Sponsored by: CEI | The Agentic Enterprise: How AI Engineering Is Rewiring Business Models

Summary

CEI.ai defines the agentic enterprise as an organization where AI is embedded in operational decisions and actions—not just isolated use cases—and argues that the gap between high AI investment and low transformation stems from missing execution infrastructure and insufficient trust and governance.
Sunil Senan presents a three-layer AI engineering transformation framework that moves from models to systems, from systems to enterprises, and from enterprises to new business models driven by closed-loop, outcome-measured AI workflows.
Kevin Larkin of the Federal Home Loan Bank of Des Moines shares a governance-first approach to AI adoption with a goal of 250 production agents by end of 2027, and lessons including the use of AI-judge models and on-premises hardware for regulated workloads.

Chapters

FAQs

What is the agentic enterprise according to CEI.ai?

CEI.ai defines the agentic enterprise as an organization where humans and AI operate together, with AI embedded directly in operational decisions and actions rather than confined to isolated use cases. This model emerges when AI execution is tightly linked to real-world business outcomes and scales across the full enterprise rather than remaining in pilot projects.

Why do most AI initiatives fail to deliver enterprise transformation?

This video argues that enterprises stall for three reasons: an early over-focus on models rather than operationalization, a failure to think about execution of AI intelligence within actual business processes, and an underinvestment in the scale, trust, and confidence infrastructure that impact requires. Addressing all three is what distinguishes transformative agentic enterprises from organizations stuck in perpetual pilots.

What is the Federal Home Loan Bank of Des Moines's approach to AI adoption?

Kevin Larkin describes a governance-first approach at the Federal Home Loan Bank of Des Moines that prioritized reducing operational friction and gaining measurable insight before scaling AI. Key lessons from their deployment include using AI-judge models for evaluating outputs, the use of on-premises hardware for regulated workloads, and a target of 250 production agents by end of 2027.

What are closed-loop AI workflows in agentic transformation?

Closed-loop workflows, as described in this video by CEI.ai, are AI-driven processes where the output of an agent action feeds back into the system to measure outcomes, improve decisions, and adapt behavior over time. This outcome-driven design distinguishes transformative enterprise AI from point solutions that produce insights without connecting to measurable business results.

