Building Enterprise AI Transformation: Governance, Data & Culture

Summary

New York Power Authority (NIPA) evolved from targeted AI projects in predictive maintenance to enterprise-scale transformation after the generative AI shift in 2022 accelerated demand from every business unit.
NIPA built an AI center of excellence and deployed self-service data products through Genie to democratize access, reducing dependence on data scientists for routine analytics.
Sustainable AI transformation requires multi-tier AI literacy programs, multi-layer governance spanning policy, technology, and organization, and unified executive alignment to balance adoption with cost control.

Building Enterprise AI Transformation: Governance, Data & Culture

Enterprise AI transformation extends far beyond individual use cases or technology implementations. Ron Carroll, Chief Data and AI Officer at New York Power Authority, and Traci Gusher, Partner at EY, discuss how leading organizations align governance, data democratization, and cultural change to move from isolated AI projects to enterprise-scale transformation.
Explore NIPA's journey building an AI center of excellence, deploying data products through Genie for self-service analytics, establishing multi-tier AI literacy programs, and implementing governance across policy, technology, and organizational layers. Learn how to balance adoption and cost control through unified leadership alignment, leverage modern data platforms for unstructured data classification, and shift from process automation to end-to-end AI-first redesign.
🤝

Chapters

FAQs

What is the New York Power Authority's role in AI transformation?

The New York Power Authority (NIPA) is the largest public power authority in the US, generating about 25% of New York's power with roughly 80% from hydroelectric sources. NIPA built an AI center of excellence and leveraged AI for predictive maintenance on large assets, evolving from data scientist-driven projects to enterprise-wide adoption following the generative AI shift in 2022.

How does NIPA use Databricks Genie for data democratization?

NIPA deployed data products through Genie to enable self-service analytics, shifting from a model where business units had to request data from the data team to one where they can access insights independently. This approach supports broader AI literacy and reduces bottlenecks in data access across the organization.

What governance framework does NIPA use for enterprise AI?

NIPA implements a multi-tier governance framework spanning policy, technology, and organizational layers, ensuring AI is governed consistently across the enterprise. This includes establishing AI literacy programs at multiple levels and aligning leadership to balance adoption with cost control.

Why does enterprise AI transformation require cultural change?

According to this video, the technical implementation of AI is only one part of transformation — the operating model, governance structures, and cultural readiness of an organization are equally critical to proving AI's value. Organizations must shift from process automation toward end-to-end AI-first redesign, which requires sustained executive alignment and employee upskilling.

