Genie & AI/BI: Conversational Analytics Explained

0:00[Music] hi Miranda hey youf how's it going all good all good so I think you're the start of the day your demo was super awesome I was in the audience people like shouting not only about aii dashboard about the messaging but also aivi Genie so please can you tell us more about how we can position aibi dashboard and gen as well and also the

0:23aibi experience yeah absolutely well first we're so excited to finally kind of pull the covers back on a IBI for everybody we're very excited to see what everybody does um but really our vision there is to make sure that the data that's in Databricks that you know for years data engineers and data scientists have been working so diligently on now those insights based on that data becomes accessible to everybody in the

0:45organization in a really easy and accurate way right so one of the things you know I'm sure that I want to make sure to emphasize is that's nice about aibi versus maybe some other things is we're able to actually flow through context from the entire platform so I mentioned those data Engineers who have been doing running jobs for probably 10 years now at your company all the context that's in those jobs it's in

1:07those notebooks flows through to the dashboards flows through to aib gen and you know with dashboards we definitely you know first of all I want to say we partner with tons we have a ton of amazing Partners we're going to keep having a ton of amazing Partners um and we're actually looking forward to more and more integration points and letting them leverage some of these pieces too but you know we do understand that sometimes there's a need to have the data visualization derive an Insight

1:30right next s of your data we want to make sure we're giving an option for that so with dashboards we want to make sure we're kind of providing all of those bi capabilities that you rely on whether it's you know quick uh assistant uh like the AI assistant for creating data sets or creating visualizations cross filtering embedding in Salesforce Confluence SharePoint sharing to users who aren't members of the workspace that

1:53are maybe not needing to see anything besides dashboards like we're making big Investments cross filtering I showed off today uh in all of those areas to make sure there's a really robust option when it's needed now what we're really excited about is having kind of Genie come online as a complement to those dashboards so while of course you can use genie on its own uh we introduced today a flow in moving from dashboards to Genie it brings all the great context

2:17from the dashboard along with it right and really accelerates that kind of setup process so you don't necessarily have to have a data expert configure every Genie space to be usable for the business user right there's already some work that a data expert has done in building a dashboard normally so now what we can do is kind of move all that context forward into a new more conversational type of interface so that those business users you know aren't

2:40just stuck at I saw these 10 insights that's great but I have this like five follow-up questions they don't have to go find a data expert and ask them 100 million follow-ups they can just go right into Genie ask their question get their answer if there's something Genie doesn't quite understand they can provide some additional clarification and if they're getting kind of the response back they want they can save that if they share that that space with others they'll be able to benefit from that same context so you know the idea

3:04idea with Genie is no dead ends right we never want an analysis to kind of stop because you need to go update a semantic model go ask a a data expert for more help we want it to be kind of a always open doors policy yeah and what I have noticed like have some customers been trying jie for quite time and what they did is like they created I think four five gen spaces one for finance one for

3:26the marketing and then each one with the specific instruction make sure that it's seller to their neede like fine T in but in a let's say in easy way yeah well exactly that just kind of having things that are tightly related because you can get more Rich detail right you can imagine that if you're saying hey search every piece of data ever it's really not as helpful as saying hey these are like really the curated sets that are relevant to marketing questions are

3:48relevant to finance questions or relevant to manufacturing line one two three questions right you want to really make sure you're kind of curating the information that's useful and then Genie helps just navigate it right we don't want G to kind of be uh making things up or looking at too many different uh tables that have nothing to really do with the context of the questions that would be coming from those users so exactly what you have your customers doing which is set up a few different

4:11spaces each specific to a topic each with some focused tables and instructions you're going to get a lot more accuracy and a lot quicker response types and also something important like with UCC integration you can have this end to end lineage from the table from the inje layer till the transformation then which queries was used in which dashboard and this is super super amazing absolutely one more question for

4:33you for folks who want to start getting started quickly with Jenny like if it's not activated is there a specific trick they can do yes okay so Genie is available to everybody on Azure in AWS right now couple things you got to have set up you see thrower serverless warehouses and then make sure that you go to the new previews panel we're very excited about that we heard a lot of feedback from folks that it's hard to know what's in preview so you go up to

4:56the top right as an admin previews panel enable Genie and then you'll be off and running and Genie can be using that workspace that's awesome thank you Miranda gcp coming soon gcp coming soono all right amazing hello hello good to see you such how you walk on the aii yep tools that we have at air bricks

5:18that's what we're calling it now yep amazing can you can you talk about it and you know tell us what you do sure um I work on uh dashboards and visualizations and more recently G so the common theme is Databricks has all this great data all these great tools we want to make it accessible to everybody not just people who know how to code who know how to write SQL everybody so

5:40that's what dashboards are for that's what Genie is for and hopefully we can make a lot more people use data for good decisions so where where can we access this aibi thing using the assistant using the Leg view dashboards um aibi is now the brand for lak dashboards formerly known as lak dashboards now okay so now all dashboards are part of aii now okay um Genie is part of AI so

6:06you can actually use both of those products today um it's just a rebranding to those things and we're probably going to refactor it out into its own section that says data bra is doing bi now so so what's the magic Source behind how does it work like why how can it you know do the text to SQL yeah um I think a lot of this is there's good foundation models out there your gpts your llamas Etc

6:30but then on top of that we have really good sort of training data and really good Benchmark data to make sure that we're generating the right um SQL out of it by default for the simple stot and then we ask uh our customers to do some of the leg work by teaching us teaching the genie uh what their concepts mean what their SQL statements look like what

6:53their joints look like Etc so that we can learn from that and do even better so you learn from the actual data you learn from the existing SQL request and that's what use they can also add so they can add like comments and instruction to the Gen spaces yeah so we encourage them to just start using it see what works what doesn't whenever something breaks they can add more instructions it becomes a Dev Loop

7:15basically well that's amazing all right so AI bi AI for bi for everybody that's right that's what we're going for text and you get the bi amazing very cool thank you thank you very much