Full transcript

[00:08] Today, every enterprise is trying to harness AI.
[00:20] how AI is operationalized. Enterprises don't fail because they lack
[00:32] At CI, we engineer an AI operating
[00:45] speed, at scale, and still keep it manageable. Every AI transformation has three
[00:57] layer. Transformation doesn't happen all at once. It happens
[01:09] compounding into an enterprise
[01:23] Where humans and AI operate together. At CI, this is what we call the agentic enterprise. >> Like your new building. >> And this is how the AI era is
[01:36] operationalized. CI delivering the AI future today, responsibly.
[01:50] Good morning all. U thank you for joining uh this session. A few um you know important things. Uh you will hear forward-looking statement from both me as well as my co-speaker. Uh and you
[02:03] will need to treat the information accordingly. Please do fill out the survey uh that comes out. It helps uh all of us improve the content that you experience uh in the summit. Um my name is Sunil Sen. I'm the CEO of CI.AI an
[02:18] engineering an AI engineering and transformation company. Uh and thank you for joining us today. The the entire focus is going to be about how AI engineering is rewiring uh enterprises
[02:30] to create new operating models, new business models. I want to set some uh foundational uh things you know just get them out of the way so that we are thinking uh about the
[02:42] problem in the same way and then we go a little bit deeper in understanding u you know what what happened so AI exuberance as we know is all-time high and we look around and yet we we don't see as many
[02:56] enterprise transformation stories as yet uh and we are here to discuss why that gap and what can be done to to bridge that gap gap as we come forward. So there are three reasons that we think uh
[03:08] you know the the the gap exists. One we in the initial 2 to three years when the geni models and the and then the rest of the wave came in we were kind of stuck in the models and we continuously looked
[03:21] at you know what's the next model etc. But operationalizing it and bringing that uh into the enterprise was limited only to the use cases. Second, we didn't think about execution, right? We only thought about how to bring those use
[03:34] cases but not necessarily look at execution of that intelligence layer in the context of the business processes and operations that we run. And the third impact requires scale and trust and confidence and that's something that
[03:47] the underwiring uh was was not present. So what is the agentic enterprise at CI.ai? We define it as a as an enterprise that has AI embedded into the
[03:59] workflows. It's not in the side view mirror that you're looking at your AI model as an add-on. It has the system that has the ability to decide and act. We have lots of tools out there that can
[04:13] do observe and sense and recommend. But decide and act is the material difference between that and what we are looking at from the agentic enterprise perspective and that it has the business
[04:26] outcome focus and not necessarily just the insights. Insights is a part and it's important for research organizations and so many others but ability to take it to the business outcomes that you're looking at becomes very very important. So how do you think
[04:39] about from models to business models? Think of models as capabilities which you have available and there is more coming every day. We will have newer and newer capabilities. Some you will create
[04:51] as well. But when you take it to execution, it becomes a system. But you're still executing for a domain for a project uh and things like that. But when you take it to enterprise level, you will often uh times think about what
[05:05] is my operating model now. Interdep departmental initiatives. AI is going to create a very fluid O structure. Things are going to converge and collage more. So what's your operating model? Where is
[05:17] your accountability? Where's where's the ownership? Where the execution going to be? And finally leads to your products and services that'll change the business model that you operate. AI is a connected product. It creates extreme
[05:30] connectivity with all stakeholders, customers, employees, partners, other stakeholders that you are interacting with. And this is the journey that you want to take in order to get to the outcomes. Not necessarily overnight but
[05:42] building up you know step by step brick by brick getting to that particular outcome. Leveraging the right foundation we believe data bricks provides that is extremely important. But if you don't
[05:54] have the execution layer on top you're often times going to see things like you can detect the fraud but you can't act on it. There are signals for care that needs to be done but they are not operationalizing the healthcare uh you
[06:07] know signals or you can see that there is an opportunity for you to price the product differently but you can't and it takes manual effort you know in order to in order to drive that and often the lack of execution that creates this
[06:19] uneven impact as we as we look at the industry data bricks provides a unified data and AI foundation it gives you governance by design and it brings production ready AI at scale but That's
[06:31] the foundation. So you have those foundations available for you to build your your enterprises on top. We believe that you need two things to work together to fill this gap. There is AI
[06:44] engineering which is how you wire your engineering part of it which covers the software engineering and everything that happens around that and how do you tie this to what we call the system of work.
