Skip to main content

Genie at Scale: Best Practices for Enterprise Rollout

Summary

  • This video covers how to scale Genie and AI/BI from a small pilot to tens of thousands of users, addressing provisioning, data access governance, content distribution, and mobile access under enterprise security policies.
  • A case study from a 70,000-employee manufacturer targeting a 100,000-user deployment illustrates the challenges and solutions for large-scale Genie rollouts, including phased adoption from pilot through UAT to full production.
  • Topics include token-based pricing with cost caps and chargeback configuration, the new Consumer role for Genie access without full workspace privileges, and embedding Genie in Microsoft Teams, Slack, and external portals.

Genie at Scale: Best Practices for Enterprise Rollout

Watch: Genie at Scale: Best Practices for Enterprise Rollout
As organizations adopt agentic AI at scale, getting access, governance, and deployment right is more critical than ever. Rolling out Genie and AI/BI across thousands of users demands clear policies for user access, data governance, and content distribution to ensure secure, consistent, and scalable adoption—including on mobile devices under enterprise policies.
We’ll walk through best practices for deploying Genie and AI/BI at scale, helping you move from isolated use cases to organization-wide impact.
Topics include:

  • Onboarding and provisioning users efficiently across business teams
  • Managing data access and permissions to ensure secure, governed usage
  • Publishing dashboards and Genie experiences for maximum reach
  • Enabling secure mobile access aligned with enterprise device and security policies
  • Monitoring usage, manage sharing, and identify opportunities to drive adoption
  • Deployment patterns and best practices for scaling adoption across the enterprise

Talk By: Dillon Bostwick, Genie & AI/BI Product Adoption, Databricks ; Xiaolei Shi, Sr. Engineering Manager, Databricks ;

Chapters

FAQs

What is the Genie Consumer role in Databricks?

The Consumer role is an entitlement that gives business users access to Genie experiences without requiring full Databricks workspace privileges. This video explains how it simplifies provisioning at scale by allowing admins to onboard large numbers of business users efficiently without going through a full IDP overhead process.

How do you scale Genie to 10,000 or more users in an enterprise?

This video walks through a phased rollout approach — starting with a pilot group, advancing through UAT, then scaling to 1,000 and eventually 100,000 users — covering AIM configuration, workspace access controls, and custom enterprise domains. A real-world manufacturer deploying to 100,000 users is used as the guiding case study.

Can Genie be accessed on mobile devices?

Yes, Genie has a mobile app for iOS and Android currently in public preview, enabling business users to query data from their phones under enterprise device and security policies. This video covers how to configure secure mobile access within enterprise mobile device management frameworks.

How do you manage Genie costs at enterprise scale?

The Databricks Data and AI platform supports token-based pricing with cost caps and chargeback configuration, allowing admins to control spending per team or department. This video explains how to monitor usage patterns, manage sharing, and set budget limits to prevent runaway consumption as Genie adoption broadens across the organization.

