Workshop - Conversational Analytics with BigQuery Agents (Professional Track 1 - English)

Added: 2026-05-18

[00:00:32][music] [music] >> Woo!

[00:02:13]>> Hello. Good evening, everyone.

[00:02:15]Welcome to the Cohort 2 of Google Cloud Gen AI Academy APAC Edition. My name is Manita Ahuja and I am your co-host for today.

[00:02:23]So, uh this time we are launching Cohort 2 with good number of success and uh this time we are taking uh this to a stage with unified data analytics and intelligence. Following the success of Cohort 1, Cohort 2 shifts the focus to unified data analytics and intelligence.

[00:02:42]This immersive journey empowers you to build autonomous data agents that bridge the gap between manual SQL and actual language discovery, moving beyond static dashboards to real-time agent-driven insights. So, today uh we are taking care of the topic that is conversational analytics with BigQuery agents. And of course, uh moving forward uh this time we have two tracks in Cohort 2, that is professional track one, that is ingest, transform, and govern, and professional track two, that is intelligent applications, multimodal agents, and analytics. And this time we are also introducing the student track, that is ADK crash course, because we also want students to build AI agents with ADK.

[00:03:29]And one more thing which I also wanted to inform you that this time we are not going with the project submissions. We only have uh assessments, that is you have to attend the workshops, you have to do the code labs, you have to do the MCQ assessments, and moving forward, there comes the prototype submission.

[00:03:49]Moving forward, will not take much of your time.

[00:03:55]Let's introduce Mr. Ashutosh Bhakre, who is the GDE Cloud Docker Captain at Google Cloud.

[00:04:02]Hey, Mr. Ashutosh, how are you doing today? Hi, good morning. I'm doing good, Monit. Thank you so much.

[00:04:08]Okay, so let's move forward with the topic that is Just a second.

[00:04:18]Okay.

[00:04:20]Over to you, Mr. Ashutosh.

[00:04:21]Great. Hey, hi all. I hope everyone is doing great today.

[00:04:25]And I welcome all of you to this GenAI Academy session.

[00:04:30]Uh yeah, I'm a Google authorized trainer.

[00:04:33]I'm a Red Red Hat certified instructor, and I'm a SUSE certified instructor as well. I'm here to talk about how to deal with uh you know, agents. So, I could say agents are everywhere nowadays, right?

[00:04:46]Uh if you see the way how the disruption is happening in the LLM or using a GenAI ecosystem is at a very rapid speed, right? Or probably a rapid space pace, sorry. So, you know, in this session, we will be talking about BigQuery, of course, uh which is an analytical platform. So, uh if you if you see how BigQuery works in a OLAP ecosystem, it usually uses SQL.

[00:05:19]Uh SQL queries and all that, right? So, recently uh I think in January itself, I'm not sure about the exact date, but recently Google had announced the agentic feature to our BigQuery scenario, right?

[00:05:37]So, even when you do a code lab, you will see that it is in preview, by the way. So, we will be doing a workshop which will be comprises of building conversational agent in BigQuery. So, in a nutshell, how I can talk to my BigQuery using a natural language or in a simple plain English. That's what the code lab that's what the workshop is all about.

[00:06:06]Ideally, it should take 90 minutes at a you know medium pace. If you are pro, if you are expert, it may help you to save little time. Probably, you may able to do it at six in 60 minutes. The objective of this is how to deal with the you know BigQuery in terms of continuous query.

[00:06:28]How to create a knowledge base.

[00:06:30]Probably, a knowledge source and how I can leverage Gemini features to deal with these kind of entire logical scenarios.

[00:06:39]Yeah, basic prerequisite is required, right? You should be knowing Google Cloud, some basic knowledge of BigQuery and how the SQL scenario works. These are the things which uh are the prerequisites for this workshop.

[00:06:54]Having said that, to do this workshop, our team will provide you or you might have already received an email on how to redeem the credits. So, I believe you will be getting frictionless $5 credits to complete this lab.

[00:07:10]Complete this code lab, right? By any chance, if you don't see the email or if you are not able to get that, please check. We will be sharing a QR code also to redeem it.

[00:07:23]Maybe a quick note, while doing the redemption, by any chance, if you have not done it, please don't use any non-Gmail ID. Make sure you're using a Gmail ID itself.

[00:07:33]So, before we talk about this entire code lab scenario, before I walk through uh using my cloud shell and the cloud console. Let's have a basic discussion in terms of how this BigQuery and analytical ecosystem are important and some of the pre what do you say discussion what is needed. So, yeah, this is what I was saying actually. So, I think the next wave of data and AI is agentic. Right? So, it's it's all agentic. So, Google has anti-gravity, Google has ADK. Right? Google had kubectl CLI. So, so these are the tools. Some of them are open source, specifically kubectl AI and ADK is open source.

