Feb 10, 2026 Enterprise Brain Connect - Transcript 00:00:00 Rakkesh Yenugudhati: are recording this same. So based on the questions what they shared with we try to map the same exercise what we did for regular scenario. Okay. Translation. Sorry. 1 hour because he has discussion. Okay, sure. had a conversation with whatever questions. Sure. No first time different scenarios visualizations you see we try to map to the what we can later what we have done is like considering the sales force as a thing and we created a different scenarios and we map it to what kind of an experience that we are mapping to the flourish example the third is like we asked them to create a questionnaire it was these are the questions given provided by them Raagaran based on these questions again I try to map these questions to the scenario question okay Is that mon? No. No. I was looking shadows. I'm average just a high level I think we have to do it with the T3 so deep level knowledge uh deep level knowledge then complete understanding work in a working model one two two weeks to two days Okay. 00:04:21 Rakkesh Yenugudhati: Exactly. Okay. Okay. research and I don't know what level of research you have to do what level has to be started executing you need to start off on that we need an alternate idea do we want to stick to flourish or do you want to explore other Python based libraries for UI itself I'm not working flourish from the angle of technology. You need similar representations. Because usually engineering teams they can yeah visualize and associate it easily rather than when I go and explain clear as This is Jmanes. So we are good at reproducing when we have clarity at least I don't know maybe one of the calls what he was mentioning is You will give the inputs to it. It will bend and it will give some URL kind of same. You can create that dashboard there and you can expose. Okay. So column E questions and different different roles. So those will be the prompts given by the se or somebody sales graph leads one lost diagram Open it. 00:07:54 Rakkesh Yenugudhati: Sales. No diagrammed implementations. Okay. Deals a bit. What is the intent behind email something right? Multiple emails has come. So because the emails can be popping from multiple I mean months back days back but the context of that is coming as a single story to me what happened and the email content it's a summary of that email in that I mean is saying yes. I think Morning. The decision will taken by the system. You define the type of graph and you pass the object what it has to render different. That is one way. That is not a graph actually. It's not a graph. for that is what how I understood. So we know that there are different sources that we connect then we'll go for a technical see there are two kinds of approaches problem the other one Okay. And I'm usually a top down person. I I give some top within and expect people to break it down and do it. 00:12:37 Rakkesh Yenugudhati: Okay. In this particular case, we are going bottom up for and we can meet somewhere mid level. What I mean by that when I'm coming from top down which is what Rakkesh was also asking or we were discussing earlier how will this be? What are my navigation? Our knee is top to bottom. Now bottom up independent of navigation, independent of login, independent of anything. A person is asking you a question. Answer understanding clear now see pure top down is pure product engineer right it's a front end layer you agree Or you think there's any that one is a product engineer. That's how I broke it into my head. This whatever we are talking is combination of product engineering and AI. So in my head two teams are involved. Let's crack this first. layer top that is not something I even we have to come up with a new pattern we are we are in sync any questions no okay then one of the things that I have been working on is ward a problem I'm trying to break it into smaller units at autonomous units and anything that can be run parally let's run it. 00:14:49 Rakkesh Yenugudhati: Don't make anything sequential. Don't make anything sequential. When morning got up when he put up message saying nobody's time 8:30 to 9:30 I immediately said I unblocked Rakkesh Nan no we we unblocked there were only two people Naven and you with whom he had to come right so I'm trying to fix things from different angles at the same like we make sequential time frame So, we'll figure this out. But tomorrow, we should be able to crack if we can crack AI first. AI has to drive first. Can this be done first? Right? You're saying it can be done. My second question is in terms of this this is description then the diagnostics then insights. This is insights and description then diagnosis then prescription then prediction eraph how will you bring all of that? Okay. Now this is a classic problem of sales. If you're connecting to Salesforce leads, MQLs, SQL, negotiation, proposals closed, Salesforce because database This itself is a decent problem for already he's doing it. 00:16:59 Rakkesh Yenugudhati: It is all about marking it to this graph. Right? Now the paths just okay multiple parts that is another one. So this is description and insights. Okay. Just this description insights. Insights. Now