Sponsored by: WisdomAI | Your AI Doesn't Know Your Business. Your Data Analysts Do.

Summary

Many AI analytics initiatives stall before production because they provide technically correct but operationally wrong answers—a problem Wisdom AI attributes to missing enterprise context across four layers: semantic models, business operations knowledge, governance rules, and decision traces.
Wisdom AI's adaptive context engine is an agentic analytics platform purpose-built for the enterprise that equips both human users and AI agents with high-quality, trustworthy, and explainable results by encoding all four context layers rather than relying on semantic models alone.
Geetesh Iyer introduces the AI Context Engineer as an evolution of the data analyst role—a practitioner who curates and maintains the enterprise context layer that makes AI analytics reliable enough for casual users and proactive AI agents at scale.

Chapters

FAQs

Why do AI analytics pilots fail to reach production?

According to this video, AI analytics pilots often fail because they lack the business context needed to produce operationally correct answers—not just technically correct ones. When the user base expands beyond a small proof-of-concept group to hundreds of casual users, the absence of semantic models, business rules, governance policies, and historical decision context causes conflicting or misleading results.

What are the four layers of context AI analytics needs?

Geetesh Iyer from Wisdom AI outlines four context layers: semantic models that define business metrics, business operations context capturing how the organization actually works, governance rules controlling who sees what, and decision traces recording past analytical conclusions. AI systems that lack any of these layers produce answers that are technically correct but operationally wrong.

What is the AI Context Engineer role?

The AI Context Engineer is a new role introduced in this video as an evolution of the data analyst, responsible for curating and maintaining the enterprise context layer that AI analytics systems rely on. This role ensures that semantic models, business logic, and governance rules are accurate and up to date so both human users and AI agents receive trustworthy answers.

What is Wisdom AI's adaptive context engine?

Wisdom AI's adaptive context engine is a purpose-built component of their agentic analytics platform that encodes all four layers of enterprise context—semantic, business operations, governance, and decision traces—and makes that context available to AI models at query time. It is designed to work on top of data stored in platforms like Databricks, enabling proactive AI agents that produce accurate, explainable results at scale.