[00:08:23]Anti-gravity is there which is an agentic framework to deal with an application development and even deployment also. But yeah, so these all are coming using an agentic scenario. Right? So, if you see how it work or if you see the history, right? We came from an analog you know, in early days in in 1950s or 1970s till 1970s we was having this analog machines and then we got database evolution. Then we got big data era. I think it was 2000. Yeah, 2000 2010 was the era. And then we got an uh uh you know, ML AI and machine learning uh AI driven decision-making scenarios uh in 2010 and 2020. And now of course, uh it's all about the agentic era to deal with the data life cycle. How you automate, right? How you complete the uh you know, data life cycle using an agentic era. Right? That's that's all about the uh eco agentic era ecosystems nowadays.

[00:09:37]And of course I think uh that's a shift, right? Uh so from uh you know, if you see from human scale to an agent scale, right? So agents, of course, some people consider it as we are replacing humans using this. So I think uh I could say no.

[00:09:56]We are using it as a helper or we are using as our co-worker for dealing with the uh data teams in terms of probably orchestration, probably workflow.

[00:10:08]Even uh we can use these uh for analytical work without context switching, right? And of course uh uh you can even as in business user, you can directly talk to the agent by giving some basic prompts, right?

[00:10:25]Why like of course, you don't need to deal with what is happening in the back end. That's the shift from human scale to agent scale what we have.

[00:10:34]So Yeah, of course, why?

[00:10:36]So if you see uh decisions. So whenever uh uh being a data, so whenever it comes to data, uh specifically enterprise data, right? Uh you need a low latency based ecosystem, right? So even if I have a chat interface, I should need or I need an instant answer, instant reply, right? So uh in in that case, the agents works seamlessly, right? Uh you can use it as an everyday tool, right?

[00:11:13]And yeah, of course, these are technically a self-service uh ecosystem or explorations. Uh as I said, it will be natural language, right? So you will be talking uh without dealing with the queries, right? That's that's the logic and then it can do the action label, you can uh get the seamless workflow, and then you can get the instant answer, so that's that's the point what you can deal with these kind of uh uh BigQuery and agentic ecosystem, right?

[00:11:47]And every employee and a customer can explore data in a plain language, right?

[00:11:51]Find hidden insights, have a partner to deep dive and turn inside into actions, right? These are the logical things which involved into this BigQuery and the agentic scenario. You know Yeah, again, so the language barrier. So, I think I think when when when I see language barrier, this is talking about SQL, right? Again, structured query language. So, rather than doing the query this way, right? I can basically do this way.

[00:12:22]Right? So, for example, the question is the query is what what were our top three selling products in North America last month, and how does that compares to the previous month? So, those who are coming from an SQL, they know that we have to write this query, right? Where you have to calculate the sum when the the particular date, then the sales, else, right? All these things you have to query.

[00:12:47]Lot of things you have to take an efforts to make it happen, right? So, I know those are uh coming from an DB experience, they can understand what I mean to say, right? So, yeah, that's that's the language barrier what we was having, right? So but if you see uh how these things are uh getting resolved or from what to what next we will be uh dealing out, right?

[00:13:17]So, if you see uh what, maybe I have a query called I need to check the sales figure, inventory level metrics. How How does it compares?

[00:13:26]Why? Analysis, correction, right?

[00:13:29]Forecast, trending, right? And then it comes to AI power recommendation.

[00:13:36]So, these are the logical scenarios which will help us to get an you know advantage of agentic ecosystem, right? Yeah, that's that's the point. So, again, if you see from data to a semantic knowledge even deep to to fix it in a proactive action, right? From operating to human to the agent scale, these are the, you know, uh advantages to uh uh current what I say, the analytical ecosystems, right? So, from system of intelligence to system of action, right? That's the best part. So, from the knowledge base to the agent scale, that's that's what we will be achieving using this ecosystem.

[00:14:22]And if you see Google Google agentic data cloud, uh yeah, I I I just want to take you little off track here. It's not about only data cloud what Google is pitching nowadays.

[00:14:36]If you see uh uh you know, Gemini enterprise, right?

[00:14:40]They have like recently we have a CX agent studio, like from dialog flow to a CX agent studio, which is again a conversational agent which you can build in few clicks, right? And trust me, guys, the revamp of the uh the previous conversational agent, now we have, you know, know it as in CX agent. Of course, it has a suit of different tools, but CX agent is one of the tool what creates, which creates the conversational agent uh in terms of dealing with the customer experience, right? So, so point is Google is not only the agentic uh era in the data cloud ecosystem. We are Google has uh a step in our Google has other products, which is again using the same agentic scenario, right? Coming back to the data, of course, uh we our Google had recently came up with this uh agentic approach in the BigQuery. So, if you if you see how it works, so uh of course, as I said, Gemini Enterprise.

