Skip to main content

AI and Genie: Analyzing Healthcare Improvement Opportunities

Summary

  • Healthcare Data Insights: Premier Inc. uses AI/BI Genie with risk-adjusted patient data from 1,300+ member hospitals to discover insights that improve care and positively impact lives.
  • Rapid Genie Deployment: Premier's team stood up a Genie instance in three days, bypassing costs and timelines of custom modeling and application development.
  • SQL Generation Acceleration: Genie enables Premier's internal teams to generate complex SQL 10 times faster than writing by hand, supporting plans to reach 20,000 users across hundreds of hospitals.

AI and Genie: Analyzing Healthcare Improvement Opportunities

Watch: AI and Genie: Analyzing Healthcare Improvement Opportunities
Talk by Jay Sharma, Software Architect, Premier Inc ; Tim Riddle, Senior Director, Analytics, Premier Inc
Improving healthcare impacts us all. We highlight how Premier Inc. took risk-adjusted patient data from more than 1,300 member hospitals across America, applying a natural language interface using AI/BI Genie, allowing our users to discover new insights. The stakes are high, new insights surfaced represent potential care improvement and lives positively impacted. Using Genie and our AI-ready data in Unity Catalog, our team was able to stand up a Genie instance in three short days, bypassing costs and time of custom modeling and application development. Additionally, Genie allowed our internal teams to generate complex SQL, as much as 10 times faster than writing it by hand. As Genie and lakehouse apps continue to advance rapidly, we are excited to leverage these features by introducing Genie to as many as 20,000 users across hundreds of hospitals. This will support our members’ ongoing mission to enhance the care they provide to the communities they serve.
Intelligent analytics for everyone: https://www.databricks.com/product/business-intelligence
BI meets AI: https://www.databricks.com/resources/ebook/business-intelligence-meets-ai
Explore the only unified platform for agent systems: https://www.databricks.com/product/artificial-intelligence
See all the product announcements from Data + AI Summit: https://www.databricks.com/events/dataaisummit-2025-announcements

Chapters

FAQs

How does Genie help Premier analyze healthcare data?

Genie provides a natural language interface that allows Premier to analyze risk-adjusted patient data from over 1,300 member hospitals across America. This enables discovery of new insights that represent potential care improvement and positively impact patient lives.

Why did Premier choose Genie over text-to-SQL approaches?

Premier previously experimented with text-to-SQL RAGbots but found Genie superior. With Genie and AI-ready data in Unity Catalog, the team could stand up an instance in three days, bypassing the costs and time of custom modeling and application development.

What performance improvements has Genie delivered for Premier?

Genie allows Premier's internal teams to generate complex SQL approximately 10 times faster than writing it by hand. This significant speed improvement enables the team to stand up analytics more quickly and serve their growing user base.

How will Premier expand Genie access to its users?

Premier plans to introduce Genie to as many as 20,000 users across hundreds of hospitals. As Genie and lakehouse apps continue to advance rapidly, Premier is excited to leverage these features to support their members' ongoing mission to enhance care.

Full transcript

0:00 Introduction and agenda

Okay, good morning everyone. Thanks for joining. Um, we're going to kick things off super quick. Just uh a housekeeping item. We're going to hold questions till the end of the session. Um, and then as people have their questions, feel free to go up to the mic at the end of the session. There's one on each side and uh you'll be able to ask. Um, but want to introduce our speakers today. We have Tim Riddle and to speak to us today.
And um without further ado, I'm going to pass it over to them. So, thank you for joining. Hello everyone. Oh, talk lower so I don't blow you away. Uh I'm be kicking it off and I have my colleague here. He's be going to be doing a live demo. So, we're risking our lives here uh with Genie as well. So, let's get started. Um, we left our QR codes because we're not going to be able to get into everything, but if anybody wants to hit us up with questions later, we welcome that.
Um, we just want to share our journey with Genie and how we've done so far. Um, we start with an introduction. So, who is Premiere? I think it's always useful to kind of understand our specific use case and how it could be relevant to your own. I know a summary of our performance, our data, a little bit more about our data. Um, since we're mainly, many of us are data and AI people. Um why did we choose Genie? Um for a couple years we did of course text to SQL ragbots.