Full transcript

[00:08] Yeah. Uh, good afternoon everyone. I'm Ron Carroll, the uh, chief data and AI officer at uh, the New York Power Authority. NIPA, you'll hear it referred to during the, uh, the presentation. Um, maybe just if you're not familiar with NIPA, just give you a little bit of, uh, background. Um, so we are the largest
[00:24] largest public power authority uh, in the US. Um, we generate about 25% of the power for the state of New York and about 80% of that is hydro, so green energy. So, you may not know who we are, but now you do.
[00:40] Yeah. I bet if you live in New York, you're grateful to have you. I think so. Yeah. Um so um the discussion that we wanted to have uh today is um we're going to go I think uh a little less technical than um maybe some of the other presentations that you've been to or um a lot of the more
[00:57] in-depth either training or um presentations that um you've experienced. We really want to talk about AI as a transformational le lever and beyond the technical side of what has to happen with AI. Um what has to happen uh around the whole the the entire ecosystem and operating model of
[01:13] AI in order for it actually to be transformational in order for it actually to to to prove valuable. Um but but but with that in mind, I think um maybe we'll start the conversation with a question that I get all the time from from my clients as well as um my parents
[01:28] and friends and so many others is you know where actually is AI um generating value in the organization and I thought maybe we'll start with you know how NIPA is looking at this maybe some of the areas where you're getting some transformational value out of AI or um
[01:44] if you're not there yet um where you're really targeting um and let's let's start the conversation. with talking about meaningful ways that that it's it's proving value. Yeah, great question. Um, so we've actually been using AI for a number of years. Um,
[02:00] primarily in the predictive maintenance space. So, uh, we have large assets that if they fail, it's expensive and the power goes out. Um, so, uh, we've been leveraging AI for a number of years. Uh but what I would say is in the past what
[02:16] we were seeing was AI was more of a data scientist activity. Um and then we had the chat GPT moment in 202 two. Um wow time flies. Um and so once
[02:33] that happened AI really started to become more mainstream and AI for everyone. And so we used to be going out and knocking on our business unit doors to say, "Hey, do you have any use cases that we can work on?" And they would,
[02:48] you know, look at us and say, "Well, I think we can do this." And we'd work on something, a PC, and we'd deliver. And they would go, "Oh, wow. That's cool." And then they would move on to the next thing. Um, but now with generative AI, I I think we're seeing I mean at this
[03:04] point we've collected over 170 use cases and just some examples of the use cases that that we're either working on or have delivered. Um, you know, looking at our assets again
[03:19] and uh a big space we're seeing is knowledge management. And so we have a workforce that is uh a lot of folks are nearing retirement or retiring. Um we have a lot of information that's uh in documents uh in our uh asset management
[03:37] system. Uh and so it's it's critical that we surface that information and that knowledge. And so, you know, leveraging AI, you know, one of the the things that we recently delivered was an AI assistant, our our um asset
[03:53] intelligence assistant where somebody that's out in the field working on some of our assets. They can they have a mobile device, they can type into it questions about the asset that they're working on, um how they would repair it,
[04:08] how they would diagnose a problem. Um so, we've delivered that. The next version of that is audio. And so we would have somebody out in the uh in the field and they would be speaking to an AI assistant and they can ask, you know,
[04:24] the same questions they would be typing. It would respond to them because if you think about folks that are out in the field, they've got protective equipment on and it's not very conducive to typing on a a you know iPad or especially if you're up at the top of a
[04:39] pole. Yeah. Right. I would I would think taking your hand might go and you kind of you know hold on with an arm or something. So um so that's the next version. Um we're doing a lot with renewables. Uh you know New York Power Authority is uh
[04:56] one of the things that we do is really look at the state and one of the things we're our goals is to help the state decarbonize. Um so we're doing a lot in terms of um figuring out where we should be installing renewables. Um where is
[05:11] there congestion within the within the grid? Um where is uh pricing um you know for battery storage? Where does that make the most sense? When should you uh store power to a battery versus when should you use power from the battery?
[05:28] Um so those are just a few of the the use cases that we're working on. we're doing things with our our customers and customer 360. It was really interesting to see, you know, some of the announcements today. Um what was I there were so many names th thrown out but um
[05:45] you know we were looking at maybe we can leverage some of the customer uh capabilities that are being introduced um for that and you know I I it it's really I I I I find it fascinating right now to look at the innovation that's happening in um a
[06:01] lot of power and utilities companies not just because of the historical problems and issues that you're always trying to solve for like you know preventing outages Right. But also, you know, there's this constant debate right now in the um the energy consumption that AI
[06:17] is producing and the um how are we um you know as a country and as a and as a globe going to be able to produce the level of power that is needed at least given the constraints of our technology today. And I I don't think that we're
[06:33] going to meet that. It's funny. We're not going to meet the need of AI without using AI to meet the energy need. and that like it it really does come full circle. Um that the one of the things that that that I take away from where um NIPA has been getting value as well as where