[00:15:48]Then agentic development, right? So so as an front face, this all looks like a Gemini uh infused uh ecosystem in terms of dealing with the agent deployment or agentic deployment or Gemini Enterprise, right?

[00:16:05]Which is an AI-powered uh backend. So, if you see, it has an uh advantage of doing analyticals operational database and business intelligence, right? And the best part is it is AI-native, cross-cloud as well. So, I can I can use not only Google Cloud Storage, I can use AWS, Azure, or any on-prem uh ecosystem as well. So, that's the whole lot of glimpse in terms of data cloud or agentic data cloud what we have.

[00:16:38]So, yeah.

[00:16:39]So, yeah, this is this is just a you know, glimpse of what uh Google has in doing and what are the agentic approaches. So, let's talk about the foundation for confident conversations.

[00:16:51]So, let's let's get into it.

[00:16:53]So, you know, uh if I have a uh if I I want to talk to the uh agent or sorry, I I have to deal with data, right? So, uh having a data or rather raw data without context is not enough. That that is anyway uh uh obviously understood, right? So, context has to build into the agentic data right? So, human analytics carry out invisible trust work, right? Agent fails to find the answer and hallucinate, right? And make a costly error. So, why why we are saying costly errors? Tokens will be burned, right? So, yeah, again, without context, data may not be data, right? That's That's the logic of the this slide. So, okay, cool.

[00:17:41]So, I was talking about knowledge catalog. So, the the logic is how you build a pipeline kind of ecosystem from the context federation to the agentic scenario. So, you build a knowledge catalog in between to make your agentic ecosystem probably a non-hallucinated and referring this as a knowledge base, right?

[00:18:07]So, how the conversational analytical agents uh works across this Google cloud Google data cloud, right?

[00:18:17]So, you know, we have a Looker, we have a BigQuery, we have a lake house, we have other databases to get this, right? So, but but if you see, yeah, of course, we are focusing on the BigQuery here. So, we have an integrated GNA centric hub for data to deal with professional productivity and publishing the agent for the BO for the business users, right? That's the logic. Even we do have an governed BI for semantics for a natural language data explorations for hundred and thousands of users powered by LookML. Of course, that's there. Even we have something similar with lake lake house and database. So, but anyway, our focus is on BigQuery, so I will not talk much on the other aspects of it.

[00:19:12]So, conversational analytics in BigQuery. See, you you are you are having an agent here. So, there can be a pre-built or there can be a template which you can select and you can even quickly convert that to an agent and then you can ask questions interactively in the natural language.

[00:19:33]The best part is it is multi-model data, right?

[00:19:37]It can be text, it can be video files, right? Uh and of course it does a deeper insights. It does predictive analytics.

[00:19:48]Yeah, of course. See, it's not only a agent who is doing that. I hope you are getting the logic behind two different aspects like BigQuery and agent. So, we are combining the USPs of both the products and that's how we are making as an agentic or or probably an analytical conversational analytics in BigQuery, right? Where again, we're talking as in human or you just put in some messages, you will get the answers, right? Of course.

[00:20:23]It even supports multi-agent workflow via API. Even it supports MCP and the Gemini CLI also. So, these are the advantages of dealing with analytics or conversational analytics in BigQuery.

[00:20:39]So, how do you build a conversational uh analytics? So, of course we have an API here.

[00:20:48]So, even as in part of code lab, you have you will be enabling this API, of course. So, once you enable this API, then you're probably I have a BigQuery or Gemini Enterprise, right? Or AlloyDB, Spanner, any sort of data can be integrated to create the conversational analytics, right? That's the whole moral of the story so far. And yeah, so you know, if you see a use case of the current ecosystem, what we have is of course YouTube.

[00:21:24]Right? So, by utilizing Looker's conversational analytics, YouTube is enabling partner managers to deliver immediate data-driven insights, personalized growth strategies to high-performing performing creators, right? So, yeah, these these are the logical things what This is a use case what we have. The Yeah, of course, if you see the streamline approach, right? So, it has saved around 1,000 hours because of the analytical or conversational analytical behavior.

[00:22:03]Right? That's the best part. And you can read it from the horse's mouth itself.

[00:22:07]So, this this statement is from uh senior director uh from a YouTube business, right? That's the best part of it.

[00:22:16]Yeah, some local use cases uh So, yes, this is a success story of a YouTube in terms of conversational analytics what we was talking about, right?

[00:22:28]And yes, uh let's talk about the hands-on workshop. Uh so, this is a QR code which you can use to redeem the credits as well.

[00:22:40]And yeah, let's get into the code lab.