Full transcript

[00:07] Today we're going to be talking about Genius Scale. So we talked a lot about uh you know all the features that you might have saw on the on the keynote yesterday, but what we didn't talk about as much as uh what do you need to be thinking about? What kind of questions
[00:19] do you need to be asking your account team or be thinking about yourself to get past that initial kind of a hundred or a couple hundred or maybe a thousand users and really starting to get towards, you know, 10,000 plus users, maybe even a 100 thousand users. And if
[00:32] you're not there yet or you don't have that user base, um, that's totally fine. There'll still be a lot of stuff applicable uh, to you here. And this is really going to be mostly tailored towards admins or or space curators um,
[00:44] who are thinking about these kinds of scale kind of questions. Um, but if you're also a business user or a consumer, um, this will hopefully be applicable to uh, as well, just so that you can kind of start to understand what's going on behind the hood. Uh, I
[00:58] am Dylan Bastwick. Uh, I'm a I'm a product adoption lead. That means I'm a kind of a conduit between the product uh, team and the field team for all things Genie uh, at data bricks at kind of the global scale. And my background
[01:10] was uh, as a solutions architect. So I was working with customers for about six years uh, before this. And then you want to introduce yourself as well. >> Hi, nice to meet you. My name is Shalay. So I'm from the engineering side working
[01:22] on the Genie one. >> Awesome. >> With Dylan, we will talk how to bring user to Genie. >> Great. All right. Uh so like I said, uh we focused a lot. If you haven't seen it
[01:34] already, uh you can check out the key the the Tuesday keynote. Um skip forward to the genie section. We launched a lot of features and the TLDDR here is basically it's all about how do we add
[01:47] more value uh to the end consumers of the data. So you've invested heavily in the data bricks data platform. You've curated your data sets. You set up governance and controls and everything. How do your business how do your direct
[02:00] line of businesses get more uh from the data that you've curated? Okay. So but this is great and everything. Uh, one of the things that we didn't talk so much about in the keynote was how do we actually start to get this into production, right? A lot of companies
[02:14] that I talked to, you start with a relatively small uh user base for an initial PC, right? Maybe you're doing uh like an AI pilot or maybe you're getting asked uh by your leadership to look at
[02:26] all the different AI tools on the market and you're probably choosing a couple of like power user tinkerer type personas. A lot of them probably already have data bricks access, right? So they're typically pretty like well-versed in the data sets and they can look at a
[02:39] generated SQL statement and probably sus out any potential issue issues or gaps or inaccuracies and it's extremely unlikely that you're going to run into any sort of cost caps uh with the way that we've priced Genie so far. It's
[02:51] basically almost guaranteed to be free at that point. Then we start to scale uh to UAT, right? You're not putting it into production immediately. Um but PC goes well, which is often the case. And then we want to start looking into maybe
[03:04] you start actually working with some of the users in the line of business that you don't necessarily have access to. Maybe some of them don't currently have data bricks access today. Um a lot of them probably still use data in their
[03:16] day jobs, but it might not necessarily be their main area of concern. So maybe they can write SQL here and there. They know some of the gold tables and you're not going to have a direct relationship with all of them at that point, right? There might be some questions that you weren't expecting. Then we get into a
[03:29] thousand users. This is the point where with our old pricing model that we're actually uh now deprecating uh you might have actually started seeing some throttles. You might start thinking about how do we actually increase the capacity. You start thinking about what
[03:42] are the things that we weren't thinking about at the smaller scale. And uh there's going to be new lines of businesses that are going to come out of the woodwork and say hey I really want access to this thing. I want it on mobile. I want it on teams. I want it at
[03:54] the place where I'm already working on a day-to-day basis. I don't want to have to get onboarded uh quite as aggressively and you're going to need to be able to set it up for them quite quickly. Then we get into 10,000 plus. This is the point where I think a lot of
[04:06] customers we don't even know what questions to ask and this is usually when Jalet and myself um and folks on my team tend to get involved and it's like what question like you tell me what question I should even be asking uh
[04:18] because I don't want to be on call when this thing uh has any issues. Well, the good news is we actually have uh customers that are doing it at this scale. Um, and we have learned a lot from getting these C company these