[06:50] you're targeting value is that um you know those the the the place that you started in looking at things like predictive maintenance um is is really classic AI, right? It's you know if you think about drones that are are looking at the lines, you know, there's computer
[07:06] vision in that. if you're looking at your your asset base and where failures might occur, that is good old classic machine learning. And I think that there's um a lack of focus that there's there started to be almost an overindexing on generative and agentic
[07:21] AI. And it's it's not I I think we need to be thinking about about those that that the usage of that because there's tremendous value there. But I also think that we need to always look at the entire toolbox of AI and make sure that we are looking at it as a the right tool
[07:39] for the right job at the right time. And I like to say like the answer for everything isn't a large language model like why bring a chainsaw to a ribbon cutting party. the, you know, if if if a good old classic machine learning model is gonna is going to get you the outcome that you need, why not just let that be
[07:55] and and not force everything into the newest the newest sexiest um technologies. Um I I'm just going to put a a couple things up here in in the background for us. Um but um sorry, there we go. Um but um one one of the
[08:11] the the things that that maybe we can shift to Ron is um when we think about those use cases both those that you've been investing in and those in the future um you know obviously you've you've got you've gained value out of them but there's been challenges along the way and there's new and increasing challenges that are emerging um in order
[08:29] to really gain the value and and have the ROI um what are the biggest challenges that that you're seeing and you know we've talked a lot about data we're at the data bricks conference maybe maybe we start with with where your data challenges or maybe uh are or
[08:45] maybe where you've started to overcome them. Yeah, I I think in the the data space, we we've had a data governance um program in place for I've I've been with NIPA seven and a half years, so seven years or so, six and a half years. Um
[09:01] but a lot of our data governance and quality in the past was focused on structured data. And I think what we're seeing now is, you know, with generative AI. Um, and I agree that a lot of our use cases actually use classical AI and then generative AI to kind of access
[09:17] that information. Um, but you know what we're seeing with generative AI is a lot of the unstructured data and the challenges with um, you know, categorization of the data. um an initi in initiative that we're kicking off shortly is around um you know looking at
[09:37] we're worried about data exfiltration and so how do we classify data to make sure that you know we we know whether it's public data or whether it's private or whether it's sensitive and who has access to it and we've done that really well with structured data
[09:53] unstructured data eh not so well right and so I was at a conference conference a few weeks ago and somebody mentioned I don't know if this is true or not I I don't know it was somebody on stage like me so take it for what it's worth but um I think they said it 90% of
[10:10] organizational data is unstructured 10% is structured so I don't know does that seem about right it's I I know it's it seems like there's more unstructured data in our enterprise than there is structured data um and the challenges of
[10:26] the versions of that you you think of the um the search or the contextual search that I mentioned earlier. Well, when you're asking the question, how do you know that you're getting the latest um version of the manual for that asset
[10:42] and the latest uh work orders are included in there? Um and and that's where capturing the metadata around that unstructured data. And what we've been been working on is is our modern data platform. um you know as we're ingesting
[10:58] data starting to capture that metadata categorize the the the data um so that we can use it for AI use cases and and I know that one of the one of your initiatives at NIPA is truly to start to democratize data to to end
[11:15] users and um you know I think that the especially with the way that AI is being put into the hands of you know kind of every employee across the organiz organization. I I think it's really interesting to think about what happens
[11:30] when you have an entire employee base that not only has access to AI tools that can help them build their own agents or or build their own small applications, but when you combine that with data that truly has been democratized for usage and especially if
[11:46] you're gaining, you know, that power of connectivity through a platform like data bricks, there's some really big things that can happen. But when you're looking at data democratization, you know, what are some of the goals that you have in in in that program and what are what are some what's some of the value you hope to get out of it?
[12:02] Yeah. And our data de democratization we started six and a half years ago, seven years ago. Um, and we started with data sets. You know, the thought at that time was well people they want to shop for data and that's going to give them uh,
[12:19] you know, self-service capabilities. And what we realized was well people didn't want raw data. So then we went to well shop for reports. Okay that's a little bit better. Um but still what people want is a curated data set that they can
[12:38] use to an analyze their business. Um and so now we're developing uh data products. And so I I think you hear that often within the industry now. Um, so you know, rolling out data products, we we've rolled out our initial data
[12:54] products and are designing our our next wave of of data products and it they're all aligned with with use cases. Um, but some of those use cases are around democratization and you know using tools like Genie. Um, you know, and this is
