardner: Now manufacturing has long been a source of great data at the edge, whether it's sensors on a factory floor or logistics and tracking shipments and deliveries for just-in-time.
But maybe that data hasn't been as well used as it could have. Is this a function of using the data that's existing better, or generating more data and therefore more intelligence?
Tim Long: Well, I really love that question because what we're seeing in the manufacturing world is this transition from where manufacturers were very comfortable operating in just on-premises, at the edge software solutions exclusively.
In fact, many manufacturers purposefully close off the network of their shop floor from any external networks such as the cloud or the internet for all kinds of great reasons: reliability, cyber threats, and so on. Now manufacturers are recognizing that the cloud is essential. Not everything's moving to the cloud, but the cloud plays an essential role, specifically in places where all of this data that was managed in the siloed systems at the edge is now coming together in the cloud.
And this is more commonly being called the unified namespace, it's a term that was coined by Walker Reynolds. For those in the manufacturing world have probably heard of him. And the concept here is how do I bring all events from the business, whether it's a transactional event in my ERP or it's a manufacturing event for the shop floor together in one place, tying together the entire stack of systems that runs a manufacturing organization.
So I can really see end to end from shop floor to top floor, what's happening in my organization. And that trend has really taken off over the last year and will accelerate into next year converging IT data with the operations technology or OT data.
Dana Gardner: This has been the nirvana since process re-engineering 30 years ago, and maybe AI has certain great capabilities on its own, but it seems to be an accelerant or a catalyst to doing this end-to-end unification, which is long overdue.
Tim Long: I completely agree with you. If we look at sort of this evolution of manufacturing, we can look at what's been known as the Third Industrial Revolution where computerization and automation really made a big impact to manufacturing processes being more efficient and improving quality. But that data always rested in these siloed systems.
And so now with AI, if I tie that data together in platforms like Snowflake, I can now ask questions that can work across the organization, across each of those silos. Bridging data from orders to the bill of materials to create that order to the raw materials to my supplier base, as an example in supply chain, where all that data can now be tied together and insights can be delivered simply by asking questions.
And I think that's what AI is offering as a value incentive. For manufacturers to bridge those data gaps.
Dana Gardner: I suppose in another area, whether it's just-in-time or whether it's build for manufacturability and design for manufacturability, that AI can have an impact there as well. And so just-in-time, faster and manufacturing with fewer components. Is that something that you're seeing as well in terms of this overall productivity that AI can increase by this unification the whole life cycle of manufacturing?
Tim Long: Absolutely. A great place that AI is being leveraged is in improving supply chain speed. How do I sense opportunity? How do I sense risk, and how do I make decisions that better position my organization to execute in whatever changes are happening within my market, whether those are geopolitical economic impacts, or even impact traditional supply chain disruption?
We all very familiar with all of those events when manufacturers are equipped with the right data and tools that they can make decisions faster that position them ahead of their competition.
Dana Gardner: Now we're all interested in finding ways of measuring the productivity. Measurement in technology has been a difficult thing for quite some time, but is there something in the manufactured control environment that you're describing that will enable more disciplined IT investment validation?
Tim Long: I think of manufacturing as an ideal place to prove the value of AI engineers in within a manufacturing process.
By operating using the scientific method every day, they can create controlled experiments to say, “If I adjust my process with this change, can I improve efficiency or quality?” And implementing AI should be thought of in the same dynamic. Can I create an experiment that proves the value of this new approach to doing my business process?
So manufacturers are accustomed to this way of learning from trial and error. I think to answer your question, absolutely, manufacturing well positioned to do this and the dividends are huge. If you think about the cost of software coming in and improving your ability to deliver more products with better quality, that cost compared to new capital equipment, expenditures, or expansions of your factories or purchase of more materials, that cost is much smaller in the big picture.
So, the ROI is there. And I think manufacturers are figuring out how to prove that as they're delivering on their AI roadmaps.
Dana Gardner: Because of the large investments required for many manufacturing environments, automotive comes to mind, where you have tooling and long-term process involvement with highly complex supply chains, is there some way that you think AI can help in moving the risk around the cost of, of investment, and then the payoff in, in terms of getting the manufacturing and products that, that you wanted in the first place?