[04:31] customers successfully into production. And for the rest of this talk, we're just going to share all the learnings and everything we know that uh you need to get to that scale. And again, like I said, even if you're not getting to 10,000 users, uh all these things are still super relevant and things
[04:44] hopefully you can take home uh to your work next week. All right, so here's an example. Um it's a large uh manufacturing company been around for over a hundred years. Okay. Um one of
[04:56] the biggest manufacturing companies in the world. It is a successful use of our 100,000 workspace user limit. Um maybe a lot of people don't even know um this this is true but our workspaces can handle up to a 100,000 users. Okay. This
[05:10] was also one of the biggest onboarding in the history of that company's IT. Right. So massive manufacturing company 70,000 employees. They work with hundreds if not thousands of software vendors on a regular basis. This was one of the biggest onboarding of consumers
[05:23] to an existing product that was historically only for technical users uh in the history of this company. Right? So how did we do it? On the left hand side, this is to me something that looks familiar to a lot of people who have kind of grown up in data bricks and sort
[05:35] of for historical reasons started to scale past a small number of workspaces. So you could have hundreds of workspaces, no no chargeback, no tagging. If this sounds familiar, I know it can be stressful, but it's something that we see a lot. And the good news is we actually have a lot of tools uh to
[05:48] get you out of there and and and implement all the best practices there. And also they only had about 2,000 monthly active users because that's pretty much the size of all of their data team. So all their data scientists and data engineers, they're already on
[06:00] data bricks, right? Um then about a year ago, they did they turned on this thing called AIM. And what AIM does is it makes it a lot easier for when new parts of the business want direct access to
[06:12] Genie, you can basically turn them turn it on for them, add them to the allow list, do a couple of really quick provisioning steps that we'll walk you through, and you don't have to do all the IDP wrangling on a regular basis. Okay? And then more recently, we turned
[06:24] this thing called consumer role. And what that is is, you know, historically with data bricks, we've always been a techn technical uh first platform. Most of our users know exactly what a parallel compute cluster is. Uh most of
[06:37] our users understand a little bit about if not a lot about uh training machine learning models and notebooks and different times types of coding languages and etc. Um but in this case right with we're actually looking at
[06:49] 70,000 employees the vast majority of the active users are not going to be technical. And so what you need is to turn this thing called consumer ro on by default. So when someone logs in for the first time, they see an interface that
[07:02] is designed for someone consuming the data, not creating tables and notebooks and all these other things. Okay. The net result of this is one in five employees at this company became monthly active users. So they used data bricks
[07:15] or Genie to solve a real business problem and help them um in in their business situation at least once a month if not weekly. And a lot of those actually became daily active users. So they're using Genie on their phone in in
[07:29] Teams on a very regular basis. Okay. All right. So what are the things that we need to do? We separated this into four categories, right? First is that platform readiness piece. So we'll talk about aim and consumer role. All the
[07:41] things that you want to have in place uh to make your life easier from a security perspective later. Then we're going to talk about where how do we meet the user where they are. A lot of users are going to want to log in directly into the UI. you saw a lot of the features that um
[07:54] that we created to make that really easy for them. But a lot of them want to use it, you know, when they're boarding a plane or in the expo center or when they're on a factory floor, right? Then we'll talk about customizing it to your company. That didn't really matter that
[08:07] much when you're using data bricks for technical users. They kind of understand they know data bricks. Um but now that's changing when when when Genie is primarily for business users. And then we'll talk about uh something that's probably on everyone's mind which is uh cost management and uh how to avoid the
[08:21] classic uh token maxing kind of situations. Right. All right. So I will hand it to Jali now to talk about the the first three sections. Cool. Thanks Dylan.
[08:33] I can keep this one. All right. Uh let's look at this one first. So platform readiness. Uh what this mean? This means actually the feature we build in data bricks as a platform to help you to
[08:46] scale your user uh development and deployment in your organization. We are going to talk about several things but before that I want to quickly talk about the prerequisites. So what type of the
[08:58] account is uh eligible or qualified to enable Genie to be ready and uh TLDDR majority of the accounts you have are fully qualified. And the reason we wanted to mention here is for example
[09:10] the first one we want we we require the tier to be premium or above. If your account is below premium which is we call legacy account this is means you