[06:56] We have system of records. We built system of engagements. You need system of work right that can take decisions act and actually operationalize as as things are needed. AI engineering I want
[07:08] to spend a little bit more time u you know just giving you a little bit of flavor of these two because that'll be extremely important to understand AI engineering most of the organizations are dealing with legacy debt and hence
[07:22] modernization is a one big you know piece of your engineering looking at how you do green field development there are organizations that have decided not to upgrade the next SAS version and build
[07:34] it out new but how do you build is in the in the in the context of your organization with the specificity of your domain etc. And the final one is the brownfield enhancements by far the most complex engineering process that
[07:46] you would see and how do you operationalize that using AI and that's the platform that CI.AI AI uh has launched in the market called Darts completely powered by data bricks as well as other platforms on the on the
[07:59] business intelligence uh side you need to bring these AI models in the center of your business operations and unlock uh you know all the AI capabilities in the context of your process right so how do you bring your process playbook and
[08:12] bring intelligence right in the center create those persona agents that map to your O structure and your personas that operate on those processes and feed it your KPIs so that it knows what is it that each person is trying to achieve.
[08:25] Are you trying to double your bookings? Are you trying to you know improve your supply chain and by you know doing what? So the system learns and and it is completely anchored in your enterprise context and the ontology and it starts
[08:38] to unlock. Combine these two you get the underwiring that you need for being an agentic enterprise and start to unlock value in in business. And we'll talk a little bit more about this as well.
[08:50] There are some examples. If you're in financial services, how you do your risk monitoring and intelligence and act on those risks every day. It's not a quarterly exercise anymore. How do you provide wealth advisory to your
[09:02] customers completely changes. It's hyperpersonalized and yet scalable for servicing all your customers. If you're in healthcare, a care agent that continuously monitors these signals coming from the patients and looking at
[09:14] how what are the other interventions and treatments that need to be enabled and actually enable those as as we go along. If you are in retail CPG, you're dealing with demand forecasting as probably the problem. Pricing, how do you price your
[09:26] products? There are so many customers who have spoken ush to us about how pricing has become very very difficult in the current world with all the macroeconomic changes around uh you know us uh across countries uh and how do you
[09:39] gauge the elasticity of products every day and there is competition that is uh you know kind of bringing many such options if you're in agri business you're dealing with the yield intelligence how do I forecast what kind
[09:52] of crop am I going to have this season and then be able to take your forward calls your trade calls and managing your inventory so that you can feed the world better. You can provide the the the the
[10:04] commodity and the food where it is required the most, right? And and look at world supply to be able to meet your world demand. So there are three or four best practices that I wanted to uh quickly
[10:16] touch on um and then go into our discussion. First, as you start this agentic journey, it is very important to understand that the agent decision making is observe, reason, act, and
[10:31] verify. And if you start with closed loop workflows, you will be able to iterate and fine-tune your AI systems to get the kind of out outcome that you need. So, lesson learned is starting
[10:43] with closed loop workflows. Fraud is a good use case in in in financial services. You exactly know what are the patterns that humans look for. Is the system detecting that and more and how
[10:56] do you adjudicate some of those you know cases as we as we go along as well. Second, you have to combine the outcomes with engineering. We see plethora of proof of concepts and
[11:11] uh initiatives taken uh across the organizations. There are enough players out there that will test the technology for you. But if you have an outcome and you're designing for that outcome, more often than not, these pilots will see
[11:24] production production scale rollouts. We think of it from a CI perspective that you start with business and you look at what are the outcomes that you're driving and coming backwards to see what
[11:37] kind of agentic system you need, what kind of data do you need and do you have them in place and how do you build the context and the ontology that is required around it so that you can confidently you know spin up those agents that will take up the work and
[11:50] start to become the assistant number one and then you know look at evolution from there. that has far more success rate compared to far higher success rate compared to all the PC's that don't
[12:02] necessarily have these you know objectives combined to that. So even when you're doing pilot there should be an objective attached to that and then you will see that that reaches you know production scale. Third best practice AI
[12:15] transformation is different. It is it is not your 18 to 24 month mega transformation initiative. It delivers big impacts but it comes in small increments. It comes every day. It comes