Tim Long: Absolutely. Any efficiency gain that you can drive back down to the floor, either making your equipment available more often more, or improving their uptime so they're available 24/7, that's a big way to make an improvement. If you can make your equipment produce products at a faster pace, increasing the throughput, that's a big improvement. And lastly, if you can improve the quality. That's a big improvement. In fact, those three dimensions I just described make up the manufacturing king metric known as OEE or Overall Equipment Effectiveness.
And to do that, you need to bring data from your equipment management systems, from your industrial IoTsensors that are monitoring that equipment, you need to bring data together from the performance of the equipment itself, how many widgets per hour is it able to produce? And, then lastly, you need to understand all of your quality, whether you're measuring that physically or virtually with things like computer vision, having that data tied all together really is empowering manufacturers to improve efficiency and overcome all these challenges that they've experienced within the markets.
Dana Gardner: How about AI for simulations and experimental pilots so that you don't have to take the risk of building out large and complex factories, but can perhaps simulate and determine beforehand where your strengths and weaknesses are?
Tim Long: Absolutely. This is in manufacturing world known as the digital twin. How can I model any part of my factory or that my actual end product to understand its performance in different conditions? Can I capture enough data over time, both in the context of the floor itself as well as the, how the equipment is performing and understand relationships that are otherwise difficult to maybe to appreciate?
But if you can do that, you can capture that -- think of it as a physics model that then you can experiment with, and those simulations can really unlock the true potential of your shop floor on the connected product side. So if you think of manufacturers, you mentioned automotive. These vehicles that are shipping every one of them is essentially a supercomputer on wheels now.
And all the data that they generate and collect has value not only to inform your lights on your dashboard, but to provide an insight back to the engineering and design organizations to say, “This is how our product's actually being used. These are the challenges our customers might be facing with it in terms of reliability or performance, and how do we learn from that to feed back into the design process, to create a better product for the future?”
Dana Gardner: Yes, that lifecycle can be a huge benefit. That historical data, we've seen it in other industries, but now going to manufacturing where the costs you say are so high, it could be huge.
So we've obviously established that there's a great deal of opportunity for AI to come and prove itself rapidly and significantly in manufacturing. But how do you expect the AI and agentic AI in particular to manifest itself in manufacturing organizations in order to prove its importance and value?
Tim Long: I think that agentic AI is just an extension of the investment and automation manufacturers have been making for a long time.
What these agents are doing now is they're able to take on repeatable decisions that an organization would otherwise need to make. And they're able to do this with rich data sets and with the guidance, and they're able to make decisions at a high speed. So again, it's all about speed within manufacturing while not introducing new risk.
Manufacturers are looking to these agents to do things like automatically reorder when materials are low. Figuring out which products should be expedited in line to achieve their commitments to their customers for shipping things on time, so how to optimize the logistics. How do I get it from my factory to my customer in the most efficient and timely way?
There are just so many opportunities where agents are starting to make an impact, and we will really see that accelerating over the next year.
Dana Gardner: How do early adopters put themselves at a competitive advantage by doing this sooner rather than later? I should think that becoming a digital-first manufacturer when you're out there in the field competing with a non-digital-first manufacturer, that you should get some pretty good benefits.
This isn't a let's crawl, walk, and run type of an affair, right?
Tim Long: I get that question a lot, and what I see are those manufacturers who have invested in the physical automation of the shop floor stand to benefit the most. And the reason why is they have now these rich data sets. And now the question is how do I synthesize these data into insights?
And that really comes in the form of integrating those data across the silos in a platform like Snowflake. Creating that unified namespace where all business events are tied together and doing it in a way that's not just fit for dashboarding, as was the previous goal. Creating these data marts to, to run these dashboards.
But you can accomplish even more by thinking beyond that and providing things like semantic models that help AI understand how to relate data from one data object to another. That is the key to success of using AI to really unlock the value of the data is to position the data in the best way on a strong data foundation with strong governance, and of course all the guardrails that are necessary to deliver the value from AI while mitigating the risk.
Dana Gardner: Well, thank you very much for that, Tim. Let's now move on to our round robin portion of our discussion today among all of our experts. Tim, now that we've heard about AI's potential across several important industries, do you agree that the ability to be precise but agile and customized to specific factory floors is a big improvement over what we've seen in the past from software and when it comes to, to optimization?