1:03 Company overview and mission

Everybody was trying those out earlier and why did we switch to Genie? Um and then going to like I said go over a real live use case with our data. It's healthcare data so it looks real but it's not. Just full disclosure and then a recap of our results. So just get started. Um, all of you all have seen this and um, remarking I'm in Chicago land, right? But I I I don't think there was a billboard that I didn't pass on the way from the airport that wasn't talking about aenic AI and AI in general.
Um, you know, any of us that have lived through like the internet boom or or other stuff know that the Gen AI stuff has hit remarkably fast. um and is beating most humans at certain swivel tier tasks, right? We all know this and this is just relevant for why we picked Genie and why we're thinking about as a SAS provider doing stuff a little different, right? So, we'll get into what Premier is, but um over twothirds of healthcare providers in America use one of our SAS solutions to help improve their care.
Um second kind of macro trend that if you you all are following is that um not even sure if there's going to be product managers aurora I'm a classically chain train trained product manager it's just really interesting as things are evolving do you need product managers I think the answer is yes but I'm biased and the third thing in our industry right is the uptick in what does SAS really mean and I just throw those three things out because that's kind of what we're trying to evolve at Premier here.
Um, we want to help our healthcare providers get to insights faster. But what does that mean when SAS is no longer a thing, right? So, what is Premiere? So, Premiere uh helps over twothirds of health care providers should think health systems, hospitals, doctors in America do their thing. Um, we also help them buy supplies for many of the pharm techch industries. So about $85 billion in pharmaceutical supplies come through our contracts.
We help them coordinate contracts with big suppliers and with that comes a lot of data that we capture in the those solutions to help them. About one out of every two people that go to the hos the hospital in America that are discharged are in our solutions. Of course we have rules on how we use that data. Um ma mainly though we're serving it back to them in a more cleansed fashion so they can help improve their care.
Um, and we have a ton of outpatient business as well. So with that, we try to look at all that data and give them insights into their bigger picture of how they deliver care in America, right? And we also a lot of times will partner with newer age organizations to bring those technologies into our portfolio to help organizations scale again to improve their care. It's all about improving the health of our communities in America.
Um, one level down from that, I just want to talk a little bit about the data that we used in this example. Um, we get billing data. So, if you've ever been seen a hospital doctor bill, I'm sure all of you have, we get that data from our healthcare organizations. We clean and standardize it. And that's kind of the thing on the left. We call it quality enterprise. But basically, it's a way to just take hospital billing data and normalize it for them and give them an apples to apples comparison.

4:26 Data strategy and AI readiness

Fun fact, we also make disease outbreak software. So that was a popular and hot topic that we weren't expecting during the COVID era. Um about a thousand hospitals, we we do real-time clinical surveillance for about a thousand hospitals in America, about one out of every five. So we take real-time clinical messages and look for where outbreaks might occur in the hospital. Um, so there we're really looking at the pharmacy, the infection, the laboratory results of people that are staying in the hospital or the outpatient setting.
All that data comes into a pot and we have a thing that has AI ready data in it. That's what we're using in our demo today. It's called insights is the product name. But really, if you think about it, this is common among all of us, right? 80% of the effort to get to where we show in Genie today is the stuff that nobody wants to talk about. It's the plumbing. Right. It's how did you get Unity catalog right? What was your descriptions?
Right. So, we're going to make it look easy, but I just want to call out behind the scenes. We were working for four years not knowing Genie was coming, just knowing we needed to get our house in order. So, a lot of wars around that. But at the end, we got a a data as a service product. So, that's one way we've evolved. So even as our SAS solutions might change, AI ready data is important and it helps our organizations deliver better care by using that in different things along with their other stuff that they have.
Um and that's referenced here at the bottom to you know 40% of organizations don't feel like their data is AI ready. You see a lot of ads if you go to LinkedIn every day somebody's bombarding me with is your data AI ready? And that's the four years of work we did before we kind of plugged in Genie is to make sure that we had a good Unity catalog metadata store, make sure our data talked to each other. And I'm not saying that's easy, but I'm saying that makes it a lot easier when you get to implement Genie.
So if you've done some of that work, it definitely helps. Now, let's talk a little bit about a use case that going to demo. So I don't know how many of y'all have been to the hospital not feeling well. Um but as many as one in five people that are admitted to the hospital for something pneumonia whatever it is come back within 30 days. And one study to make it a little more missionary is 30% of all children that get admit to the hospital get bounced back and they could have been prevented.