[00:22:42]Sorry. Let's get into the code lab. So, the code lab is what we will be using is all about introduction to conversational analytics in BigQuery.

[00:22:54]So, you know, there are some prerequisites uh we will be having to complete this code lab.

[00:23:03]So, I will be doing a demo from start start to scratch. So, I have redeemed the credits already. So, of course, I will be skipping that stuff. But, by any chance, if you still have an issue, you can please contact our team. They will help you to redeem the credits, right?

[00:23:21]And yes, having said that, let's begin the workshop quickly.

[00:23:25]So, in this workshop, as I was talking about, we will be building a big query agent catalog, right? Which will deliver or which will have an AI-driven insight through the conversational data analytics, okay?

[00:23:46]A basic prerequisite, of course, you should have an understanding of Google Cloud. Let me zoom this out. Yeah.

[00:23:51]Then, you will learn on how to navigate a big query agent catalog.

[00:23:58]How to create a custom agent. And how I will be making a knowledge base or knowledge source, right?

[00:24:07]That's one of the I think key important aspects of the whole discussion. And and of course, how I can use a Gemini Enterprise to deal with the semantic metadata.

[00:24:18]I will be using an LLM, which is Gemini in this case.

[00:24:21]How to add system instruc- instructions and verified queries.

[00:24:28]So, I will be using not only a knowledge base or or I my agent will not use only a knowledge base to uh do the conversational agent.

[00:24:42]I can have some sort of queries, which will be referred as in knowledge bank or how it behave.

[00:24:53]So, I can add those verified queries, which will be a what I say guide to an agent.

[00:25:00]So it's like an official documentations, right? Or it's like a what do you say?

[00:25:08]Something manual like user manual. So this will be agent manual, right? And then we'll talk about how to publish and share the agent. This will be the last logic what we will be doing. We suggest that you should use a Chrome here, that's the suggestion recommendations.

[00:25:26]Then of course a Google Cloud account and Google Cloud projects projects must be there.

[00:25:32]Basic knowledge of BigQuery and SQL is required.

[00:25:36]So sorry, huh, yeah, of course this is the point and nothing else. So I think let's talk about the next logic. So this is my console by the way.

[00:25:49]This is my GCP console. So I will be doing everything from scratch. So I will go to a select project and I will create a new project here.

[00:25:59]So probably I could say gen AI academy, right?

[00:26:08]So this is my billing account. So I have lot many billing accounts. So I know this is an active billing account.

[00:26:14]Please ignore for that.

[00:26:15]No organization. Of course even while you are creating a project, no need to create a organization. Just create a project name and yeah, this is a project ID.

[00:26:29]Please make sure that you understand project name and project IDs are different. So in some cases we need a project ID, right? Which is immutable in nature by the way. Project name is changeable but project ID cannot be changed, right? That's the point. So let's create the project.

[00:26:50]So it says it's creating.

[00:26:54]Right? And if I say select project, yeah, I will go to that project itself.

[00:27:00]So, make sure you're in the the project what we have created just now, right?

[00:27:05]So, uh yeah, that's what is all about.

[00:27:12]Then, uh if you come to the third step, right? I'm sorry, I'm switching my screen just to make sure uh the Yeah, just to make sure that there will be uh ample of uh uh you know, uh logic between my console and the instructions, right? So, when you uh Yeah, of course, this is very important stuff. So, in GCP, we have an IAM identity access management, where we decide who can do what on which resource.

[00:27:45]Right? So, uh being an owner uh of a project, even we need to add an IAM access called Gemini data analytics data agent owner. This is the role we will be adding to the root user what you log you are logged in as it, right? So, if you if you see my case, if I go to IAM and if I see the first IAM part, yeah, it's it's loading. It's a little slow here.

[00:28:19]Let me close my cloud shell.

[00:28:22]So, if I if I go to IAM and if I see edit principal, so I have to add a rule here. This is what the logic is.

[00:28:32]And the role is Gemini data analytics, so you can even type data here. Gemini data analytics, we need a owner. So, I can search for something called owner here. Yeah.

[00:28:52]Gemini data analytics query data data agent owner, sorry. Data agent Yeah.

[00:29:04]Okay. So, in uh don't get confused. In the code lab, it says beta. Now, it's not beta, it's GA now. So, that's why you will not able to see a beta in the bracket, right? So, I'm selecting Gemini data analytics data agent owner.

[00:29:21]Yeah, and you just need to save it. The best part is you can add some conditions. If you want to assign some time that I want to do do allocate this for specific uh time or specific period, that is also there here. But anyway, we are skipping that step.

[00:29:40]Uh anything I miss? Okay, yeah, that's it.

[00:29:43]So, I'm just updating so I did this this logic, which is a role added as in Gemini data analytics data agent owner.