what is the diagnostic layer that can be on it and what is my goal? Okay. gun count then why did I not meet my goal is agnostic layer how will you present it here that it's a narration this narration can be narration This is okay. Uh narration is like a story in which there is no interaction but diagnostic he needs to give us what he's trying to diagnose the problem. So user input then prescription user input prescription user input prediction user. So is one problem. Now a problem for other problems. What will it evolve into across the pyramid system similar problems principle problem is how do we apply this graph and all of this first thing you started with that same key diagram and you clusters And I assume that whatever the diagram the first graph whatever we are plotting assume that this is a descriptive just information insight based on how we want to descriptive diagnostic there should be some triggers I want to analyze based on this giving 00:20:24 Rakkesh Yenugudhati: another contain to answer that I need to go to a different metaphor representation will change the representation of the graph it depends upon the context of the question and the type of representation that is needed to explain that answer because it's all dots you feel like the name graph. There is another place I will show you where the theme changes. But will the user be able to relate to that problem? One problem. graph. It all depends upon the narration. What is the story you wanted? If the story demands two villains, we get two villains there. What I would recommend is W take this as the example. There should be narration because Manisha Anyway, Satyasri Prabhakar Mantripragada: narration is going to be a descriptive that is very clear from there how can we narrate all of us should we take that it will be Yes. MQs SQS in progress. Uh SQL is itself an opportunity. Then there is proposal one and last. 00:24:27 Satyasri Prabhakar Mantripragada: Okay. No description. Yes. inside. Why that particular lead is not moving into a different it's been so much in that in progress state not to state number of days that it it was the number of days that is in waiting but diagnostic where the n number of leads not converting into n number of mqls and downwards. Yeah, because that is a user input. Whatever is the descriptive layer is presenting that is nally prop that should be the goal. If it is not, if it is deviating then that's where we need to find out the first layer of description. Not here. Do you need a narration here? Still we can give the narration because from which of the US rest of world people by location. By location. Yeah. Okay. Okay. Yeah. Okay. There are two multiple assets to this. 00:27:09 Satyasri Prabhakar Mantripragada: One is revenue goal. Each person may or may not have a revenue goal. If they have a revenue goal, what will your narration be? Each salesperson accomplished their quarterly target. How many there how many clients they met? How many are in positive statement to state? What is the projection? Diagnosis is low. performance of an individual. We can say the company target is to generate this much of revenue. So far we have n leads which are contributing to x revenue and we are deficit by this or we are we have already exceeded the target. Yeah different analysis process. Yeah. Then we have to say my typical questions would be either how is this different from how is this AI first and I'm expecting user to click everywhere how is it AI first now if it is AI first what should be the alternate solution this would be one question I'll ask first you need to and when you're coming up with the solution in the middle of the solution normally Don't question continue the solution then come back you should always have a checklist satisfy satisfy 90% no to drop it then start all over it that is how you refine your solutions 00:29:25 Satyasri Prabhakar Mantripragada: for example in this example we are showing the first level information level one now the Second one narrative story already some predefined flow is defined how you want to drive if you looking the how the the goals with the team flashard I don't know what to call it assume that you are creating a different flash cards one is like you want to drive that from the employee individual goals perspective how it was converting or you want to target in terms of the revenue that each lead is generating assume that there are multiple cards If user clicks on that that will he's entering into that story narration which will able to tell the about as a holistic how we are achieving as an individuals how they're achieving and whether they're aligned with how much deviation they're aligned with the goals answer questions. What is AI should throw the information at you inputs have to be uh it'll become which is also fine not denying a person solution solution with minor differences and then differ what needs to be done. 00:31:17 Satyasri Prabhakar Mantripragada: Next question because for me the AI first because AI is throwing that suggestions to me because what story that you want to because you right now looking into the leads pipeline okay if you consider that that's AI first what will be user first user will be