Does AI give us an opportunity to fine tune at a much more granular level our efficiencies?
Tim Long: Yes, it's an interesting question. Manufacturers tend to have a very diverse set of systems from one factory to the other. They generally grow through merger and acquisition, and as a result, it's really hard to take one solution from one location and replicate that across the network.
AI is going to help with that. We're already seeing AI help with things like data migrations, mapping source system to target system. Rather than doing that manually across thousands of data objects, AI can take the lead and get you 90 to 95 percent of the way there. And so I do see that that complexity simplifying over time with AI playing a big role.
Dana Gardner: How do you think the adoption patterns in 2026 will help organizations value AI, establish itsbenefits, and perhaps remind people that they may be undervaluing AI? I mean, we often hear about a wide variety of perceptions in the market as to AI's performance and impact. It seems to me that there's still quite a few people that are underestimating AI's impact.
Do you think we'll move past that in in the next year or so?
Rinesh Patel: I think we will. There's a broad acceptance that AI is clearly a powerful technology, and that's not something that we need to prove anymore. I think it's now about how we leverage it strategically and how do we make sure we put the right focus on the metrics.
I fully expect next year to be an acceleration of adopting the technology, of implementing the technology, of measuring some of those outcomes. And we're going to see a lot of winners and a lot of people having to learn very, very fast.
Dana Gardner: Okay, well, we're about out of our time, but I wanted to ask you a question that will be addressed across these different vertical industries and that's who's your poster child?
Who's doing this well now that is a harbinger for others to look at, even if you can't name them. Describe a use case or a situation that you're aware of that that highlights and exemplifies the potentials that we've been talking about.
Rosemary DeAragon: One of the most advanced retailers that is a public reference for us is Under Armour. They are super advanced, with credit to Patrick Duroseau, the Chief Data AI Officer, for pioneering a lot of the AI efforts at Under Armour. They have done things like data sharing with their logistics providers. They've gotten into clean rooms and now they are actually building out fully agentic workflows between internal and also with their customer service use cases as well.
So when somebody has a question about, “Okay, how do I download Zoom,” or something like that. It is all automated. So being able to replace a lot of the more manual effort of looking through documentation, they're using the RAG method, using AI on top of all of their internal data sets with Snowflake to be able to automate those workflows.
So big shout out to Under Armour for being a pioneer and really a good partner for us as well in testing out a lot of our AI features.
Tim Long: There are so many, but if I had to pick one, I'll pick one that just illustrates the effectiveness of AI at restoring high throughput on a shop floor.
We have a semiconductor customer leveraging years of history of equipment maintenance logs, and all of the repair manual information about these very expensive and complex equipment. And they're loading that within Snowflake and using Snowflake Cortex and Snowflake Intelligence to help them maintain peak level performance across their shop floor. There's no better example of how AI can help manufacturers than to make sure the shop floor is running at peak performance.
Rinesh Patel: The one that really stands out to me is the focus on unstructured data. Now, the finance industry has a lot of unstructured data that resides across an enterprise, across an organization, and I think the one that's really compelling right now is our organization uncovering new potential sources of alpha by leveraging gen AI in LLMs to uncover untapped alpha.
The unstructured data beyond traditional natural language processing. I think that's going to become very much a theme, a pervasive theme, of opportunity for financial services. There are a lot of organizations, specifically on the buy side, that focused on that, and I think that's a really interesting area of exploration going to 2026.
Dana Gardner: Well, thanks so much to our latest Data Cloud podcast guests. We've been here with Snowflake’s Rosemary DeAragon, Global Head of Retail and Travel; Rinesh Patel, Global Head of Financial Services, and Tim Long, Global Head of Manufacturing. We so much appreciate you sharing your thoughts, expertise, and experience with us all.
Want to see what's next for apps and generative AI and the data cloud? Check out BUILD, Snowflake's annual developer event. Dive into the latest innovations like Snowflake Intelligence, now GA, and explore how developers are building powerful apps, data pipelines, and machine learning workflows for the LLM era. Watch on demand at snowflake.com/build.
(Snowflake supported the creation of this discussion).
[Listen to the discussion or watch it.]
Comments ( 0 )