[00:29:54]After that, once it is added, uh yeah, I have to enable the API. So, you know, by default, data analytics API with Gemini is disabled, which you need to enable it. So, if you go to BigQuery, if I click on this three uh horizontal lines, and if I go to BigQuery, and if I see the agents here, see.

[00:30:14]Yeah, please note as of now, till today, it is in preview. Okay? So, make sure when you use it, you will note or you will keep this in mind that it is in preview, right? So, I'm just clicking on agent. So, uh I think in my case, it will not ask for the cool. So, because I think I disabled the API early. So, it will say enable data analytics API with Gemini. Just enable it.

[00:30:42]So, yeah, Data Analytics API with Gemini, enable Gemini for Google Cloud API.

[00:30:48]You just need to enable both the APIs.

[00:30:53]Yeah, so just to just to make sure I am on the third step, this screen, right?

[00:31:01]Cool. And once it is done, I should able to see Okay, cool.

[00:31:07]So, this is this is the uh screen which explain everything in one go. So, this is where you can create a new agent, see.

[00:31:16]Right? What sort of agent, what sort of source knowledge source I want to add, right? Or any any sample agents you want to use it, right? So, you can see your uh your created agents or any agent draft you have. This screen is itself self-explanatory.

[00:31:34]Uh but yeah, this is uh what we will be following. And now, let's say new agent.

[00:31:39]So, if you compare my code lab, this is what This is where I reached so far, right? Now, we are jumping to the fourth step. We are just creating an agent, right? And yes, we are following these instructions step by step. So, this is my agent name, by the way.

[00:31:59]Agent description.

[00:32:01]As I said, you need to build a knowledge set, right? So, by the way, we will be using uh this international public data set, which is already there with us. So, if you even open this link, it's all about uh Google Trends uh what we have. So, we will be using that data set. Let's see.

[00:32:22]Here, this is the data set what we will be having. So, anyway, so first is agent name, agent description, then we'll add this as in uh knowledge source, right? So, let's do that first. So, let me come to this.

[00:32:39]And I'm just copying the agent name, just doing a copy-paste just to save the time.

[00:32:45]And again copying the description.

[00:32:47]Right? And then it says add source.

[00:32:50]So, what source I will be adding? As I said, I will be adding this BigQuery public data, which is in Google Trend, right? Uh Trends International top terms. I will say send.

[00:33:05]Query here. I got the International top terms data sets.

[00:33:09]I will select it and say add. So, this is uh basically a a knowledge bank. This is basically a knowledge source of my agent, right? So, I will be adding it. Once it is added, yeah, cool. And can you see that here is my agent? So, even if I say if I say something here at this moment, as in as I had not done anything, I have only given agent name and a description, and if I say if I say hi, who are you?

[00:33:50]Let's see what it says.

[00:33:53]So, you can see that Gemini icon icon and which is pulling see, which is pulling the data, see.

[00:34:01]So, it it got an uh knowledge source, but even yeah, of course, this is not enough for our code lab, so I will be basically uh trying to improve the data accuracy by structuring the context, right? So, how to do that? I will click on the customize button here.

[00:34:22]Right? And yeah, see.

[00:34:24]The best part is the entire table description is suggested by Gemini. I could say accept, and I will simply say okay, cool. Go ahead, right? The table contains daily top trend search term for Google Trends for various countries and regions around the world. It provide insights into what people are searching for globally.

[00:34:46]Simple. And then it it got those columns uh sorry, table names, right? And the description of each table name.

[00:34:53]So, this will help my agent to uh get a context. You know, context is always important, right? So, what if we started this whole lot of session directly jumping with the code lab?

[00:35:05]So, we have we need a context setup, right? So, same thing we are doing here.

[00:35:09]And I could basically say, let me see.

[00:35:12]Yeah, if I want I can regenerate the suggestions. Or I can select all and say accept suggestion. Even I do have an option for reject suggestions, right? I can I can make it uh according to my convenience as well.

[00:35:26]But as of now now, I'm relying on what the description is generated. I will simply say accept suggestions.

[00:35:35]Yeah.

[00:35:36]And I could simply say update.

[00:35:38]So, just taking you back, I did this just now. See.

[00:35:44]So, we had uh added a BigQuery data sets. We have accepted the default Gemini description.

[00:35:53]Then we have accepted the tables and their descriptions. Sorry, table's description as of now.

[00:36:00]Right?

[00:36:01]And and if you see uh yeah, that's what we did. And now we need to give So, by the way, just a quick update. If if I So, if you see this, uh let me save this. I will click on save. And if I scroll a little bit, see.

[00:36:21]Uh sorry, here. Instructions.

[00:36:23]It's almost blank. I had not given any sort of instruction to this agent as of now.