[12:28] every week. So the micro changes that makes it easier for humans to also you know absorb this embrace it and then see how humans can improve and how agentic systems can improve together. On day one
[12:41] you will have the trade the trust deficit between humans and AI and that's just natural. But giving this micro changes every day makes both the systems work better and they they improve over a
[12:53] period of time as well. So these micro changes lead to macro transformations and that's the AI transformation that that you're going to experience. Finally, it's how you put AI system to
[13:07] work but you also can have ownership and accountability to it. In the end state, you're going to probably deal with a fluid structure because agents can transcend your departmental boundaries,
[13:19] your application boundaries, your data boundaries, oftent times even your enterprise boundaries. And it'll be difficult to say who owns it and who is accountable for the decisions made by agents. And that's the structure that we're going to go to. But it's extremely
[13:31] important to understand how we embrace this today. Most organizations we have recommended that we start with what is called the persona agent which is to amplify your existing personas. They
[13:44] belong in your rock structure. There is a responsibility. There is a KPI and there is a governance model that is around that and that's a good way to get agentic in because it participate in that structure with your governance around it. And of course, agent needs
[13:57] certain other governance layer to be able to see that it's making the right decision and so on, right? Ethical decisions, safety, IP considerations and number of other things. But it works in your governance structure. But as the
[14:09] system evolves, it then looks at how do you transcend that from persona amplification to process reimagination. How can you simplify those workflows? Just because something was done in sevenstep workflow yesterday doesn't
[14:21] mean that we do that today. The system needs to be able to identify those opportunities, those bottlenecks, those barriers to value and how you improve from there. And finally getting that
[14:33] into changing the value delivery. What do you get from those business processes? These are all value creating processes that you have nurtured, implemented, standardized over a period of time. The ERP era promoted
[14:45] standardization. So we put everything on one template. Agent is reverse. You personalize it and yet make it manageable. And that's the cycle that we're going to see amplify roles do
[14:57] business process reimagination but with your agent tech personas in place and that becomes your transformation backbone and changing the value delivery over a period of time. Finally, experimentation to enterprise level
[15:10] impact requires a a thought process requires an approach and a framework that can in that you need to engineer in the organization. We often times see that there are lot of experimentations
[15:23] that's done by both IT and business. It's because the AI access is democratized but it's for the enterprises leaders us to think about how you harness that and I think of it always as two engines. one engine that
[15:37] is working innovation at speed. You want to enable speed for our teams so that they can experiment but all on your you know quote unquote playground. So there are some boundaries in which those experiments are happening and every experiment has an objective so you can
[15:50] then select which ones go into the next engine which is innovation at scale and you need to create those funnel so that you're proving out the ideas but you're very quickly scaling it. And if you have those two teams and that framework
[16:03] within the organization, you will see that most of these move into production. At this point, I would like to invite one of our esteemed uh clients. I'm very proud to have Kevin Larkin, the chief
[16:15] information and operations officer of FHLB and in De Moine. Um, and I'm going to, you know, provide a brief introduction to to this, but there are three things that I can say uh, you know, about Federal Home Loan Bank. It
[16:29] is the second largest issuer in the world after US treasury. Second, they act as the shock absorber for the economy. When economy is, you
[16:41] know, wobbly, it is facing challenges, headwinds. Federal Home Loan Bank works as a cushion to make sure that the societies are running. And third, AI for good. I haven't seen a better example.
[16:53] And it's my immense pleasure to invite Kevin Larkin to share that story with all of you. >>
[17:06] have undertaken at FHLB. So first of all, congratulations for being that role model for being that leader who's pushing the envelope and and looking at what are the new things that we got to learn. Um you want to give a brief
[17:18] introduction to yourself and to Federal Home Loan Bank for for the audience here? >> Certainly, but I'm not going to repeat what was just said. You said that very well. We uh the Federal Home Loan Bank is an organization you may have never heard of. Uh and that's somewhat
[17:31] intentional. We support the uh financial system as a whole in the United States. Uh like was just said, we are the largest second largest issuer of debt in the world. Why do we issue so much debt? We take that those the bonds that we
[17:44] issue. They take that money, we uh lend it into the financial system, your credit unions, your banks, um community investments, and we give 10 to 20% of our earnings into the affordable housing