Full transcript

[00:08] Good morning everybody. I said good morning. There you go. I know it's the last day of the conference. We've all been to the party last night or some of us have. Some of us have rested. So, I'll
[00:21] keep it appropriate to that. Um because it's the last day and I'm a huge consumer of social media. I absolutely loved that video and that is for three reasons. Number one,
[00:34] it's a spectacular video. It's damn funny. It makes me laugh every time I see this. Second, I'm a musician and I see that this is absolutely real. Uh there is a technology called
[00:47] autotune, a technology called Melodyne where you could it could enhance your vocal that you see in the output. So, yes, this is real. This happens in real life and I have used it on my voice as well
[01:00] on some occasions. And the third one, probably the most pertinent one to this conversation is that it's it talks to a personality trait or not a personality trait, it talks to something that we all humans are very close to. We all want to
[01:12] shoot for more than what we have. We have the vocal ability which is probably a four on 10, I guess. But then we want to make it look like 10. That's the world we live in and
[01:25] particularly for this kind of an audience, you know, a lot of data practitioners here, this is very real. Is it enough? Is a conversation is a question that all of you receive. Right? Uh we have compute for 20 people, but
[01:38] hey, you know what? We are going to have 100 concurrent users. How are you going to manage that? Is it enough? So, same goes with, you know, what kind of LLMs we use. We are going to use a simple model. No, we need to do complex reasoning. We want to do more with less.
[01:51] So, is it enough is something very very very common to particularly the data practitioners and that's what I want to spend our time today doing. Wait, I think I need to figure out how to go to the next one. There you go. So, anyways,
[02:03] with that preface, I want to introduce myself. My name is Gitesh Iyer. I'm the founding product manager at Wisdom AI. Now, Wisdom AI is an agentic analytics platform. It's built for the
[02:15] current and for the enterprise. Our adaptive context engine is purpose-built for analytics in the age of AI. Now, we've built this for two primary personas. Number one, the humans
[02:28] and the agents who are the end explorers. So, for example, a chief marketing officer, a head of sales, a finance manager who wants to ask questions on their data, they need to get extremely high-quality
[02:42] trustworthy, explainable results. That's number one. And number two is we want to address the data teams. We want to build the right tooling and the right engines for the data teams so that they can
[02:54] enable number one. That's what Wisdom AI does. And today we're going to speak a lot about some real-life use cases. And it's going to be very interactive. So, I'd like you guys to participate as much as you want. It's the last day, so I'm
[03:07] going to keep it this way. Sorry, are we ready? Are we ready? It's going to be a lot of music. No, I'm kidding. Anyways, what are we going to talk about? Now, the ChatGPT moment happened I think 3
[03:20] and 1/2 years ago, right? ChatGPT came in and we looked at the possibility that this technology gives us. It told us that you can have your data, you can have your systems, and then you can connect all of these things together to
[03:34] an AI ecosystem, and it'll just give you the answers. It's been 3 and 1/2 years. We're still here in June of 2026 talking about how to make data work. It's worked fantastically in some industries, but
[03:47] for data it's significantly harder. And what it takes to get from where we were 4 years ago to an AI-based system that does analytics is what we're going to be talking about
[03:59] through very real world examples. And I want you guys to be part of it. Sound okay? Let's go. Now, raise your hands if Okay, how many of you have been part of a POC that worked spectacularly well
[04:13] in an AI system? You bought a new tool, you built it internally, but you built a POC for agentic analytics and it worked brilliantly. Show of hands. I mean, it worked for me. All right.
[04:26] That's a significant part. I think half of you guys say that. Now, keep that hand raised if the same quality that you got in that POC extended to production. Where are those hands?
[04:38] Like not not one? Okay. Now, that's not funny enough. Imagine a world now instead of those 10 people who did the POC, now you have 500 people who are casual users. The 10
[04:50] people who did the POC were power users. Now, you have casual users who are coming in. The questions get weirder. Uh the edge cases get hit every single time or more frequently. And the AI really starts sweating. Right? And is this
[05:05] sounding familiar? The moment this happens, trust becomes a question. Everyone's like question, "Hey, is this something we should do? Should we hold back?" And wait, that's not enough. The kicker is when a VP of sales is
[05:17] essentially looking at this and saying that hey, I'm going to report this to the industry, to the market, and then finds out that the data is wrong. A head of finances actually going to put it in the earnings report and they find out the data is wrong with everything
[05:30] that we've set up. What's going to happen? A bunch of you folks are going to get a Slack message at 11:00 p.m. on a Friday night. Deal with it. This is a very common occurrence in enterprise ecosystems. I
[05:42] mean, all of us have been through this. Uh but then there needs to be something that needs to be done for us to go from where we were a few years ago to where we want to be. Let's take a very real world scenario.
[05:55] How many people over here use any of these agentic harnesses like Claude, chat GPT, Gemini? Raise your hands. Everyone, literally everybody. So imagine you folks are the data leaders.
[06:07] How many people who work in data here? Raise your hands, please. Okay, again, most of the class. Sorry, I'm making it look like a class here. Uh but yeah, imagine a world where your company has given access to Claude