[00:36:30]So, what I will do is I have instructions which talks about synonyms, key field, what to exclude, what to filter, and how to deal with joint relationship, right? So, these are my instructions. So, if you see this this is a system instructions.

[00:36:48]Right? What What is your job, what you are doing. So, I'm just copying it and pasting it here in the instruction field. See.

[00:37:00]Then, core model is refresh date, select the daily top 25 terms, week plus score, and historical weekly data. So, all these things in a step-by-step manner uh we had added as an instruction. This is the most important What do you say? Uh uh point to make sure that your agent is working as expected. So, without instruction, your agent will not work appropriately, right? So, yeah, this is what the instructions I copied.

[00:37:35]And uh you know now we have instructions. Now we have knowledge base.

[00:37:41]But I need uh if you remember, we talked about verified queries. I need a manual uh I need a SOP to be added to my uh conversational agent.

[00:37:58]So, how? So, see uh these are these are examples also. These are in in a uh layman terms uh these are known as examples. So, when you say examples, these are uh the what do you say? Procedures to be followed or these This is a source examples what agent will rely on. So, as it is a big query or or we are using conversational analytics as in big query, we will be adding some verified queries. So, if if for example, a user asked that what are the top search terms in UK right now? So, what sort of query my agent should execute is this.

[00:38:48]That's That's the verified query, right?

[00:38:51]So, let me let me see if I can add it now. Okay, before that let me save it.

[00:38:56]And yeah, can you see verified query? It says zero. I could simply say add query.

[00:39:02]What sort of question? So, I could say this is the question.

[00:39:05]What sort of query you want to execute?

[00:39:08]This is the query.

[00:39:09]I could say add.

[00:39:12]Uh sorry, I think I forgot to Even if I if I go to this, I can even say run, which will run the query.

[00:39:20]Query completed, and I could say update.

[00:39:25]Yeah, in the same way I will be adding one more uh query. By the way, if you see run here, I don't know why I don't have an Sorry, I think I missed one thing.

[00:39:39]Uh if you run it, the data uh check.

[00:39:44]The query output. So, the screen was not completely visible. I'm so sorry for that, but yes. This is how the query result is coming.

[00:39:52]Right? So, I can say update.

[00:39:54]I could try to add one more uh what I say question. Right? Let me add it.

[00:40:04]Add query. Another question. So, show the last 12 weeks of uh interest for the current top five terms in Auckland. So, I could copy the SQL query. So, this is Sorry, my bad.

[00:40:19]This is the query.

[00:40:20]So, where you will be dealing with lot of group by, order by, limits, right? Some where conditions and all that. So, I could say add. Oh, sorry. Again, I I forgot to run it. No worries.

[00:40:34]I could I could edit and see if it works.

[00:40:39]Yeah, it works.

[00:40:42]Right? So, let me update this query.

[00:40:45]And let me save this again.

[00:40:48]So, yeah, this is what we added as in verified queries.

[00:40:55]Correct?

[00:40:57]And yeah, even if I scroll below, here it says view Gemini generated suggestions, right? So, if I say it is telling one more suggestion saying that identify specific search term executed an autonomous term popularity index, which will be calculating in a given week. So, I could say, "Okay, fine. No worries. I will add this also."

[00:41:24]So, these are the three are the three verified queries what I have been added so far.

[00:41:30]Right? So, I did this.

[00:41:33]And cool. I think that's it. So, what we did is a quick revision or maybe a quick revisiting what we did. We gave agent name. We gave a description.

[00:41:45]Uh we added some knowledge source. Then instructions.

[00:41:49]>> [clears throat] >> Then we added some verified queries.

[00:41:54]Right? That's the best part of uh the current scenario.

[00:41:58]Uh I think, yeah, we can even talk about glossary, right? If you see, glossary is basically uh by any chance if you're using Data Plex, right? Uh for those who are not aware of what exactly a data place is uh uh data place is I think it's a old term, right? If you see uh now it is known as knowledge catalog.

[00:42:24]Right? Uh which is an AI-powered uh uh uh data source from a distributed uh scenario, right? So yeah. So uh what I can do is I can add a glossary to my knowledge catalog, right? So maybe like an index or maybe like a metadata here.

[00:42:44]So I will copy this.

[00:42:46]And yeah, I could say glossary term, refresh data. What is the definition of this glossary is this.

[00:42:54]Uh let me remove that double quotes.

[00:42:56]Sorry.

[00:42:58]Uh it says correct. No problem.

[00:43:00]Then synonyms, what sort of synonyms?

[00:43:05]Yeah.

[00:43:06]So this is my agent catalog. Uh that's what I was saying it is also known as knowledge catalog, right? That's the best part of the uh logic. And yeah, I think add and then save. So this is my glossary which has been added. Right?