[13:11] where I think it's it's a little bit of a culture change for our our organization because some of our business partners, they use Excel and, you know, Excel is what they know and we're building our modern data platform and they're like, "That's great. You're going to bring
[13:27] this together as a data product. Then I can download it to Excel and do what I need to do." And I'm like, "No way. That's, you know, now you've exposed the data and we we've lost control of that." So um you know we're really trying to change the culture so that people can
[13:42] look at data differently and the way that they access it and the way that they um they use it. You know our dashboards and BI reports going to be what serve the same purpose in the future? Probably not because I think people are
[13:59] just going to start asking questions of the data and you get your response back. So, you know, the way that you're using, you know, Gemini or Chat GPT or whatever you're you're using. Um, so that's that's the goal. Um, I think one of the other things that we've we've seen is in
[14:17] this culture change that we're trying to go through is talking about data products and people get it, but then it's like, well, how do I use that? What does that mean to me? Um, another thing that we're trying to do is to allow
[14:32] people to bring their own data. So, here's a platform where you can bring your own data. We have data products. If you have access and should have access to that data product, you can bring those two data sources together and start to, you know, perform your your analysis. Um,
[14:50] we need to help them do that and we don't want to have it in the middle of that. So, we're actually building a knowledge agent so that somebody can go to the knowledge agent and say, "Well, I have a data set. It's in Excel. What do I do?" And we will provide them the
[15:08] instructions, not just a link to, "Well, here's a 20page document that tells you how you can upload your your data and and start to use it." It's more of a, you know, interactive conversational uh knowledge agent um to to help them. We're hoping that will help to uh you
[15:24] know promote the usage of the data. Yeah. And I I I I think it it for for so long um the ability to to to utilize to to to get access to data to to utilize data has been so focused on I'm going to take data and I'm going to move it here
[15:40] so that I can manipulate it in this way. I'm going to take data and I'm going to build this dashboard so that I can see it in this way. And when you look at things like Genie as an example and it giving you a completely new way to interact with your data and to you know a ask to to to gain insights from it, I
[15:57] think it's it's a it's another step forward in the right another step forward to data real data democratization and and I I think we're going to going to going to keep seeing um a lot more um a lot more technologies
[16:12] and approaches that allow us to do that. Um I I was also um we were talking about this earlier but you know when we when we think about the advancements of of integrating with data and and and truly using data as a mechanism for AI transformation um I think you know all
[16:28] of us have have heard a number of different really interesting um announcements this week from data bicks um on things that that are are are new in the portfolio and um you know I was really excited yesterday to to to hear um about Unity Gateway because the cost
[16:45] of of AI is something that everybody is worried about, right? Um and you know having the ability to look at um you know both costs across MCP agents etc and be able to have that observability layer I think is is really important but
[17:01] cost is is is a is a a big topic right it's not just about how much your platform costs it's not just about the tokens it's about so much more and um both from a cost perspective as well as from a real implementation perspective
[17:18] I actually think that the bigger challenge to cost is culture and talent and people and change. Um and so so maybe maybe we'll we can shift away from you know data and technology for a second and talk about what it takes to
[17:36] really drive transformation and and maybe we can start with with with culture and talent as as a as a key enabler of that. Yeah. Yeah. And I I mean it takes investment and I think what we were experiencing was um I would joke that
[17:52] you know this is a hobby for everybody. Um you know when that chat GPT moment happened we stood up a few different work streams focused on governance and risk management and change management and technology and and platforms and use case um prioritization.
[18:10] But it was uh it was a hobby for folks. it was an add add-on to what uh everyone was doing. So, you know, the the first thing I would say is we we're making investments. So, we're hiring resources that are focused on AI. Um we've stood
[18:26] up an AI center of excellence. Um part of what that center of excellence is focused on is AI literacy. And so we have a uh multi-ter strategy on how we will uh you know upscale the
[18:41] organization um starting from you know how do you spell AI because I I think there's you know all of us in the room we live and breathe this every day right so it's it's kind of natural to us but our workforce I mean we have folks in in
[18:58] you know management we call it management and union so um you know kind of the administr ative type of jobs and knowledge workers and then we have union workers and the range of knowledge is vast and so we need to kind of level set
[19:16] and and baseline offer that know that information to everyone. Um and then looking at well how do we start to um you know focus on leaders and how do they drive change within their organization
[19:31] and then um democratization. How do we train people so that they can become citizen developers? Uh and then overall, you know, our hands-on technical, you know, like folks in this room, up
[19:47] upskilling them and then our executive management um group, we need them in the right headsp space to drive the strategy. I mentioned earlier that we have about 170 use cases. A lot of those are coming from the kind of the ground level. And so we need to meet in the