6:49 Addressing healthcare challenges

Now I don't know about you that keeps me up in the morning as far as like making sure kids don't get sick and come back to the hospital. Now we serve our healthcare organizations and all of them are always trying to do the best. So it's not like this is things that they want to happen. What we do in a second I'll show you is we risk adjust that data so they can show opportunities to improve their care where it did happen.
Like how do they close that gap, right? And this is how we do it. This is an eye chart I know but basically we take those one out of every two discharges we have in America and we we risk adjust it. Right now, you're probably asking, why didn't we go use the fanciest LLM? And the reason right now is because our healthcare organizations want explanability in their models. So, we're still doing a model. We're trying LMS, but we're still doing a model.
We can explain away all our variables. In this example, the patient had cancer. And that increases their risk depending on the outcome, a specific amount. Anyways, what we do is we calibrate all of our data. We roll it up and we get to a model that says did the patient have a readmission and what do we expect the likelihood of the them having that readmission was right so you know imagine a one they were readmitted imagine a 20% chance that they would have been readmitted you roll that up you can see your service lines your hospitals your doctors you can look for performance improvement opportunities and we got a billion reports that they could look at but we wanted a different way because Genai the one thing about Genie is it's allowing us to have people that are non-technical, not just pull reports, but actually have a conversation with their data and that's what we wanted to do.
So, we've had some successes with this data as a service, right? There's some success stories because we want to plug ourselves shamelessly, but basically we've done a lot of work with data as a service to help our healthcare organizations automate their manual work already that had been done over the last couple years. But really, we wanted to think about what could we do next. So with 60 plus tables, 2100 columns of different data, 2500 different measures for patients and billions of records, it gets kind of hard for all of our healthcare organizations that are already strapped to figure out what their opportunities are.
So that's where Genie came in. And we we originally started doing this ourselves, right? For a couple years, we were doing text to SQL, rag bots, we were trying to figure out and then Genie came along. I'm like, great, I don't have to do that. I don't have to try to out innovate where my download to Excel button is. Data bricks is going to take care of that for me and I'm going to get to focus more on how do I improve my data which was nice.
We want simple, right? A lot of times we make things a little complicated in tech world. We want to just we want to get to production. I'll talk a little bit more about in a second. and it has all the governance and security that keeps us up at night so that we don't have a list of HIV positive patients that get released. So that was the other thing that was big. So really just before I wrap up and give it to kind of some of the again the highlights of the benefits of Genie, have any of y'all heard of the Voyager probe?
To get super geeky the Voyager probe we said okay so a while back maybe about a year ago I started describing that to people in my organization. And it's like we want to use data bricks as a gravity assist, right? We don't want to out innovate data bricks. We want to innovate with data bricks for our partners. We're a midcap company. I mean, sure, we like to think we're better than we are, but we're a couple billion dollars in revenue.
We can't innovate everything, but if we can use partners and use Genie, then we can act like Voyager Probe. We can slingshot ourselves, right? We can move faster. And so we we we're we're doing this already internally with our case study. We unleashed Genie for our our our business users so they can have a conversation with their data and our advisory team that helps our organizations improve their care that I talked about.
It's a reusable UI, a transparent SQL. Um, and we'll get into the results at the very end, but um, we really road tested Genie and got 100% accuracy. A lot of that I pushed back to the four years war we had civil war to combine all our data in unity catalog and have the metadata and stuff. So that was very helpful. Um and it can continuously learn and kind of via feedback loops that we'll show in the demo and we'll talk about a little bit more.
But with that I want to turn it over to talk about how we configured Genie and then to give you guys a live demo of what we did. Thank you. Check. Yep. Thanks. Hey guys. Um before I I take you to the demo of Genie and show you how Genie responds to some of the healthcare questions that I'm going to ask, I just want to walk you through uh some of the steps that we performed in order for Genie to be accurate and precise in the response.