asking the question because we are not giving any hint to him he's asking that what he Same. Yes. First input has to be first dashboard can show this graph without even user asking. first then in just example again user first law user will ask for a question hybrid law I don't we'll have to figure okay question thought process actually one more for example whatever the current example that we do is actually like for example if I'm having any sales force or anything I will log and it will give me that you know any deals any deals information right what are the hidden facts right one of the questions that I have in my mind is is this is the right use case that we are I'm not use case is right okay use case is right okay Okay. 00:33:28 Satyasri Prabhakar Mantripragada: What do I care? Am I making my revenue goals or not? Okay. Goal is always the same. Imagine it like this. Everyday morning I have to come to office. Sir goal. Whether you are taking a fly or whether you are taking a local role. Is there a fly over is the only difference. AI person the or enterprise brain is like a fly over which will reduce your time by another 15 minutes or by 50% or 100%. End of the day you're making innovation is in how we are solving the problem. All of us have the same problems end of the day. Some problems when they are automated we get new problems. For example, if your car is automatic, you're not sitting and driving it and it is a nice car where you can take meetings. Why would I care if I'm coming to office at what time, right? I have a meeting place. 00:34:39 Satyasri Prabhakar Mantripragada: I can take all the meetings. I don't have to worry about driving. It's not it is stress right because driving and taking a meeting you have to present it is an automated car it's as good as working from home you're understanding the difference so problems are always the same enterprise brain is making it extremely convenient for the user to get to what he wants within few seconds. That's the advantage of yeah I'm aligned with it. Sure. I'm not saying this is the right solution solution. Yeah. Then we have to start somewhere. We are starting starting. Sure. Then my question my checklist how is it AI first? How is it designed first? How is it different from sales salesforce dashboard? I have to be different. Yeah. If I'm giving the same graph as Salesforce dashboard, now why would somebody buy enterprise? That's your checklist. 00:36:01 Satyasri Prabhakar Mantripragada: Yeah. Okay. I usually like to conclude on a solution, go check the checklist, drop the solution, rebirth. That's how innovation happens or not. during coming up with a solution when you ask thousands of questions right you'll feel I'm lost I usually like to converge get a solution go against the checklist half of them failed this is not the right solution restart you understood you're clear right we have to start somewhere and our brains right usually go only backwards so We know only a certain way of solving the problem. So it'll always go back to that. That's why your checklist comes into picture. Your guards come into picture. Now your next question is how is this a solution? Maybe not. It's not an AF solution. Will my second question is will Salesforce also give this? Absolutely it can give. Then what is the differentiator of enterprise? What is a user first? 00:37:14 Satyasri Prabhakar Mantripragada: What is the hybrid first? That is why the narration that you're talking won't come into a 2D enterprise but the narration has to be seamless. That's why I was rakish when he said user input is the name user input is as good as drill down then it'll become like another sales dashboard so it can't be that then what is it so negation not this not this then what not this not this then what that's one way of innovation Yeah, sure. We'll we'll continue. You need my help further. Uh okay. I mean at least to me we are still starting stage in I gave a framework. I think whether we were at starting or ending. Yeah. So what are your frameworks? Prescriptive or descriptive one. That's one framework. Yeah. Experiential innovation is another framework. Then any solution you're coming up with if it is AI first, user first. 00:38:42 Satyasri Prabhakar Mantripragada: That's a third. Then the patterns be recognized where will narration come because narration is very unique to enterprise or any NLP and AIdriven solutions, right? It won't be it won't be in 2D. Look narration is not easy. Narration usually is maybe they'll start including it's not a big deal to include that into 2D but a regular dashboard's local now they can include narrations. Narration we saw one kind of narration. What are other kinds of narration? Dal is You guys anything that without user intervention something that we are throwing to the user for me It's like a very first anything that user is asking for me to do something that's in user first maybe both I don't that I just came up with because I'm assuming anything that fall into this and this will fall into high now this is the if you start with this graph the two ways to do it because either the assume that if you believe that this is the common thing that a user will want to know on this platform. 