[00:43:30]Cool.

[00:43:32]So I completed till this step.

[00:43:36]Right?

[00:43:40]Yeah, then we have one more logic called labels. Labels are the key-value pair uh to organize the uh Google Cloud resource in the logical grouping, right? Uh as of now I'm not talking or dealing with labels. I'm just changing this maximum bytes built to this value and let's save this.

[00:44:01]That's it.

[00:44:03]So we will be using this value which is in query uses per data. It's like a quota.

[00:44:09]Of course, because I have to do a capacity planning and again a cost center should not be impacted, right? So, that's the reason I can fine-tune this. This is a bytes maximum bytes billed per query, right?

[00:44:24]That's what we gave here.

[00:44:26]Yeah, so we reached till the fourth uh step which we completed so far, right?

[00:44:35]Cool.

[00:44:36]Now, uh yeah, before we go to the next step, can I talk to the agent?

[00:44:52]Let's see. I I'm doing this randomly.

[00:44:54]Let me uh check if uh the whole lot of configuration we did correctly or not.

[00:45:02]And if that in case, this should work without an error.

[00:45:09]Yeah.

[00:45:16]Great. So, uh yeah, we were able to publish uh the agent just now, right? And now, if I want to share, I can share uh using a Looker Studio if someone wants to access or maybe if someone wants to access using a BigQuery, I can do that, right?

[00:45:32]Now, let's try to create a conversation here. So, here, right? So, maybe uh I am on the sixth step now, where I'm talking about how to deal with the conversation. So, I could say I will copy this and I could ask a question saying that, "List on top 10 terms in England. How did they trends for past 3 months?" Let's see what it does.

[00:45:56]So, it is basically doing the reasoning.

[00:45:59]Uh it is checking the summary. It is generating SQL query, right? You can see the data result. So, I I think Oh, sorry. It says the agent in the either deleted. Ah.

[00:46:11]My bad.

[00:46:12]One second.

[00:46:19]Let me see if I can do it here.

[00:46:23]Ah, yes. So, you can you can see it's uh analyzing now. So, probably I did a small mistake there.

[00:46:31]And once that is done, you could see Yeah, it's doing You can see a uh It's doing Let's wait. Ah, cool. See.

[00:46:51]Right? So, you can see the chart, you can see the table, right? You can even get the details of it, right?

[00:46:59]So, yep.

[00:47:01]I don't know why the conversation is not This is the Yeah, let me see if I am doing anything wrong here.

[00:47:12]One second. Did I Yeah, I published it earlier also, but not sure why it's not working here.

[00:47:25]So, yes, guys. Uh Yeah. So, I was facing one error when I went to a conversation uh stuff, it was giving some error that uh unable to connect. So, uh the issue was very simple. I have to give couple of uh roles, couple of uh uh I am roles additionally to this. So, one of the uh role what we have to give is uh the Cloud AI companion user.

[00:47:54]Right? So, by any chance, if you see the error and you will not able to see a data here on the right hand side, make sure you assign a role called here.

[00:48:05]Cloud AI companion user and I think this if it it is not added, add this as well.

[00:48:10]So, these are the two roles which we need to add which I did and it is working now. So, if I gave the same question, so I was giving that based on the top 10 terms in England, how did they trend in the past last 3 months?

[00:48:24]So, I can see this is the query. Can you see that, right?

[00:48:27]And this is the answer. Cool.

[00:48:31]Fine? So, this is this sort of output you can expect from this conversation what we have. The same step is given here which is our sixth step, right?

[00:48:42]So, yeah. Even even if you can you can come to the agent and directly ask probably here, can you predict and visualize? So, for example, if I want to here, I can directly push the question.

[00:48:58]How Monopoly board will trend in the next 4 weeks, right?

[00:49:07]Yeah, here. So, it's doing a forecast.

[00:49:08]You will see a graph also if I am not wrong.

[00:49:20]Let's see how it goes.

[00:49:29]So, I think uh it's basically uh doing all the analytics in a back end and then it will predict Yeah, see.

[00:49:41]And even it will visualize uh the stuff as well, right? Let's see how it goes.

[00:50:10]>> It's taking more time. I don't know if everything is uh it's thinking, yeah.

[00:50:15]You can see that.

[00:50:17]So, I Yeah, no worries. I think it's working.

[00:50:21]I thought it is not working, yeah. It's doing.

[00:50:30]So, just for meanwhile, just for your information, we could able to complete our sixth step, yeah. We are expecting some sort of this result so far.

[00:50:41]Let's see what happens.

[00:50:46]It's doing the work so far. Let's see.

[00:51:02]It's not yet done.

[00:51:10]It's taking bit long because it's basically checking the next 4-week trends, right? That's what the point is.