[17:57] arena. So, uh just I think everyone has a story where it's tough to get that first house or even that second house. uh we help with a number of different programs from down payments to u building affordable housing for people
[18:10] who are in on the lower income side. So that's that's essentially what the home loan banks do. We do it quietly. Um but we're super excited about Agentic AI. It's going to help us do do more good.
[18:24] >> That's awesome. It couldn't have been put uh better than that. And we are so proud to be associated with you uh and the journey that you've taken up at FHLB. immensely proud. Um let's get into
[18:36] how did you think about how you set the objectives right because the art of the possible is so far so vast and so wide. How did you think about what are the objectives that you want to achieve and
[18:48] how can this be a gamecher for Federal Home Loan Bank? >> Certainly uh I' I've said it a couple times. It's the golden age to be an information technology right now. We've never had these tools at our disposal
[19:00] before. Uh and now that we have them, okay, what can we do? Um nothing's free and cheap or easy if it's going to be good. The way we think about our objectives, we want to reduce the friction uh within our operations.
[19:12] That's certainly an easy one. Uh we want to get insights into our business model uh insights into patterns that we never saw before. We know we can get those. And as we build out workflows, okay, let's make these workflows even more
[19:24] efficient. Uh so today, that's the way we're going to measure it. Uh at the same time I jumping into the ROI space. So all of a sudden when I get asked, hey, how are we gonna what's the ROI on that investment? There's extent of uh
[19:38] we'll say sunk costs. If we don't invest now, we can't get it later. This is that step one. We're going to invest now. >> Yeah. And those are great examples of the closed loop workflows that we were talking about. So you know where it is
[19:51] going in, what it is going to deliver, but it's also a change management process, right? to be able to, you know, start doing those small things that lead to that bigger impact uh as as it goes along. That's that's well said. Uh
[20:04] Kevin, what initiatives have you undertaken given, you know, these objectives within within Federal Home Loan Bank and and how do you think about what objectives you'll you are seeking
[20:16] to get to, let's say, in next 12 months? >> Certainly. So we're a financial services organization and not only a financial services we're as I described we kind of underpin the economy to an extent. So
[20:28] our regulator is let's just say very interested in what we do and how we do it. Um because of that our step one was let's get a governance model in place before we do anything specific in AI
[20:40] even a simple chatbot let's bring our risk department let's bring our audit department let's get everybody uh together understand and created a governance group say okay you make the decision on what we do with AI because once they're comfortable with it we know
[20:52] we can continue forward that that helps step one uh the next thing uh when I think about what we've done to prepare um didn't know it at the time but now can leverage I I'll use the phrase identity and access management. So every
[21:07] individual in the organization has access to certain things. They have entitlements as as we call it. So personas is another way to think about that because we've invested there. We because we have to best practice those
[21:19] best practices can now be used for agentic AI. My expectation is for every agent we're going to be in our same identity and access management systems. Um, so my my Kevin agent is going to be,
[21:32] you know, managed just like an individual if they were a human. Uh, and then lastly, policies, procedures, and, uh, clean data. All were done for best practices because we have to now we can
[21:45] really take advantage of that, have those agents read those procedures and know how to execute a workflow and, uh, we don't have to retrain them. >> That's phenomenal. Uh Kevin, I'm going to go off script a little bit to uh talk
[21:58] a little bit about you know some of the profound changes that I experienced in in in our journey as well as in conversation with you. Uh you know in the industry we always think about build versus buy in certain ways and it's very
[22:12] very interesting uh to hear how you have thought about it. You want to share that thought process with everybody here today? >> Certainly. Um like many organizations our strategy had been when it comes to applications development uh solutions we
[22:26] want to uh buy first SAS would be our preferred uh approach and then build only if we have to only if we have to agentic AI and just AI in general we
[22:40] flip on our strategy. We are now thinking I think I can build that thing first. I can build it cheaper. I can build it explicitly for what I wanted to do. I can embed AI where I wanted to embed it. Before could not do that. I
[22:53] was at the mercy of the vendor. The vendor gives us what we've got, mess with it, hope for the best, complain about it daily. Uh so now our strategy is is is flipped because of AI. The
[23:07] engineering economics have led to different choices. That sounded so unnatural just yesterday. Yes,
[23:19] an enterprise scale the scale that you are at that you are able to orchestrate strategies in keeping with the with the evolution of AI as well. That's that's a phenomenal story. The other paradigm that we briefly spoke about is that of
[23:33] this knowledge management >> and how that plays out in the enterprise with people and all the you know human cycles that we see. uh why don't you share that uh you know thought process?