[06:19] for everybody, right? Everyone's excited. Hey, you know what? I'm actually going to connect to the database. I'm going to ask questions and the data teams who are typically the people who are getting the insights now are not the ones who are the
[06:31] blockers. The blockers. Uh and everyone has access to this data. So there is Adam from sales who's decided that he's actually going to go and build something with Claude uh or some agentic harness here. He
[06:43] feels the output the chart is amazing. It is exactly the way the VP wants. It's got everything that you need until the VP looks at it and says, "Hey, are you sure that number is right?"
[06:56] This is something we've all dealt with. This is the nightmare for the data teams. And if anyone has worked at Amazon, that question mark is symbolic of what Jeff Bezos apparently I mean it's it was a legend there that he would just send out a question mark
[07:08] uh to ask if what they were saying was actually sensible. Anyways, so this question is asked of you. The first thing that that leader does is it looks at Adam. He's like, "Hey Adam, you've built this. Do you think it's accurate? How do you think it's
[07:20] accurate?" And Adam's like, "Okay, I had access to these tools. I connected a bunch of things. Uh I had access to the MCP server for that data source. I asked a bunch of things. Yes, I have some skills built
[07:32] and I got this response." Uh and Adam obviously doesn't have an answer. Adam looks at you, the data teams. Now the data teams over here have no idea what all has gone into building this. He's actually connected a bunch of
[07:44] things that are local to him and he's actually gone ahead and built this. This is something that happens on a regular basis in enterprise ecosystems. The first thing, let's act as analysts
[07:56] here, right? So, let's try to break this down. If you were in this scenario, what is the first thing you would do? The first aspect is like, hey, you know what? The model is hallucinating. This is something we heard a lot of you years ago. So, we go, we change the model, let's say it's a GPT-5, we change it to
[08:09] Claude, we say excellent, we've changed models, we've tried a bunch of things, still the same answer. We see that uh model is actually doing the right thing. It is doing exactly what it is asked to do.
[08:21] If it is doing exactly what is asked to do, then let's look at what is being fed into it, the data. So, the data that is being sent into it is something that we test next. So, when we look at what's happening, we look at the data modeling practices there. We
[08:34] look at, hey, we provide the right kind of join conditions, we provide the right kind of column descriptions, stable descriptions. We remove a bunch of we hide a bunch of columns to save cost. We do a bunch of these things so that we feel that this is now ready to be sent
[08:47] to an AI system. Then we realize that also is not working. So, TLDR, we've reached a world where the model in many cases is not the problem. The data is also not the problem. Then what is the problem? Hey,
[09:01] wait. Another team actually told me that this is a behavior they are seeing at their end as well. They've had multiple such scenarios where there's two sub-teams, so a sales finance and a sales team, both of them
[09:13] are asking for ARR and both of them are getting different results. Is it always wrong? Not necessarily. ARR means slightly different things in finance than sales, but we have documented it as ARR. So, when an LLM looks at it, it basically
[09:26] makes up its mind. Same goes with having different results on different days. I might have uh the third case is I might have like 10 skills that do something. Ravi might have only a couple of skills that are
[09:38] set up. And then both of us are asking the exact same question on the exact same database even if the sequel is exactly the same, the filters are off. And that happens and you get different results. So, long story short, we live in a world where
[09:51] there are so many things that we have to keep doing and just then someone comes and says, "Guys, you're missing the S word." The semantic model. Great. Let's go and build the semantic
[10:03] model. Now, we are thinking like an analyst here. We say that let's go and build a semantic model here. Uh and when we build a semantic model, everything is great. You've set up the dictionary, you have put in the artifacts, which has all the information
[10:16] that you need about the data and it tells you at what time you should use pick ARR for sales versus ARR for finance. It tells you everything. But then still you when the AI says use a particular table
[10:29] from CRM, the finance team doesn't trust that data. Uh revenue dropped in this, but hey, the reason why revenue dropped was not because something changed, it's just that we've changed the definition. Again, semantic definition is fantastic.
[10:42] It's perfectly done, but still you're getting a response that's not accurate. So, is this enough? I mean, just like we saw in that video, is this enough? We set up the model.
[10:54] It didn't work. We tested We changed the data. We modeled the data. We worked on it. It's still not enough. Then we set up the semantic model, which is very, very important for certain use cases. Uh semantic models were built for a
[11:06] particular uh you know, the dashboards use cases of the previous years. Uh and we set up everything and you're still missing that one particular piece. That's not because of which you're not
[11:18] able to make that leap. And what is that piece? We need something that is the living record of an organization's decision traces. Uh like in this case, it needs to capture why a business why a business
[11:30] decision was made, more importantly. So, for example, it's the end of the quarter, your sales leader comes and tells you, "Hey, you know what? Give that guy give that company a 15% discount and not a 4% discount. I need to close my quarter on a positive."