00:40:49 Satyasri Prabhakar Mantripragada: It can be a one of the way when he land on other things. This is the one way to tell that this is a quick view kind of a something so that AI is showing something as a checklist ask what are my different than the first show that the user first be calling. A first is it only because from a natural language you are able to extract the information or is there any other parameter that's based on how things are happening and system will uh what so system is giving you the information without any human intervention that we are calling it as dashboard system is giving some sort of information. So regular dashboard what should be the difference no matter what somebody has to code right now show me list of products that are sold in this particular area or by category so that we code it actually uh whereas the code is static the code is static whereas as an orient at least we we decide the direction that this is how it should be that it is assumption or not we should decide the direction that this is how uh whatever output that you are presenting should be dynamic to the context that is asked for it need not necessary that every time it 00:43:25 Satyasri Prabhakar Mantripragada: presents a sank diagram for a similar question not same question yeah it can pick anything able to generate because it has generated a different query now different API altogether It decides what to generate, how to present. Fine. System will behave. What need to be shown to me? Not only system personalization. It's not a system preferences is contextualization. System preferences. Contextualization. What? Setting the context. Context. Contextualization. System preferences. User preferences. User preferences is personalization. I mean I'm trying to learn that what you liking what you're not liking that is the person I put it preference is like I I'm interested in these domains sales marketing domain sales and marketing what kind of articles that you read that is the person to the Uh no name so that all of us are on the same page. No your whatever you're talking about evolution of the content based on your usage is a yes okay personalization is nothing but user preferences where I define them. 00:45:30 Satyasri Prabhakar Mantripragada: Okay. Contextualization system reflexes environment whatever contextualization context is different personalization different contextualization is different because for me as an organization it can set at an organization level what is a color branding and all that here personalized contextual uh means like because based on what I am triggering in the next drill downs. Okay. You define your regular words. What is contextualization? What is personalization? And you define Jesus. No already personalization solution. Yes. Okay. What is personalization? Showing me what I want. Personalization general. Yes. Personalization. Normal. normal system. That's what you're trying to say. Okay. What is details? Preferences. Personal preferences. Personalization system preferences is contextualization. Now first law one of the places that we could do is no last these are the no no based on your role these are the things that you need is what is personalization. 00:47:47 Satyasri Prabhakar Mantripragada: person. So first what we are trying to say is then the question key answer which is um the question key a question answer how is this different from 2D 2D 2D regular system 2D if you have to Answer that question first. Regular system key AI system key what's the difference? Regular system your dashboard is always constant. The data within the the graph is always constant. The data that you shown will keep changing. So the only dynamic here is the data. Yeah, we agree in a regular system there's a choice of graph is constant, API is constant, information is constantly constant. Now a first law what are all the dynamic things? One definitely data is dynamic graph or data is going to be data is going to be common. Yes. What are the different data is dynamic in both the cases. It is no longer a differentiator. for lack of a better way of doing it. 00:49:37 Satyasri Prabhakar Mantripragada: Four KPIs or three KPIs uh revenue goal. What is the deviation? Okay. So, amount of data you're showing could be different. Mhm. I could shift to graphic. So in the end what you're trying to say is AI first law there is a lot of dynamism driven by the system clear now dynamism dynamism. One data evolution will define it. The more and more data, the more and more the better the model is. Okay. Uh for example talking about how the uh how the sales pipeline is there in an organization. That is a use case that we are looking at. How is sales performing in an organization? Yeah. So I understand the system another parameter that I think that got into my mind is so you do a lot of analysis and then you come to a conclusion that is a kind of a reactive current situation, current day, not you but generally in organization driven we will system will proactively say that you knowap that is what we are saying it is an example Just hold on to your thought only. 00:52:56 Satyasri Prabhakar Mantripragada: Okay. You are on par with your revenue deviation almost zero. So you're good to