[23:45] >> Yes. So knowledge management um a challenge every organization has. I got a couple individuals who've been with the organization 40 45 years. They're about to retire. Oh no, what about all that knowledge? The good news is we've got policies, procedures, guidelines. If
[24:00] I could teach an agent those policies and procedures, maybe just maybe I can replace some of that knowledge that just walked out the door. Um, and I don't have to reply rely on a service now, you know, knowledge base or
[24:14] something else that only gets me so far because a human has to go find it and search it and they still say they can't find it. So excited that an AI agent can help with that, I'll call it key person
[24:27] risk and that knowledge uh knowledge loss that happens when we have these long tenur or individuals leave the organization. >> Yeah. And and we see this with our customers that customers that invested in knowledge management today are
[24:40] sitting on a golden pile of information that is well curated governed over a period of time have become the playbooks for the organization for years like you said policies procedures and and the
[24:52] risks etc and your journey to agentic is only as good as the knowledge that you can feed it. So I think it's it's now finding a lot of relevance in the in the world that uh we live in. Um you want to
[25:05] share any of your best practices, lesson learned, things that don't get talked about in the consulting world but are real practical insights. >> Certainly uh lesson learned and some of these might be obvious. So I I talked
[25:17] about governance, governance first that that really helped us go faster. At first when we started that, as you can imagine, the uh technology and business folks started rolling their eyes when we said, "Okay, we're going to create another governance group that's going to
[25:29] watch this. Oh, it's going to take forever." But what's played out is once that was in place, now the uh use cases are getting approved much faster. They get implemented much faster. People are getting much more comfortable. That was
[25:41] one lesson learned. Um, uh, I'd say the other one was, uh, you've got to be able to challenge your AI solution in order to feel good that it didn't hallucinate. And this one's got most people nervous.
[25:55] So, uh, we're going to have a judge model, if you will, evaluate the uh, the large language model that came up with the solution. So, AI to check AI. Uh, and maybe lastly, uh, and maybe time
[26:07] will tell if this was a good lesson learned or not. uh last year got really concerned about this the token tokenization in the cost model for AI. So we went out and bought our own uh hardware. So we've got a a public I'll
[26:19] call public cloud AI solution. I got a private cloud AI solution on the private uh I've got uh some hardware. It's not the beefiest but it's good enough uh redundancy and that's where we're hoping to to manage the costs and sensitive
[26:33] data over on that side. So that that that should be interesting. engaging great partners um is also another lesson learned. Uh although you got anxious people that are doing the stuff at home, that doesn't mean you you've got enough
[26:46] experience. Great. But get a good partner. Uh and maybe uh lastly uh when I think about lessons learned was try to leverage as many tools and technologies that you already have because we're busy
[27:00] as it is. This is kind of new unplanned work. It's exciting. It's great. But we already have priorities that go past the list of of what you can do in a given year. So instead of ripping everything up and starting from scratch and saying,
[27:12] "Okay, we're going to bring in data bricks and it's going to do everything." Uh, okay, maybe let's leverage what we've got for most of what we need, I can build agents with current tool sets, but I still need some extra uh AI solution. So that all that last lesson
[27:25] learned was don't uh don't start from scratch from a technologies perspective because that's even more expensive and more disruption. It's going to really slow us down even more. such uh powerful insights I think coming from real practical experience of trying to bring
[27:39] that transformation that we're looking for within the enterprise but I think what really stands out to me is the responsible way in which you're approaching it and that was the first point that you've shared I think FHLB being a responsible organization I think
[27:53] it comes naturally to you to see you know how we anchor this so that it's not not just not hallucinating but also creating safe envir environment and uh you know the the kind of effective
[28:05] outcome that we are that we are all looking for. It kind of leads me to the next question that I have for you. AI and and humans are going to be interoperable but often times we treat
[28:17] them as competing forces right and doing this at an enterprise scale even if you're starting with certain divisions first etc. I think you reached that crossroad of how do I run the change
[28:30] management process? How do I get my people excited about doing things that they were doing that agent is going to do? You want to share your thought process around that? >> Certainly. So, I'll go with that last piece. How do we get people excited about this uh using AI? So, we we've had
[28:45] uh yeah, this month two different promptathons as we called it. So, uh we get people together like hackathons for the developers who might be in the audience. So, get people uh in a room and we give them a challenge. Okay.