[09:22] are a longtime with us. Uh thank you for being together with us for many years. And uh we also have a pivot pass to help upgrade you to the premium or above which is the latest stack we have. And
[09:35] you probably also notice several requirements related to the permission which is a unity catalog as well as identity management because we are going to deploy and enable tens of the
[09:49] thousands of user in your organization. So identity management going to be very critical. So there are several features already enabled by default in your account and if not you can also talk to
[10:01] us. So there's also a very good uh pivot pass for you to be enabled there. Cool. Uh now we're ready. Let's walk through this uh platform feature one by one. So
[10:14] in total only three steps is required. Let's assume I'm a admin of my company called Acme and uh CEO told me that we have a business uh users from our marketing sales wanted to use the data
[10:28] that our data team and engineering team built. So how do I do this? Just a three steps several clicks you will be able to do that using all the platform features we built. The first one we called
[10:41] provision user using AIM. So, Dylan just mentioned uh with our pro previous uh customer. So, I stands for automatic identity management. So, this means you
[10:54] no longer need to do any uh wrangling with your IDP scheme. So, we will talk about details later. And second one is uh consumer access. Similar to uh the
[11:07] previous customer what they do this is give you a very good control as admin for this number of the user you don't worry about their abuse of the usage and lastly we also offer something very
[11:20] special called a customer URL this means you do not need to remember this like a long like uh you know different like a long code of the URL for data bricks anymore you can use your company name as
[11:33] branding company.databbricks.com databicks.com. All right, let's take a deeper look for AIM. So, automatic identity management. What this means? Basically, this means you can seamlessly manage the identities
[11:46] directly from your IDP. So previously if you are on sim which is the solution that we offer for our uh previous customers or some of our existing customers essentially you copy your data
[11:59] from your IDP into database and this require a lot of uh you know maintenance and overhead as a mean this is a netmare. So what we are doing the first the pinpoint is no longer there. So you
[12:13] can directly just manage everything in your IDP and the data bricks will have this manage the sync directly pull from your IDP and so as a user once you got provision
[12:26] you have to wait for the next sync schedule in order to access data bricks. This is no longer the case in the aim which is just in time provision will happen when you log in or when you got
[12:40] authorized everything will just like realtime pull from your IDP with the latest uh current uh information and membership. And lastly this is the ultimate goal why we want to do this. We wanted to unblock
[12:54] all the collaboration within your organization which you can seamlessly and frequently sharing and collaborate with your dashboard and other data.
[13:06] With that I want to quickly walk through what this looks how this looks. Uh it looks very great. Uh but uh what's the user flow as a domain? I already enable aim in my workspace and account. So as a
[13:18] new user I added to the workspace the first time user login in so login through the single sign on through the SSO. So we will route this request to our account level which is uh the
[13:30] central managed uh aim um place that we will do the next place which is to authenticate the user directly to the IDP. So in this case we use ATRA which
[13:43] is um from from um from from Azure in this case. Then we got the latest data we put the data back into our system. So even though this data is not in the
[13:55] system we proactively manage and sync into our data brick system and bring it back to the workspace. So on top of this we have a more than that uh one more
[14:07] thing which is we will manage that and actively sync this membership with your IDP. So no longer have a customer script in your scheme anymore.
[14:23] All right quick wrap up for this. What if you are admin what this mean to you? So TLDDR you only need to manage one source of truth in your IDP. You do not need to manage any scripts any sync
[14:35] scheme sync and it can also help you to solve the problem like a nested group which not supported very well in the scheme sync as a user. You will always get the real time sync or just in time
[14:49] provisioning. So there's no like ping your IT saying I still cannot access but your it saying I already put you into data bricks. So that's not no longer a thing with your team member. You can
[15:01] directly share things and collaborate with your team member even though they are not in data bricks yet. Cool. Um yeah so going going back to my example I enable the aim for my account
[15:15] and workspace I sync with my IDP for the users I want to bring in. But the problem now is they are not technical user. they don't know what they create or they just like accidental accident