[11:43] Where does that piece of information getting stored? Uh there is an analyst work around. The analyst has realized that, "Hey, you know what? The tables that we have set up, the aggregate tables, they are far better than the transactional tables that are generally used to report to the
[11:55] leaders on a Friday evening. These are things that are not documentable, if that's a word. These are things that exist in people's heads. They are there in some documents here and there. And what do we call them?"
[12:07] Enterprise context. Now, the reason why I took some time to get to the word enterprise context is there's a lot of literature that's out there in the industry that talks about semantic models, that talks about context layers.
[12:19] But, the rationalization of each of them happens for because a certain problem comes up. Semantic layers, data, model, all of them solved a certain purpose. They solved something.
[12:32] But, when it comes to solving analytics and AI, you need this one initially amorphous concept and now fairly defined concept of an enterprise context for you to understand what it takes to go there. And it's one of the components of that.
[12:44] And what is enterprise context? It has four major components, I would say. The first one is a semantic context. We spoke about it. Semantic models capture a lot of those things. Metrics definitions, ontologies, rules, and
[12:57] a lot of things. And many of you have spent a lot of time in building this over the years. So, a lot of these things exist. So, we can't expect teams to essentially start from scratch when it comes to context. No, we are going to leverage a lot of that. The second
[13:10] thing, like business and operations context, are the ones that I was talking about. Those things that those hallway conversations that become part of your deal. >>
[13:23] the box, but we need to spend some time in getting it out of the heads of some people who know this context. And the last one is obviously governance, which is access controls, auditability, role level security, versioning, and whatnot.
[13:36] So, is this enough? Now, we went from data, we went from data to model, model to semantic layer, semantic layer to now we've realized it's enterprise context. Do you think it's enough?
[13:49] Maybe not. In order to make this come to life, we need that toolkit, that engine that actually makes it come alive in the world of AI. Now, as I said, some people
[14:02] some teams have invested excellent amount of time in building their semantic layer. Some people have not done that. Uh some people have data still in their CSVs or rudimentary databases. Now, we need a way where you can bootstrap from
[14:14] literally any source that you have. And once you've bootstrapped, you want to refine and adapt, give them uh give the context a lot of love, if you will. Uh tell what works, what doesn't work, and then finally have a
[14:28] context development life cycle kind of a setup where you are able to govern, maintain, and publish that context so that it doesn't sit in that system. It is available for other systems for you to consume.
[14:40] And that, ladies and gentlemen, is the adaptive context engine. You need that engine on top of enterprise context that you build from so many places, that toolkit that the data teams uh and the engineering teams
[14:52] essentially have to make all of this come to life. So, the context engine has to be a platform agnostic ecosystem. When I say platform agnostic, uh it should pull
[15:05] from any kind of sources. It could be structured, unstructured, semi-structured. It could be from third-party MCP servers, SaaS apps, whatever you say. And then it should In second, it should be accessible on any surfaces. Like when I say any surfaces,
[15:18] Wisdom AI has a We do a fantastic chatbot, uh conversational BI, AI powered dashboard agents, but that's never going to be sufficient. You will also need to have this accessible on top of Cloud, Copilot, or any other
[15:31] platforms that you're already working on. Is this enough? Is this enough? We're getting there? I think a few of you are nodding that we're getting there. We're almost there, but now this is the most important part
[15:45] of this entire ecosystem. What is the most strategically valuable asset in the AI era in your team? It is that asset which tells us which tables to trust. It tells us which metric definitions changed and why.
[15:58] It tells us how leaders operate. You know what? If you want to get this, go to this person. If you want to get an approval to go uh for get this budget, go to this leader. These are not things that can be documented anywhere.
[16:11] That asset is your data analyst. The data analyst is that central cog that sits across so many different systems and platforms within your company uh that they are able to look at it from
[16:25] a perspective that other people don't get to do. And when we talk to data analysts, what we've noticed from our customers is that 87% of them actually feel that their work has become
[16:37] more. I mean, they they have to do a lot more work. Their work has become far more strategic, not just operational that in previous years they were doing. So, it kind of suggests something.
[16:50] The data analyst who was essentially required to build dashboards, answer ad hoc questions, respond to requests. I mean, this is classic data teams, right? Hey, you know what? Uh I'm going to run a campaign on this particular company. I
[17:03] need data yesterday. So, sends to the data team. The data team runs a bunch of sequels, builds a bunch of static dashboards, and then you send. Takes a time. Takes a lot of time. But, that's a world we lived in previously. Especially with AI platforms
[17:17] like Wisdom, that's not a world you live in. You need to encode business knowledge. You are basically bridging that gap from traditional analytics to AI-powered analytics. You need to encode business knowledge. What means what? When should
[17:30] I use what? Uh who should use what? Who is allowed to use what? Uh and they are the ones who power AI accuracy at scale without any doubt. But, there is a small nuance to it.