[28:57] Who's got the best prompts? Create some teams. Create some energy. Uh we had prizes for who had the best one. Created posters. Use AI to create the best poster for this organization. What do you think? So that was fun. Competition seems to bring out something us humans.
[29:10] So we we we chase that. So that was helpful. Promptathon to get people excited. Um second thing is our philosophy is we want the uh the agents or the agentic AI to do the the hard
[29:23] work, the toil if you will, and we want our humans to make the decisions. So the agent says this is what we suggest you do. The human says I agree or don't agree. So that's that's going to be our approach. Um and lastly, my vision is I
[29:37] want my end users configuring agents. I want those individuals saying, "Yeah, this is my agent. I know what it does. I know why it does that." And then that'll just that'll start to spread when somebody says, "Oh, you got you have an agent. What do you mean you have an
[29:49] agent?" Uh so that's uh the last piece where we we expect that to play out. >> Yeah. No, that's phenomenal. And and I think the step that you've taken to say most of the business operations or the
[30:01] core processes by end of next year must you know experience the agentic transformation I think gets people thinking and and they start working towards their objective right and and that's what we were earlier talking
[30:13] about that when you are taking small steps you're taking with the objective in mind uh otherwise those small step will remain in the labs or in the PC stage right so as you think about the enterprise transformation uh at scale,
[30:28] right? Uh how have you thought about those, you know, one year out end objectives that we want to get to and then break down into your into your smaller initiatives? >> Certainly. Certainly. So, uh when we try
[30:40] to think about what we want to measure, it's a little bit tough at the very beginning because you're not sure exactly how it's going to play out. So, one thing that I uh challenged my team earlier this year, it was I want two at least 250 agents in production doing
[30:55] production workflows by the end of 2027. And you can imagine the look they gave me. And they said, "Well, what what should they do?" I'm like, "I don't know. That's up to you. You're going to figure that out. But that's that's your goal. I want I want I want agents in
[31:08] production. You figure out which best ones to do. Some are going to be low risk, some are going to be high risk, some are going to be simple, some be closed loop, some might some are going to fail, they're going to try it, they're going to fail. Some are going to try to game this idea and go, "Oh, okay.
[31:20] I'm going to have one agent then talks to five other agents." And most of fine, that's fine. Let's do it. Let's get those in there. Uh so that's one way uh to to get our agentic uh AI enterprise going.
[31:32] >> Um I also think about uh how we're going to measure uh the outcomes. How many processes are agentic? How many people um are no longer doing certain things?
[31:44] You inspired an idea of compliance. As you can imagine, this organization that's financially regulated, lots of compliance. I'm sure many of you in organizations have a a compliance piece. Most people don't love it except for
[31:57] maybe uh you know, the compliance team and the internal audits. The rest the the the people have to execute on compliance and double check that you're following your processes and procedures. No fun. perfect use for Ajentech AI. So
[32:10] I'm now thinking about I want my whole compliance department to be the AI compliance department. So thank you for that inspiration. >> Yeah. No, absolutely. I I think you know it comes through a lot of uh um information and idea exchange and
[32:22] thinking about you know what we are seeing in the industry. Um really inspiring to see you know the the steps that you're taking bringing the future today but doing it responsibly. Um the
[32:34] the other element that you know I think comes out very clearly in what you just narrated uh and I want to paraphrase this a bit as well is how you're leading with experimentations that have the end objective of getting to where you want
[32:46] to be at the end of 2027 but you're also thinking processes and you're also thinking about how you scale these ideas once the experimentations happen and that's the piece that I was talking about earlier with innovation at speed
[32:58] and innovation at scale engines and they are different they can be on the same uh bedrock of innovative platforms that that you have which are data and AI platforms but they are different
[33:11] objectives you want to be able to experiment fast fail fast like you said the ideas that don't work we need to really get them out of the system very very quickly but you need to discover those that work as well and it comes
[33:23] with that experimentation and then there is an art of scaling you know as we as we think about uh you know bringing this up so thank you so much for sharing the insights it's been phenomenal uh you know uh and very inspiring story
[33:35] to hear how FHLB which is a very key part of our economy not just US but also world economy and how you are working to make sure that everybody's lives are impacted positively and thank you for
[33:48] bringing AI as a as a mechanism to be able to do reallocation of capital and and all the you know uh being the shock absorber you know for the economy and and really putting AI to AI to a good
[34:01] use. So I love it when you said let's do more good >> with >> AI for good. >> AI for good. And that is the story. Thank you so much. We do have some time for any questions uh from the audience.

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.