[20:05] middle. we need the you know our executive management group to kind of drive the strategy and meet meet in the middle there and it so so it's been interesting for me to look at a lot of organizations and what their strategy has been around education and AI literacy and and and
[20:21] where it's happening in the organization and what's what's been interesting to me is um is that um I'm seeing the you know new entrance um you know newer um to the workforce um resources as well as executives being really interested in
[20:39] being upskilled and being, you know, knowledgeable beyond just I know how to write a good prompt, right? Um, in fact, you know, at EY, I was I was asked to provide reverse mentors to our entire executive committee, to our entire
[20:55] management committee. And the the mentors are like senior associate manager level resources that are hands-on keyboard AI engineers. Um, but they are every one of our management committee has one of these mentors. And um, at first I thought what they were
[21:10] asking for was somebody that would, you know, show me how to use this tool, give me some advice on how to prompt better. And and then I started digging into what the actual curriculum was that this team was providing. And, you know, I I I talked to one of these these reverse
[21:26] mentors one day and he said, you know, tomorrow's lesson is the difference between training and inference. and the lesson next week is we're going through regression. And I'm like, whoa, wait a minute. Like, you're not just like how to write a good prompt. Like, you're actually going into this. And they're like, this is what they want. Like, they
[21:42] they don't want to just be peripheral. They want to get deep. And a lot of our newer um you know, uh uh our our junior workforce just coming in. Like, they've been they they in 2022 started to be told that it's not, you know, AI that's
[21:58] going to take your job. It's it's people who know how to use and leverage AI and they've embraced it. And so for years they have been focused on that. And what what's most interesting though is that the top and the the newer entrance are the ones that are diving in the deepest
[22:14] and that's our middle layer of organizations that are not digging in as deep. And in fact we did a survey probably six nine months ago at this point. But um one of the questions in it was it was an AI poll survey to get you know feedback from from companies on
[22:30] where AI is is um is being utilized and and what the challenges are. And one of the questions that we asked um was related to management of agents. And what was interesting was 67 or 64 or 67%
[22:46] of the employees that responded to the survey said they did not want to manage be in a position where they were managing an agent. I looked at it, I'm like, "What is that?" And I I I I kind of sat back and tried to do my own synthesis on it. And I reached out to some colleagues of mine that are behavioral um behavioral change
[23:03] management specialists and I said, "Can you help me unpack this statistic? What does it mean?" And it was interesting because what they told me was on one side of the equation was that um there is a fear that if I manage agents then
[23:20] eventually who the person managing me is going to manage my agent meaning I'm going to become an agent and so if I refuse to if I don't want to if I refuse to be a manager of agents then I'm not going to be replaced by an agent. That that was part of it. The second part of
[23:36] it was that um when we look at individuals that are in a like one to five or six year um into the market into their careers increasingly they don't want to manage anybody whether it's an
[23:51] agent or a person. They want to be individual contributors. They want to provide value. They want they they want to contribute to the value of the organization. They want to be respected as as highquality professionals, but they don't want management responsibilities. And so it was it was
[24:07] an interesting just just onion to peel back on culture and really look at the way that both the imple the the um the application of AI as well as the way that the workforce is evolving is really changing the landscape of of how people
[24:23] are becoming more literate and where they're where they're applying that knowledge. Yeah. I mean, we've been talking about managing agents within NIPA and okay, are we going to have performance reviews for agents and yeah, and um
[24:38] but yeah, I mean I think the workforce is going to change. You mentioned it like are people going to be re replaced by AI? No. But their jobs are going to change. The way they do their work today is going to change. The way they get information, the tools that they have at their hands is changing. Um, I mean,
[24:55] we're trying to in addition to the AI literacy, we're also looking at different ways that we can um tap into maybe that new to workforce um talent uh so that we we can infuse that into the
[25:11] organization. We have a fellowship that uh we kicked off earlier this year with uh nine students at the University of Buffalo uh PhD students and they're working on four different use cases. It's funded by uh a clean energy
[25:26] workforce development program. Um so we're training them uh you know on AI to help with AI or or help with um climate and and clean energy use cases. Yeah. So, you know, to
[25:42] go back to your point earlier about sustainability and and I mean, NIPA is in a uh unique position because we generate power, but we're trying to use AI to generate power more efficiently and and uh you know, managing agents.
[25:58] Part of what I want to manage in the future is what is our sustainability like what's our footprint? Um, you know, perfect world. I want to have a dashboard on everybody's laptop where you know they're they're using AI and it they have a dial that shows whether it's uh you know good use
[26:16] of of AI or not and and so anyway I don't know how why I went there but no I I and and you know if so if I think through you know we've been talking about some of the challenges we've talked about we've talked about data as a challenge we've talked about you know