[00:51:23]Yeah, it's finally done. See.

[00:51:26]Right?

[00:51:32]So, I can even see can you forecast the other stuff?

[00:51:36]Top trend trending games in the Philippines right now. So, it's suggesting some questions, right? So, that's what I can easily do. I can even go and create a new conversation if I need, right? So, yeah. In a nutshell, this is how the agent our conversational agent with BigQuery works.

[00:51:54]And if you if you head towards the last step now, how you explore the agent catalog, this is what we have.

[00:52:03]So, if I go to agent Yeah, these are my different agents.

[00:52:11]Yeah, this is the agent which I had created and these are the different sample agents what we have, right?

[00:52:17]I think yeah, that's it in terms of the current codelab.

[00:52:23]So, we could able to complete it a little little early, right?

[00:52:27]And this concludes that we were able to build a conversational analytics data agent successfully.

[00:52:36]Right? The only one additional thing what was missing is the role. I hope you got the point. Uh the maybe by any chance if you're getting the error while doing a conversation, make sure you have a data agent user role and the Cloud AI companion user role is added.

[00:52:55]The one which I showed you from the documentations, right?

[00:53:00]Yeah, I think that's all from my side, guys. Thank you so much for listening to me patiently.

[00:53:13]Okay.

[00:53:14]Thank you so much once again, Mr. Ashutosh, uh for an excellent session and thank you to everyone for joining us today. I hope uh you like the session.

[00:53:23]And if you have any questions, you can just drop in the comment section and we'll reach out to you as soon as possible. And if you also uh want to have some other queries related to the session or something, I'm just adding a ticker here so that you can just reach out to this email address and then our team will uh resolve your issue as soon as possible. So, I guess that is it. And once again, thank you so much, Mr. Mr. Ashutosh, and for this session. I hope everyone liked it. Okay, so from now onwards, let me give you a little bit of instructions of how you have to perform the code labs and the entire journey of the session.

[00:54:06]So, uh we like we hope that this workshop gave you a clear view of how conversational analytics and big query agents are transforming the way developers uh interact with data.

[00:54:18]Uh as you have seen, uh like Cohort 2 has already entered around specific term that is unified data analytics and intelligence. Uh this is a theme that is designed to help you move beyond traditional dashboards and manual analysis toward intelligent agent-driven systems that can understand, analyze, and act on data in real time.

[00:54:39]Uh the journey starts on the participant dashboard. As you have already registered there, you just have to go to the dashboard, which will serve as your central workspace throughout the program. And this is where you'll find your session recordings. Because uh right after that, you can find this live session as uh on demand on your platform on the interaction session. There you'll find the recording, uh your learning modules, code labs, assessments, and all important updates related to your selected track. And you should already have received the credits and access instructions uh to complete the uh hands-on code labs. And we strongly encourage you to work through the lab while the concepts are still fresh, as it it will help you put today's learning into practice in a structured and practical way. Uh in addition to the core curriculum, you will also have access to optional Google Skills Lab, which offers deeper exploration and additional hands-on practice to strengthen your understanding of the technologies covered in the the And once you complete the workshops and code labs of your selected track, make sure if uh support professionals, I just wanted to let you know that it is mandatory to complete one track, but if you want to uh go with the track two, that is totally up to your choice. Uh you can do that. And once you complete the selected track, you'll be able to attend the MCQ assessment, which serves as the final milestone of the academy and validates your understanding of the concepts covered throughout the cohort.

[00:56:03]And participate participants who complete all required components in at least one track will earn the GenAI master certification. I'm telling you again, in order to get the GenAI master certification, you have to uh complete the entire academy that consists of code labs, workshops, MCQ assessments, and prototype submission.

[00:56:24]And of course, professionals may choose com- to complete one or both professional tracks, while students will follow the dedicated student track.

[00:56:33]After completing the academy, you'll advance to the cohort hackathon, where you'll collaborate with fellow participants to solve a common problem statement that will be again available on the uh platform.

[00:56:45]And it will build innovative and real-world solutions using Google Cloud technologies. And all graduates will be invited to the graduation ceremony where top performers will be recognized and celebrated for their achievements.

[00:56:57]Participants who excel in both the academy and the hackathon and rank among the top 100 performers will earn an invitation to the GenAI Elite Club, our exclusive alumni community of outstanding developers and innovators across the ecosystem. So, if you have any questions whether technical or program related, please reach out through our Discord community or by email, which is mentioned in the ticker.

[00:57:18]And also, the Discord channel link is there on the platform. Once again, thank you so much for the session. I hope you liked it.

[00:57:25]Uh my name is Manav Ahuja, and I'm your host. Thank you so much. Bye-bye.

[00:57:32]>> I see you.

[00:57:43]I see