11:38 Configuring Genie

And like any other like uh rag based architecture, there are three basic steps that you have to do. You have to configure, you have to operate and then evaluate while you're operating and then go back to optimize again. And that's what we did with Genie as well. The optimization involved here. Um, basically Genie gives you three handles and we will um, you know, we will show you all those. Um, optimization process is is very simple.
You have to provide enough context or enough information for Genie to be able to process your business. So if somebody is asking a business related question, Genie is able to interpret that and convert that into accurate SQL and then go and fetch data for the you know the person who is asking uh the question. um our data h u basically we have a um a driver table where we have pre-calculated measures thousands of them um one just my colleague just talked about was readmission you go to hospital and you don't want to come back um there are thousands of measures like diabetes sepsis and everything and if you want to uh learn you want to access that measure you have to understand what the code for that measure is That's the identifier and then that measure table spreads out and connects to other tables if you want more information about the patient who's part of that measure.
So the challenging part for us here was to uh provide enough context to Genie so Genie can take a question from the user and then translate that question into the correct measure code and then you know go to the correct tables to pull all the information that is being asked by the user. So the first thing what we had to do for this to be for Genie to be efficient was to make sure that we have um the tables and the columns that Genie has access to to me meet that objective right is clearly defined.
And here the definitions that we are uh we would want to put here is um why this table is there and um if if you could describe the content of the table. Now I'm not talking about uh some obvious things like if the if the if the variable is says gender column and then the values are f and m there is clear that f female and genie and llm they understand that but if you have very specific values say it says gender description but the values are zeros and ones which is which genie doesn't know that so you have to you know provide those here so and uh you know uh data bricks provides It's the AI uh assistant if you would like it can access your data and give you a nice description that you can you use that for table.
So we put that all in unity catalog and once we are done with defining all the important columns and tables we uh utilized uh the second uh handle which is the instructions. Now instructions are you can you you can provide your business um defaults there. So like for example you know in healthcare there are so many dates you have discharge date you have admission date you have birth date you have death date event date all kind of date but say if you're interacting with the with genie your genie and somebody just says give me all the patients for 2022 now which date are we talking about like it could be any of those dates so I instructed genie that when somebody doesn't specify what date go ahead and use discharge date so those defaults are set here and then you can see uh you know I have I've put some instructions as to how to uh uh you know decode the measure code uh from the instruction set.
So there's this a measure code readmission hwecd as a measure code and the instructions are there for genie to when somebody asks a question and I will show you when we go to the demo how it does that and you know we also added some business rules. So we we provided all the instructions here and the last uh and very critical uh uh handle here is the your SQL examples. So you don't have to like anticipate every single question and put SQL for that question.
A genie is smart enough. You just give like a one qu one SQL example and it can tweak it to adjust around it. So the idea here is that what we what we uh you know in my personal experience our it was that you have to give as many variety as possible. So it has genie has the context and remember the instructions that we were talking about is pretty heavy like it it has some limitations. But if you put a lot of instructions um it may perform slower.
So you have to balance between the SQL examples uh functions that you can also put and the instructions. So once we were done with uh you know the optimization steps then uh the next thing was to evaluate and operate which basically kind of go hand inand um also so Genie also has options where you can automate the evaluation. Um one of that option is benchmarks. So here you can see we we ran some evaluations there. Uh here you can uh put the questions and then put the right answers and then run it in automation fashion and Genie would go ahead and generate the answers compare that with the benchmark answers and we'll tell you if it failed like you know what percentage or what's the accuracy was.
So you can um then further go and tweak and optimize uh you know your instructions more examples maybe. And then we also use uh uh good old-fashioned QA testing here. Um we evaluated the output against the known reports making sure Genie is uh accurate and um you know and then then we have uh this monitoring activity that Genie it's available in Genie space and what it lets you do is it lets it gives you a view if you expose Genie to external users all the that interactions were is documented and you can see that a user has a has an option to rate genie's response and based on this good or bad rating it gives you an opportunity to further tweak and make genie further you know improve the accuracy of Genie's output.
So um these are the steps that we performed to make Genie work. And now I'm going to take you uh to the demonstration. But before I take you to demonstration, I I will have to promote myself to uh vice president of quality uh in a hospital. If you some of you have been to a hospital or work for a hospital, you understand the responsibility that role comes with. Um so I'm going to uh share Genie here. I'm going to start the warehouse really quick.