[26:32] cultural and change management as a challenge but I think One of the other things that's holding back a lot of companies from really investing in transformation, not just small use cases, not just tools put into the hands of employees to be individual employee productivity tools, but real
[26:48] transformation using AI is is risk and governance. And you know, there's there's there's a a lot of ways to look at this. And you know, I'll put up a kind of a a a wheel of how how we think about it holistically, but you know, governance isn't just one thing. And
[27:04] it's it's not just about, you know, who's watching the models. It's it's not just about who's governing your data. It's it has it has so many different facets to it. I wonder as as you've been working through transformation with AI, you know, what pieces of the governance wheel have have you been focusing on,
[27:21] you know, personally and as an organization, um where where are you where where's the focus coming for you? Everywhere. So um no what I realized is and and I've heard the word governance at this conference a number of times uh and it's
[27:39] often used in terms of technical governance or model governance. Um but what I realized as I started to talk with our executive uh leadership team about governance, it depended on who I was talking to. And
[27:55] so I talked to our CIO, he would think about technical governance. I talk with our um innovation officer or head of our EPMO and they would think about use case intake and prioritization and value
[28:12] realization as governance. um you talk with our our risk organization and they think about uh legislation and uh policy and so we've kind of looked at governance from a you know policy perspective an
[28:29] organizational perspective technology and then oh yeah there's there's cyber and there's data that goes along with that and project delivery um so when I say all of it I I joke but we are actually looking at it from those
[28:45] different lenses. Um because I I think you need to cover all of that. The good thing I've heard a lot at this conference about the technical governance which is great because I don't want to spend time NIPA doesn't
[29:01] want to spend time on the tools um to to govern the technology. Let's let the vendors do that and we'll we'll leverage that. But the governance within our organization uh our policy you know we we created a a generative AI policy um
[29:18] you know that's more specific to NIPA and in our culture and the way the our risk tolerance um so I'd rather focus on that and then just let's leverage the technology providers and integrate that into our our environment. Yeah. And you
[29:35] know I I I actually love your answer that it you know it's everywhere. everything right because I think this is one of those unique things that touches so many different parts of the organization right like you said it's you know the syso is thinking about that
[29:50] that that privacy and security layer and your CIO and co are thinking about how data is being leveraged properly and your you know CFO is thinking about the costs and and growth that can be generated by it and your risk and compliance is thinking about regulation
[30:07] they're also thinking about policy like it really is something that touches the entire organization. And what's interesting too though is that in organizations that do not have their governance structures really well defined with all of these leaders coming together on the same mission with the
[30:25] same strategy. Um you you start to see misalignment. And I I had a I had a interesting scenario just just a few weeks ago. I uh I had a meeting with um the the CIO and the CFO of of an organization. And then in the same vein,
[30:42] I had a meeting with the head of AI for the same organization. And when I talked to the head of AI, what he was focused on is how do I get people using AI more? How do I get it? How do I get them using the tools more? How do I get, you know, how do I get
[30:58] that adoption up? When I talked to the CIO and the CFO, their question was how do I control cost? And it was literally competing against each other, right? It was, you know, one was how do I reduce tokens so that I reduce costs? The other
[31:14] was how do I increase tokens to increase to increase, you know, experience and and adoption. And you know, we when we talk about governance like the answer to that question of where is the balance of those two things that is a governance question, right? But if you don't have
[31:30] these robust governance structures in place, those are some of the things that that that fall to the wayside. Yeah. And one thing I would add, um, so we had a lot of governance in place prior to AI in the chat my chat GPT
[31:46] moment. Um, we're not because of AI recreating new governance where it doesn't need to be recreated. It's like let's leverage what already exists. We have an architecture review board. AI is just another technology, another
[32:02] solution. Let's leverage that architecture review board as it is. Um, we have policies that are in place. Our acceptable use policy. It applies to AI. We've made changes, but we haven't our our guiding principle
[32:19] was let's not create new unless we absolutely have to. Um, our Gen AI policy was a case where we created new because there wasn't really anything that covered generative AI. Um, and so we thought it was so new and so
[32:35] different that we created new. So, um, and I actually think that this is a a some this is a place where there's a very different there's a very different either challenge or um opportunity
[32:50] depending on what industry you're in. Because if you know if I look at organiz some some of the companies that we've been working with that are in regulated industries be it power and utilities be it healthcare and life sciences be it financial services you know heavily
[33:06] regulated industries they are they are tending to have better more robust governance processes that are to your point just the next evolution of processes they already had whereas
[33:22] Whereas when you look at less regulated industries, there's there's a little bit of a, you know, oh crap moment because they don't have the same types of existing governance processes in place. You know, I I like to say like we're we're talking a lot about eval now,