[15:28] accidentally run some very expensive query I receive a surprising deal what how do I prevent this that's how we introduce the consumer role the consumer role is a I would say
[15:42] least priv privileged entitlement but also sufficient enough entitlement for people to use genie so list in a way that if you compare with SQL access and
[15:54] workspace access which is a two higher author driven uh entitlement. So consumer role cannot create anything. So they cannot create a security they cannot create a dashboard. So you don't
[16:07] worry not not even mention to create like a cluster. So you don't worry that computer cost from there but they'll be able to have the access to chat with Genie one talk to Genie agents
[16:22] look at the AIBI dashboard and use the data bricks apps. So that's still a lot which means uh this is the truly uh managed uh user row that's why the previous company deployed 70,000 users
[16:35] using consumer row going back to here yeah as an what this means um so does that require a lot of work to enable consumer row so uh what I
[16:49] should do um three steps again here but caveat you only need to to one time manual migration here which all also have a built-in flow and number three
[17:01] and number uh number two and number three is automatic after you enable aim. So let's take a quick look for the first one. The reason we have a one-time manual migration for consumer role is
[17:14] because existing when we have the SEO and the workspace access which is higher priv privilege entitlement. So we also have a default user group which means everyone within the workspace will
[17:26] always have this uh be part of this group and this group always have one of the uh higher pervision per privileged uh entitlement even though you add another user to be consumer only they
[17:39] will inheritate the higher permission from this group. So essentially what this process do um is to keep the existing existing user with the
[17:51] same permission no change at all but also change the default group to be either consumer role or no permission so that you can always add the new user as the consumer role. This is a very safe
[18:03] is already launched in our uh admin settings uh in the advanced section. So you can take a look after you finish that basically your job is done. You can use the existing way
[18:15] that you bring user or group into uh workspace they will um and mark them as a consumer. You can see the consumer only checked here uh on the second row.
[18:27] So they will be the consumer and if you use the group over there they will automatically think when you add the people into the IDP group and then AIM will bring this user into data brick
[18:39] system and uh automatically everything is end to end working. All right. So now I'm happy because I know so even though I bring a lot of user they won't uh abuse the compute and
[18:55] uh I won't receive any surprise um from the the building side. So how do they remember how do they access data bricks in this case we used to have a super long super super long uh different like
[19:09] even myself cannot remember the the the URL for for data bricks. The only solution is like you create a goal link but now so we offer this um good custom
[19:21] URL based on your company name. You can have in this case acme.deb.com is our example company. So we can have this one. We actually provisioned pre-provisioned 400 accounts. So uh with
[19:35] their company name plus uh deadbris.com and this feature is u fully enabled uh in the account console. So uh we call this like gated public view because we
[19:47] want to you talk to us and verify this name is the name that you want to go with before we enable that for you. Um going to the last slide to wrap this up.
[20:00] So quickly mentioned the status of all platform features we have. AIM is a GA for Entra which is uh fully available on Azure but if a customer on AWS and GCP
[20:12] using entra this is fully support as well for octa is a popular IDP solution we are public preview uh starting this week and uh with AWS and GCP for
[20:24] consumer actually since last year we already g that one. So that's the reason the previous large customer already adopt this but we keep improving the migration flow to make it more seamless
[20:36] and easy. Lastly, customer URL. Uh this is the public preview and please talk to us. We already have some prep work done for your account and um we can also um enable that with you uh fairly quick.
[20:52] All right. So this is the platform. So I'll bring the users uh we want to quickly mention so how your user will access Genie and in what way. So of course this is a web uh we have uh our
[21:04] genie one available and uh they can ask the questions and talk to the data from our web solution. So we also uh support a genie within your existing tools like a business
[21:17] tools like Microsoft teams and slack. you can ask a genie from there and we also have this um brand new Genie mobile app. So it's uh publicly available in uh
[21:31] Apple app store and uh Google play and uh this is all also launched a public preview everyone can download but this is a brand new surface so this is the one thing we wanted to mention to you
[21:43] how you going to deploy the Genie mobile in your organization um since the summit and especially after yesterday's keynote we have a lot of questions about this and we want to quickly clarify something