[17:43] And that nuance is what we're going to talk about. The data analyst is now not a standalone data analyst. I would call them data analyst plus plus. And here's, ladies and gentlemen, introducing a new family
[17:55] of roles called the AI context engineer. More and more, as we speak to our customers who are, you know, data teams, they're seeing that the kind of work that the data analysts are doing are significantly different when it comes to
[18:08] AI workloads. They are spending a lot of time doing a lot of what you're seeing. The previous world expected analysts to build dashboard, and then users come and ask questions. A lot of the context lives in people's heads. That person leaves the
[18:21] company, you've lost a lot of those things. And then you're just scrambling to figure out what is the right thing to do. But, we live in a world, or we already do, I think it's not tomorrow, it's today and tomorrow. They're already there where proactive AI agents are
[18:34] monitoring KPIs. You all have systems set up where there's a bunch of proactive AI that's happening. Uh it's monitoring KPIs, anomaly detection uh is happening, root cause analysis is delivered to your fingertips.
[18:46] Uh and it is trusted because context is encoded. So, in summary, the AI context engineer, who was previously doing only analytics, is now
[18:58] doing a lot more. They are spending time on data architecture. They are spending time on governance artifacts. Uh and not just that, they're spending a lot of time in being that strategic
[19:10] business advisor to the company. And in order for all of this to be successful, just having a data analyst who is now a strategic advisor, who's an AI context engineer,
[19:22] is not sufficient. You need the right toolkit to be able to do it. And I think this is something that people don't talk about. This is that dirty brunt work that needs to be done for them to take whatever's
[19:34] available. And I mean, it's think of it as like the kitchen with the right utensils. You need the right electronics in the kitchen for all of this to happen. So, the chef is excellent. The sous chef is excellent, but they are very powerful only when they have a lot
[19:47] of these tools in order to make that happen. And in summary, what the ACE actually does is it translate business intent.
[20:00] Uh business leaders don't speak SQL. I mean, they speak outcomes. So, we need uh people who are talking that language. Why is our West region declining? An ACE converts this business intent into a
[20:12] precise, governed definition that AI can act on without ambiguity, without hallucination. Then, they build metric definitions, join rules, business
[20:24] policies, ownerships. This is the work that no AI model can do or no AI can do. It requires human judgment, domain expertise, institutional memory, and the right human judgment that an ACE essentially
[20:36] has. An ACE is the perfectly enabled human in the loop that makes AI trustworthy at scale. And I know a lot of you fully agree with this. It's just that this particular job role is a slight nuance or not a slight, a
[20:50] decent nuance to our quintessential data analyst roles. And this is what we think more and more companies are doing. We have a lot of ACEs in Wisdom. I'm wearing an ACE t-shirt. So, uh we should definitely know more about it. And the
[21:03] third thing is they validate and govern. They QA. Not code QA, business QA. Does the AI's answer match how this company actually makes decisions? Would the CFO put this number in a board deck?
[21:16] If not, why? The gap is a context bug and the ACE fixes it before it becomes a bad decision. And it is the ownership of that particular job job role to essentially go and make that
[21:29] change. And finally, bringing it all together. The unlock. Once enterprise context is built and governed, you don't just have a better chatbot. You don't have better agents that proactively surface insights.
[21:42] That catch anomalies before they become board-level problems. This makes your entire organization smarter continuously without anyone having to ask anything specifically. And
[21:55] talking about ACE, uh we have a cohort of courses that we're going to do on adapt on AI context engineering. This is a course. Please pick up your
[22:07] phone and you could register. The first cohort is actually uh fully booked. So, you could please sign up for the uh waitlist. And looking at the number of people who have signed up in the past couple of days, we might do a second
[22:19] cohort very soon. So, we are doing it so a background on this. We actually interviewed close to 200 Fortune like 200 CDOs, CIOs uh and
[22:31] adjacent of companies that are more than a billion dollars. I do more than a billion dollars. And there are a few things that actually came together. And one thing that came from that is how do you build that skill set that makes all of this valuable. And
[22:45] a lot of this this cohort or a lot of this coursework or pedagogy is built based on what we've learned from them. Uh we have people from a lot of uh massive companies who have actually signed up for it. So, this is something
[22:57] that is going to be extremely essential for the next generation of analytics with AI. So, to summarize, let's talk about what we took away from this conversation.
[23:12] Number one, the multi-surface future is here. If someone tells you dashboards are dead, if someone tells you that my organization just needs a chatbot, we can just ask a question on a chatbot and
[23:25] we'll get all the answers, you know that they don't know what they're talking about. Everyone has a different relationship with AI. We all agree? What I do on my cloud code is very different from what my wife does, what
[23:38] my dad does. And the way they consume information is significantly different. The modalities are different. And because of all of this, and because of access to all of this, everyone is going to try out a bunch of
[23:50] things before the organizations actually feel that, hey, you know what? A a certain flow is cost-effective, a certain flow is more productive. So, that's going to take some time, and we are nowhere close to that. So,