go. Your cost of selling is extremely high. spending 20 hours is spending another 20 hours spending another 40 hours. This was not there earlier. Okay. The has to throw it out and say watch out your while your your your revenue is increasing your negotiate your negotiation of profit is reducing information which will impact my decisions. We know that it's a that is the language that we need to tell to the user for example I need to show when you're talking about the revenue or sales cost I should know for me to tell it as a require I don't have to know it as a requirement. AI is finding the insights and throwing it at me. showing because how it is going to prioritize that this is most relevant to tell okay otherwise I'm good with other things and retire But my assumption is solution. 00:55:23 Satyasri Prabhakar Mantripragada: AI solution is all about helping me discover certain things that will influence my decisions and help me take better decisions. regular systems. Even now. airdriven solution. People can build in the dynamism and discover the most cost of selling may go high. So the regular so data maintaining data patterns is what I am anytime it goes out it should call out and you should understand what is the impact that is what AI is to me I'm pitching Nice. I said MP 40p. Hey, nice. Okay. Now these are the guards for these are the guards. See then it get it will get slightly easy. There's no end of the day we are defining the guards. I'm not saying they're already different. So next and where did we stop? Uh a first approach. What are the parameters to consider and how do we define a first and Okay. 01:01:59 Satyasri Prabhakar Mantripragada: A first versus user first versus first solutions. Third thing in tech first and for lack of a better air first solution. It has to be intelligent and automated. Primary drive user first human need and experience. Experience should be the driving factor in human need. Thirdly architecture starting question for an AI first. What can AI predict, automate or optimize? What can he predict? Automate optimize user first. What problem is the user truly facing? Yeah, first data models learning systems drive everything. Use first usability option key outcomes and smarter decisions. First use first lema higher adoption retention satisfaction and loyalty. Air first. It's the central brain of the system. A use of first an enabler where A is giving recommendations, insights. technology role in AFS is platform to operative intelligence operative intelligence I don't know what exactly that okay technology role in user versus seamless experiences for the user typical use cases predictive analyt if over success metric. 01:04:37 Satyasri Prabhakar Mantripragada: A first success metric is accuracy learning rate from automation. Success metrics. What is common across all these platforms and design thinking has to be common independent of tech user uh everything has to have clarity and usability business objectives have to be common scalability security measurable outcomes cross functional collaboration to be common Here we go. The system is designed around learning continuous model prediction, reasoning and adoption. Manual decision is action performance. User perspective. User perspective regular products or platforms user manual decision. Typical traditional products. Yes. That is recommendation exempt. So automated recommendation per action. So UI first it has to explain AI decisions and human overt AF solution of prediction failure is one of the things that we have to monitor and we have to continuously try to reduce the prediction failure rate. Increase productivity, reduce optimization cost. Continuous learning loop observes behaviors learns from outcomes, adjust the predictions based on that. 01:09:36 Satyasri Prabhakar Mantripragada: Improves recommendations. If the system is not getting smarter with usage, it is not a A first law AI proposes human validates, overrides and refineses if required. Human override system learn it will automate. That is what it is saying. It is critical for trust, accountability and adoption. Air should always explain why show confidence levels surface tradeoffs cost of that is what is surfacing tradeoffs allow inspection of signals trust is not a feature is a product. Shall I ask him to come? Continue. They may be waiting for us. brainstorming. One of the questions 20 line. That's fourth point. Two things of right. Yeah. together. Yes. Well, that is always an abstract thing. Sorry. funny. Yeah. I'll get it. Okay. October there is 16 crores that they have to collect that government has stopped. So 18 but they don't have because 18 collectibles government you know apparently in the retail industry the government pass something to say to right now just suspend all the commission payments until they're they're doing some investigation and investigation suspend better you know but they're just not able to raise that. So chairman is going to come back because there's corrections. There is no cash flow. close. would be no business they can achieve. Transcription ended after 01:15:53 This editable transcript was computer generated and might contain errors. People can also change the text after it was created.