[21:56] as the main of your organ organization, you would you would want to work with your IT team to set up the access for your mobile app or if you don't have any restriction to access your data bricks
[22:10] account and workspace the Genie mobile app will work out of the box. But we still recommend you put something you know as the managed app or uh at least a single sign on for this this one. Uh
[22:23] several things to quickly mention. So the identity is a single sign on such as uh you know allow the sign on to Genie mobile through your IDP um VPN control.
[22:35] So this is the solution that we have this called IP ico which is a IP access uh list through the MDM management and through the distribution channel either from the app store or from the MDM. This
[22:48] is up to your choice. So roll out flow is something we want to quickly mention if you wanted to work with your uh organization to roll out this one because this is actually very easy to get the insights from your pocket uh
[23:02] when you have the Genie mobile but you need some prep work and help from your IT team. So as admin you can enable the preview and it will should work and with
[23:14] you and help you to set up the right network solution MDM and going back to yourself you can start rolling out as a pilot with a smaller group and then do the full roll out within organization
[23:32] and this slickly mention the Genie mobile we followed every security protocol within data bricks we do not bypass anything. So every technology we use is a trusted technology within data bricks. So uh such as you see everything
[23:44] don't worry for any concern from here. All right. So now your user can access Genie anywhere and uh one more thing. So as a do I always wanted to make the
[23:58] workspace be a better and a beautiful how about make the workspace looks and feels like your organization. So we have this uh four things that you can curate
[24:10] and customize your workspace and a genie one is your brand color your company logo and markdown text is that announcement you want everyone to look
[24:22] and read and paint the content. Maybe you have your dashboard and your genie spaces or um different genie agent and a different asset you want your user always access. We also offer the ping
[24:34] content and advanced customization is coming soon. Great. Uh now the most important now we have a lot of user how about the cost? How to manage the cost? I will go back
[24:47] to Dylan talk about this. >> Great. Thanks uh Charlay appreciate that. All right, we're going to wrap up with cost. Um, we're going to spend a couple minutes on this. All right,
[25:00] so the first big thing is we're changing the cost model. Um, if you didn't know how we used to do uh limits around Genie before uh you you know you don't even bother learning this, but we used to
[25:13] have this thing called questions per minute. We still do this. It's going to come live um uh closer to July, right? we're going to start introducing this. Um, but we had this thing called question per minute and it was our own
[25:25] construct and we set limits uh around this notion of questions, right? Um, what we're doing instead here is we're removing those limits altogether. We're basically going to make the amount of tokens you can consume on Genie
[25:38] limitless and instead we're going to make it consistent uh with the rest of how we do pricing in data bricks which is basically uh consumptionbased. So, we're going to have usagebased pricing
[25:50] that's going to be centered around the amount of tokens that you use uh across any managed uh Genie service. So, whether that's Genie 1, Genie agents, and Genie code, it'll be a completely consistent uh limitless pricing model.
[26:03] Um and in exchange for removing the hard limits, we obviously need to give you some way to control caps. And uh so the way that we're solving that is we're going to integrate Genie directly with
[26:15] Unity AI gateway which is already uh where you would be doing cost management for all of the different other um AI aspects of uh the database platform. So getting a little bit deeper into it uh
[26:27] again like I said before we it was QPM based. We maxed out at five questions per minute uh in the API. Again, if you're not familiar with that, um, doesn't really quite matter because
[26:39] we're getting rid of it. Instead, it's going to be token and consumption based. We're going to give every single user on data bricks 150 DBUs free every month.
[26:52] In practice, that usually equates to a little over $10. Okay? So, in US East, it's about $10.50. That reset resets on the first of every month. Okay? In practice, this means
[27:04] that about 80% of your users will pay nothing. That's what we found in our internal simulations. Um because that's per user, right? And if you have like let's say 10,000 users, that basically
[27:16] means you're going to get uh over a million dollars for free per year. Okay? If assuming that those users are maxing out uh what they can and when you hit that limit, you automat pay metering automatically kicks in until you hit the
[27:29] first of the month and then it resets. Okay. Um, like I said, we're adding cost controls uh at every level of granularity from user all the way up to the account level. And the other thing we're doing is we're unifying it under a