19:07 Live demo of Genie

We didn't want to wanted this to be timed out. So, um this is Genie's interface. This is our opportunity analyzer. You can ask questions here. It's very clean in need here. Uh tidy um I would say very intuitive. So, as a uh VP of quality advisor, I have I have saved this question here. don't want to see me typing with uh with the spelling mistakes. Maybe Genie can come up with some sort of spell check. That would be something uh you know for future.
But um what I'm going to do here as a you know I as a VP of hospital I I I'm responsible for quality of the patient. So in readmission is is one example. We don't want a lot of patients to be readmitted and I want to see what's going on in terms of readmission. So um I can you know it's going to be a conversation with Zeni. So I want to zoom in and I want to find out uh in terms of opportunity readmission opportunity which are uh this service lines that I need to you know uh focus on.
So now you can see here as Genie is working through the question there are things that happened here. The first thing that you can see here, which we just I just love this is a English version of what Genie just is is trying to do. You can go through this, but I don't have to because I know that the measure codes that was generated is correct. It's right here. This is the highlight here in the right version. So here I'm asking Genie to give me the opportunity for 2022.
I did not say discharge date. So genie is supposed to take the discharge date here and then I said sort that opportunity by service line and it clearly tells me here that general medicine has the highest opportunity. Few other things over here as you can see um there is if you are if you want to make sure 100% sure uh that whether the code is right or wrong you can actually see the code in the back end that Genie generates.
Um I have I have seen this many times. This is the right calculation. But you see how we have we are here uh generating the measure code based on uh somebody says hospitalwide readmission. This measure code gets uh gets generated and then the version here is um you know V4 2023 and then you can see the discharge date which is default for the year 2022 and this gives you the answer. The other thing that I really like here is you can download if you want to you know share in a presentation you have an option to either download this or you can also copy the data in a as a CSV if you have a you know larger data set you can do that as well you can regenerate a response if you think that something else needs to be you know tweaked.
So these options are already available and then this is the this is the icing on the cake. You get the visualization as well and when uh when we started doing this and since then there are a lot more features uh available here. Now you know if you um if you tell Genie um to uh uh like say so for example if I if I uh come here and see if there is a visualization option the bar is default but you can you have option to choose other vis visualization as well if you want to take a screenshot for presentation or anything like that you can do that uh you can also um you know uh sort this if you if you like there's a sort option available and then take the screenshot so This is really cool and it keeps getting better uh you know and um it's just we like it because uh you it's so easy to take a screenshot and just go to the presentation.
I always like to you know pat Genie on back if it does a good job here. This is where the feedback is and that's so important. Uh these feedbacks helps you know to improve the query uh further. All right. So now that I know that uh general med medicine is where I need to focus on I want to you know ask few other questions though however like for example the service line is a big thing right but which particular disease should we focus on within that service line right so let let's ask that question with genie and um I think I got it here it's basically pre-loaded question I'm just saying you give me uh show me the opportunity for a particular disease for general service line and see what happens here.
So uh now Genie is processing that and thinking about it. Um we are hopefully we'll get we will get to see the disease that is causing this our readmission to um in general medicine service line to go high. Now if you if you look at this there are few things that you can um you know you can appreciate here. I did not have to mention readmission. Okay, I just said opportunity and Genie has that context that the conversational context we are talking about readmission.
Then obviously the I'm following up on readmission. So it generated the code based on the conversational effect and then the right version here and I I said genie to give me general medicine. So that was right here. So we got that and again all those panels exactly the same. uh you can download see the code. Um the thing I want to focus on here is um septasemia because septacmia has the highest opportunity and it is also one of the leading cause you know of death in in hospital.
So we we want to focus on septasemia and I want to I want to know which facilities have high return because of sepsis you know the the patients are coming back. I'm just going to copy. Oh, well, you know, I just already have a pre-loaded question, so I was going to type it up and save you some time. Um, and I'm going to just copy this question again for you. And let's um let's see if Genie can tell me which facilities which which of the hospital are doing poorly in terms of septicmia based readmission.
Right. And um and you can see it's it's strikingly fast like it takes like what three 10 sec some seconds to convert that um you know u that question into SQL and then go get the data. There's a lot of work that's happening uh you know behind and we didn't have to um train or or or create a rag based architecture did the vector search put all the documentation it's like a lot of work to do that I mean if you are involved with this like I am you would understand that the work effort for us to get Genie here was a lot lesser than And if we had if we had to do the rag system from scratch.
So based on this um you can see that South Premier General uh this is the facility where um we have we are seeing highest readmission in terms of septic for septicmia. Um and there are other hospitals too that if you want to you know focus you you can and so they're like you have about all the 15 facilities uh listed and you can see um that there are some ones some hospitals that have negative opportunity means they're doing really good.
So you can compare what's happening uh here versus what's happening here process-wise, right? And occasionally when you have this data um you're going to have to actually dig making sure that when you share this with physicians responsible or um you know whoever is the in charge they would like to see well yeah you're saying this that we have the highest opportunity right 35 but who are those patients like tell us right so you want to get into that detail as well so you can I mean I I just want to prepare prepare for that conversation with that whoever the in charge of that facility is with all the data available.
So I'm going to ask Jeie um you know and this is the last question and then I'm going to have summarize this and you all can ask your questions follow-up questions. So I'm going to ask Jeie uh to give me the list of patients that returned because of septicmia and I have provided this that says like observe to expected is greater than five. What it really means is we are only interested in patients that had very likelihood of coming back was um less than 20%.
But they returned anyways. So we are only focusing on on those those patients. So uh let Genie run this really quick and then we can and this is a loaded question by the way. There's a lot of things going on and if you can see the code here uh all that genie is doing with additional uh you know um factor of greater than five which is you know likelihood of five times less likelihood and they still return. So you can see now I have these patient information, their admission and discharge date.
And with all these information uh I can have the next uh conversation with whoever you know the uh the excuse me uh the next the person is this is this dings all day long when you're at work right? So um you you saw in all of these at uh we were able to do that in in not that much effort like in if you if you have to do rag it takes months to get it right you know with all the right documentation. So with uh with this I'm going to have I'm going to ask my colleague uh to come and uh do a final summarization and then close the um uh the meeting for us.
Thank you. So, how long do you think that took to do that when he started in G with Genie? That's what might get lost a little bit. Three days. He did this in three days, right? And I I know as a non-practicing reformed data scientists, I couldn't do this in three days even in my peak. and neither could, right? Three days we did this, right? And so, let me just talk a little bit about the results. And that's the great thing about Genie that I really love is it took all of our work with Unity Catalog and metadata and built us a bot that could help our our organizations and our advisory folks help people come home to their families.