[33:38] right? Like how do I evaluate these agents? How do I make sure they're doing what they're supposed to be doing and they're doing it in a way that is trusted and secure? And I think well in financial services, you know, they've been doing model stress testing for decades, right? They've already got
[33:53] these things set up. And you know, that's very different than if you're, you know, a consumer goods and retail company. You probably haven't had those things because the way that AI has been used in your organization hasn't required that level of rigor. And so I think there's there's a different
[34:09] mountain um or just molehill depending on what industry and sector that you're in. Yeah. Yeah, I came from financial services before uh my time at NIPA and yeah um you're absolutely right there a lot of governance almost too much governance um but probably where we are
[34:27] today that's that's a better thing. So so so um we've talked about a lot of challenges we've talked about um a lot of different elements of um of of what's needed in order to actually capitalize on on transformation. um big world ahead
[34:43] of us, lots of things changing. You know, what are you either most excited about or most scared for? The I mean the opportunity I I've often said at work like AI has just
[34:59] increased the intensity of work because things are moving so quickly. Um so I'm excited about the opportunity. I'm nervous about just the change, the the rapid change, um, and trying to keep up.
[35:14] And that's that's something that we're, you know, the technologies changing rapidly. Um, you know, whether it's state legislation, federal, whatever the case may be, that's changing. Um, and it's just trying to keep up. And so, you know, we leverage partners like EY to to
[35:32] help keep us informed on on what you're seeing in the industry. Um, you know, I'm excited about, you know, many of the things that I heard this week. Um, you know, ontology is is one that, uh, you know, we've created a
[35:49] taxonomy within NIPA. Um, we've used it within our our SharePoint sites um to help with search, but it hasn't really brought a lot of value. I think you know looking every you know conversation I have or
[36:06] conference I go to taxonomy ontology context is more important and you know okay great we have a taxonomy but having technologies that can help us to identify and and we talked earlier about
[36:22] it's never going to replace the human that's you know giving the the business context to say you know customer um you to the marketing department means something different to the financial department. Um
[36:37] but the technology can help to surface some of those questions in those those areas where you need to yeah so I'm excited about that. Um yeah I mean we have a number of use cases like I said we have our our center of excellence we um have stood up fusion
[36:54] teams that are focused on it's it's basically agile teams that are you know it's it's multid-discipline IT our innovation office our business units um that are working on delivering use cases and you know we've kind of changed the
[37:10] way that we work where we used to be a you know waterfall big you know do the require requirements, do your design, you know, 12 months later, here's here's what you get. We're much faster, more much more agile. Um, so it's and we've
[37:26] stood up three teams so far. We're looking to stand up more teams. Um, so excited about that and what we can do. Um, you know, the value that we can deliver to our to our business. Yeah, it there is a lot to be excited about and I
[37:42] think as as as I as I look forward what what I'm most excited about is us moving past use cases and moving into actual endto-end transformation. I think um because you know this the space is moving so fast and there's so much excitement and there's been so much
[37:57] democratization of AI into everybody's hands that it's really easy for us to think about the those small use cases, those individual productivity types of gains that we can get. And I feel like we've been running a little bit on that hamster wheel for for a bit. And it
[38:13] feels to me a little like the 10 or 15 years ago RPA story, right? Where we looked at the existing processes we have and we said, "Hey, look at this big beautiful well- definfined process and I can use uh an automation bot here and
[38:30] here and here and here and here." Right? But, you know, that got us some value, but it wasn't really transformational. And I I think there's too much of that going on today. But if if we can shift our mindset to be where do I start in a
[38:46] process and what's the ultimate goal I want to get to and forget the way we do it today and rewrite it with an AI first approach, you start to get to something that really is transformational because you're removing out those layers of where technology became uh a a a reason
[39:05] to change your process, new talent became a reason to change your process, a regulation became a reason to change your process. process and acquisition became a reason to change your process. And we've built up these processes over time that are these giant spiderw webs. And if we can just wipe the slate clean
[39:20] and go, this is where I start, this is where I want to end, and I can use an AI first approach to get to it, we get out of use cases, and we get to transformation. And I that's that's what I'm most excited for. Yeah. And I see that I I mentioned earlier that the use cases that we've gathered, a lot of those are kind of
[39:36] bottoms up. And that's where we need that top- down view to really drive the strategy and have our leaders be in the right mindset to to think that way. It's not just you're slicing out a little part of a process and automating it. You're just wiping the slate clean.
[39:52] Um and with that, thank you for for for coming to our session today. Enjoy the rest of the conference.

Learn more about the Databricks Data and AI platform.

The information provided herein is for general informational purposes only and may not reflect the most current product capabilities or configurations.