[27:41] single skew. Um, for those of you who have used other parts of our AI uh features, you might be familiar with serverless real-time inference, which is the skew that we use for basically any token. So like FM API, agent bricks,
[27:55] etc. They all use that same skew. This also means that if you have a discount on that skew, that discount actually now applies uh to Genie across the board. Okay. All right. So, how far does $10 get you?
[28:09] Uh like I said, about 80% of users will pay nothing in the simulations. Obviously, that number might go down as uh your users start to get more mature and start using Genie in more advanced ways. um the median cost for a question. So a
[28:23] roundtrip, you know, how many sales did I make in Q4 kind of question, it's only about 10 cents. Remember that's the median. Uh there's there is some skew there, right? Being honest, there's a there's a there's a fair amount of variance. Um but only a small number of
[28:37] uh users in a typical scenario are actually going to be doing complex enough stuff um to hit that kind of extreme end uh of the spectrum there. meaning that if you have a large user base, those business users that
[28:50] longtail, they're not going to pay anything and then you'll have a small number of power users um that might be making up the majority of the spend, right? So that means that you can ask about 100 questions per user uh per month for like an average business user
[29:02] asking relatively simple questions. Um and it can also for coding sessions, you can you can get about a coding session a day uh without paying anything here. Obviously, if you're doing more complex vibe coding, if you're running like a Ralph loop, you're going to max that out
[29:14] pretty quickly, right? So there's a kind of a um uh there's an extreme case there, but this is kind of uh how we've been we've we've set up the free tier. The other thing I mentioned is cost controls. So out of the box with the
[29:27] integration with Unity AI gateway, um you get caps at the user level, at the group level, at the workspace level, and at the account level. And those can be hard or soft. It can be hey this is how
[29:39] much I want um that user to spend and then I want to set a limit there. Or it could be, you know, when this group approaches 90% uh of the budget I allocate to them, I want to send some sort of notification.
[29:52] And the other thing we're doing is we are building in a cost monitoring dashboard directly into the product. Uh so that's what this is the this is a picture of that dashboard. Pause if you want to take a picture of it. Okay.
[30:05] And then this is a picture of um the email that that the admin can get um or the user can get. Uh and looks like Kian spent quite a fair amount of money there. Uh so we said a $10,000 limit on
[30:18] him. So in practice, what should you guys do? What's the kind of the ideal recommendation here? Well, if you think about your typical business user, generally they're not going to need more
[30:30] than $10 a month anyway. So what we recommend you do is just set the cap to the free tier. That way you don't have to even worry about the budget or chargeback or anything. Just let those people stay underneath that in that free
[30:43] tier. And if they hit the limit of the free tier, just cap them for now. Then as you start to grow and your use of GD matures, you'll you'll start seeing certain power users, maybe certain lines of business that have a valid need to go
[30:55] past that free tier. And then what this lets you do is it allows you to make decisions around those as you go. So if chargeback is set up, that business has a has a budget. You can basically control there and then you can start to kind of give them overrides without
[31:08] having to necessarily worry about budget across the entire user base. And then finally, service principles do not get any free tier. Okay, just as a as a side note, if you're using GD in the API, always make sure you pass the user ID so
[31:21] you can save some money there. Obviously, they have to be real users, of course. Um, but if it's an SP, you should basically always have some sort of cap that's commensurate with the kind of budget that you're setting for that application or that agent, um, or
[31:34] whatever it is. All right, so with that, uh, we went through all of these different areas of concern. I mean, just to summarize, this is this is new for a lot of us, right? This is, uh, um, I mean, number one, LMS have only been
[31:47] around for three years. So, fundamentally, the way that we need to be reasoning about them is a little bit different. It doesn't quite look like a web app. doesn't quite look like a data platform when you think about cost and controlling it and scaling it. And the
[31:59] other thing is for a lot of us this is the first time that non-technical users are logging into the data bricks platform and so it's really important from an adoption perspective that we give them a first experience out of the gate. Uh we don't want to scare them off
[32:11] with something that looks too complicated or everything. So that's that's something that um is really top of mind. Um all right so with that uh we can uh take some 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.