30:12 Recap, results, and next steps

Not to be too missionary about it, but that's why we chose this example, but like three days, guys. And while I think and I are awesome, if we can do it, probably most people in the room can do it. If you've got uh if you use unique catalog and use Genie and describe those instructions, you know that we're good, but we're not that good. And so what I would impart to you guys is this is a real tool that can really accelerate.
I mean, everybody's doing projects, not that much stuff is getting to production, especially internally. This can get to production immediately, right? Um so I just want to impart that. So we did a case study with data bricks because we were impressed and they they asked us to and so results again from initial configuration it took three days of effort. Uh when it write SQL it's 10 times faster than I am writing SQL.
Now you might say I'm slow. We tried it with other people too. You know I'm an XDBA. Um we said well surely it's not accurate. There's got to be something wrong with it. We tested against production reports that have been hit more than a 100,000 times, right? And none of none of the answers were wrong. None of them. We ran through with our trainers and our tech non-technical people. Every single answer from that bot was right.
Now, it doesn't mean it won't ever throw an error, but for the stuff we tested that had been VNVed through our production development process, it was 100% right. So, um, what are we going to do next, Linenie? Well, we're working on now, um, we we this is an internal production release. How do we get it through to our 20,000 users, right, that are in hundreds of hospitals. That's that's making it really real. So, we don't want to have a bad day.
And we're working on with the data bricks team on some features that will allow us to more easily expose Genie rather than rebuilding Genie with the API thing, which I know we could do. um we want to really define when we want to use Genie and Genie for aenic AI versus Mosaic ML and stuff like that. There's a a fine line. So we want to make sure that we're doing that right and we want to look forward to some of the new features which I'll talk about in just a second here.
Um really what we want to do is make sure we get a path to production all the way all day long. And so one feature that you guys may be aware of if not they've got a CRUD feature coming for Genie. So it's fine, but we don't want to deploy this by hand 400 times. We're talking at scale, right? So we want to have a CRUD feature that allows us to do API work, but we want to deploy at a scale. We have already trained our internal team on genies, but we want to consider what operations process like any other bot that has a thumbs up, thumbs down, who does this go to?
Somebody thumbs downs it and goes nowhere, that's not great. So we want to make sure our operations teams are set up to kind of scale this when we go externally. And then as we roll out more genies with that CRUD feature, um, data bricks has coached us maybe to consider a genie of genies. We haven't got there yet, but like you know, usually you want to start with your bots small and then subject area focus so you know you trained them.
Their accuracy is good, but like when we get 10 or 20 of these, right? We want to loop it into a genie that can know where to route the question. That's what we're thinking at least initially. Um, and you know, we're hoping for big announcements um, tomorrow like everybody. We don't know if there's going to be any big announcements on the AIBI tool, but we're hoping that we continue to see data bricks improve the journey.
Um, you know, we like engineering too, but we want to get insights in the hands of people that can make a difference, right? So, we like Genie because it helps accelerate our production timelines and it's a tool. I mean, I'm saying it's the tool and the only tool, but it's a tool to really get this into real life production, you know, and make sure that it's bulletproof. Um, so that's kind of it with that. Um, you know, check out our case study if you want.
It's up on the web as well. Again, LinkedIn. Um, and then we have a few qu minutes for questions if anybody has any questions.

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.