[24:02] uh we need to uh acknowledge that we live in a world that it could be cloud, it could be your own React-based apps that you've built internally within the company. It could be any surface that you have built or
[24:15] you've bought. So, the world is moving to that moving away from that quintessential world. And in order to make this happen, you need a context engine. Wisdom or not, you need a context engine. You There is
[24:28] one where you're repositorying a lot of the context, but there is the additional work beyond that, wherein there is an engine that adaptively brings in context from structured, unstructured, semi-structured, MCP, whatnot. It brings
[24:41] it from a lot of these places, and gives you the tooling to keep it simple and keep it in a governed uh ecosystem, which sounds a lot like your
[24:53] SDLC. So, that's the context development life cycle. So, this is basically building your context as part of a life cycle and maintaining it, versioning it, governing it, so forth. So, you need that toolkit, that platform to be able to do what we saw in the previous uh
[25:06] slide. And all of this could be on any of the surfaces. Could be a chat, it could be embedded, could be agents. For all you know, it could be a headless world. People would be using uh cursor, for all you know. Or it could be dashboards, and
[25:19] dashboards of the future. And the third one, probably the most important one, is analyst is now an AI context engineer. I mean, I can go on and on and on about this because in the new role,
[25:32] understanding of business is paramount. Understanding the data architecture, which an analyst was not doing earlier, was not required to do earlier, is going to be extremely critical. The analyst is also going to be the curator of history, which means what has
[25:45] happened, what changed, what was important to be added into uh my context ecosystem. In Wisdom, we call the context building our unit is called a domain, and we often
[25:57] see that the moment someone sees that you can kind of put in a bunch of your documents, they start with adding anything and everything that they have. And then they realize, hey, and if whether they are running it through
[26:09] Wisdom or they have their own LLM key, they see that they're running through tokens a lot. And that's because they they think that, hey, you know what? I'm just going to throw everything in my organization within the domain context, and that's actually going to take care
[26:21] of it because everything will be thrown to the LLM, LLM will give the answer, Christmas is over, we all go home soon. It doesn't work that way. But it takes iterations for us to get there, and the adaptive context engineer, the ACE, is
[26:33] essentially that job role which is going to be the curator of this workspace. They are also the quality controller. What is right? What is wrong? Now, the hard problem is when you have a strictly probabilistic system, which is
[26:46] the underlying uh engine, like you like if you have an LLM, which is strictly stochastic in nature, and if that is something that's powering your entire world, you need something that grounds it in
[27:00] reality, something that adds that determinism. And that is how you can actually build trust, you can build that quality bar, and the ACE is the one who's doing continuous evaluations, filling in those gaps
[27:13] whenever that is needed. And the last one is they have to be the right storyteller. You have a system that's generating all kinds of probabilistic systems. But then when the VP asks for what happened Q over Q last month, you need to be able
[27:26] to give that right answer. In order for you to be able to give that right answer, you need someone to understand what part of the system was probabilistic, what part of the system was done outside the probabilistic uh ecosystem, which is what is deterministic. How can you add
[27:39] determinism into it? So, you need a platform that lets you add that, a context engine. And then you need a person who's the storyteller from that probabilistic system to uh
[27:51] uh to leadership when you're kind of uh summarizing this. Does that make sense? One last slide. Why did we play this video?
[28:04] With all of that we spoke about, there is value in looking at this video and contextualizing it, no pun intended. Contextualizing it to this talk. Uh
[28:17] the song itself, whether you thought it was brilliant, average, or terrible, is like the output of AI. And AI outputs are only as good as everything they're built on. The person egging the singer, the person
[28:30] the person who's standing up there, that's the world around us, constantly telling us, "Use AI. Try the latest model. Adopt the newest tool." And that's important. Models matter. But it is you but if you look closely,
[28:43] the real magic isn't happening there. It's happening behind the scenes. There are two people There are two components over here which are extremely important. I feel the most important person in that room is the sound engineer. The one who
[28:55] understands every nuance, every signal, every piece of context required to create great music. In the enterprise, that's the AI context engineer. And even that isn't enough. The sound engineer is only as powerful
[29:09] because he operates a very potent digital audio workstation or they call it DAW. So, you need a context engine, an adaptive context engine for it to be powered. Now again, I'm part of Wisdom. I'm talking about the context engine
[29:22] from our our platform. But one way or the other, you need that toolkit beyond the repositorization of context in order to make your context come alive in the world of AI.
[29:37] I leave you with this. The assets best positioned to take you to the next phase of agentic analytics are already on your payroll. The only question you have to answer is whether you will redefine their role before your competitors will.
[29:50] With that, ladies and gentlemen, I'd like to end this and open this up to Q&A. Thank you very much. >>

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.