Skip to main content

Scaling AI Field Assistants: Eli Lilly's Platform for Enterprise Change

Summary

  • Eli Lilly deployed an AI field assistant powered by the Databricks Data and AI platform to 12,000 global sales professionals, consolidating data from dozens of fragmented systems into a single interface.
  • The platform uses Genie for natural language queries, Unity Catalog for governance, vector search and RAG for unstructured data, and agentic workflows for task automation, with the flexibility to shift between Llama and Sonnet models as needs evolved.
  • Thomas Raborn and Todd Walker share lessons in change management, governance, and the collaborative frameworks — including the ADKAR model — that drove adoption across diverse field personas from territory managers to medical specialists.

Scaling AI Field Assistants: Eli Lilly's Platform for Enterprise Change

Watch: Scaling AI Field Assistants: Eli Lilly's Platform for Enterprise Change
Eli Lilly deployed an AI field assistant to 12,000 global sales professionals, consolidating data from dozens of systems into a single interface powered by Databricks. This case study reveals how the company balanced business requirements with technical innovation, using Genie for natural language queries, Unity Catalog for governance, and agentic workflows for task automation.
Thomas Raborn and Todd Walker share lessons from partnership, governance, and change management. Learn how Databricks' flexible LLM support enabled them to shift from deterministic Llama to Sonnet models, how vector search and RAG surfaced unstructured data, and how collaborative frameworks scaled adoption across diverse field personas from territory managers to medical specialists.
🤝

Chapters

FAQs

Why did Eli Lilly build an AI field assistant for their sales team?

Pharmaceutical sales reps often have only seconds to a couple of minutes with healthcare providers, making it critical to surface the most up-to-date information quickly. Eli Lilly built the assistant to consolidate data from dozens of fragmented systems into a single interface so reps could access everything they need instantly.

How does Eli Lilly's AI field assistant use the Databricks Data and AI platform?

The assistant uses Genie for natural language queries, Unity Catalog for data governance, vector search and RAG for surfacing unstructured data, and agentic workflows for task automation. This gives field professionals a unified platform rather than requiring them to switch between multiple disconnected tools.

How did Eli Lilly switch LLM models in their AI assistant?

Eli Lilly leveraged Databricks' flexible LLM support to shift from deterministic Llama models to Sonnet models as their needs evolved. This flexibility allowed them to upgrade to better-performing models without rebuilding their core platform.

How did Eli Lilly manage change when rolling out their AI field assistant?

They used the ADKAR change management framework to guide adoption and created structured engagement initiatives including First Look Friday and Testing Tuesday to give field teams regular touchpoints with new features. Collaboration between the business operations team and the tech team was cited as a key driver of successful adoption across diverse field personas.

Full transcript

[00:08] All right. Good afternoon, everybody. Thanks for coming on the end of the day and joining our session. Um let's see. My name is Thomas Rayborn. This is Todd Walker. Uh we both work for Eli Lilly and Company. Um they make medicine. We're based out of Indianapolis, Indiana. Anybody from
[00:24] Indianapolis here? From Yeah, there we go. Okay, a couple couple people. Anybody pharma in here, too? There we go. Okay. More fun. Anybody on the operations side in here? Okay, you guys come to this side for me, please. Everybody else can go
[00:39] to that side for Todd. But, uh I work on our US field operations team. Todd works on our US um tech team. And he leads that group as well. So, he and I've been partnering for many years uh building a lot of different capabilities. Um
[00:57] de- deploying new processes as well. Um we've got one recently we've been working on is kind of our um AI assistant. Um we've deployed it. We're in the middle of enhancing it, um scaling it up right now. That's going to be the focus of our conversation today.
[01:13] So, I'll start. Just make sure you know what you're getting into. Just a little bit of um why are we doing it? What are we trying to solve for? Todd's going to hit up on some of the platform, the technical components of it. And then we'll cover um a lot more about our collaboration, some of our
[01:28] lessons learned as well, and our partnership and why we think that's a key to success for what we're doing for our field teams. So, starting um with kind of the why. If you're if you're not um if you don't know pharma, just give you a typical day in the life. We'll
[01:44] talk about our sales reps specifically. Their primary job is they spend their days out in the field to try to see as many customers as possible to make sure that our customers are the health care providers, their offices. Make sure we share the most up-to-date information
[01:59] with them in their office so that they can make a good informed decision for their patients. Um if you understand pharma or can just relate, um offices are busy. If you've ever been in a doctor's office, they're busy. They've got important work to do. They're kind
[02:17] enough to let us come in, but the challenge we have is time. So, on an average, what we might get just generally, you may get seconds, maybe a couple minutes with a provider to talk about your product. Um we can't control that. So, that's something we can't
[02:33] control. It's just our environment and our nature. What we're focused on though, and what we believe we can control, is on the back end, making sure our sales professionals are the most productive and effective they can be. That moment and the few that we get with
[02:48] them is so important, so we want them to have the information at their fingertips so that they can have an engaging and a very effective dialogue. Now, the few things we've experienced over the years that get in the way, we all know data's fragmented. It's in
[03:03] so many different places. Our sales professionals have multiple tools they have to work in every day. The ones related to customer engagement, but even other ones related to the administrative portion of their job as well. Um so, what we've tried to do and we
[03:19] think about that, that just adds the administrative burden and it takes away their focus of what we want them them solely engaged with is how do I get an awesome customer interaction. So, the visually how this looks right now is on the left side, um you've got
[03:36] the territory manager in the middle, and that's a pretty true representation of the number of tools our territory managers have to interact with. Whether it's um customer information, sample information, uh employee like expense reports, policies and procedures, um
[03:53] training, you name it. They've got so many different places they have to go to get information and process um for their job. What we've done is we've shifted and obviously put our um AI engine in the middle of it, uh new user interface powered by AI as a field
[04:11] assistant. So, what that does is it's going to consolidate all that information so the territory manager doesn't have to manually go out, find the information, it's right there for them. More importantly, though, if you look at the left side, think about one territory
[04:26] manager, but if I give five territory managers the same set of customer information, I know for a fact they will interpret it differently and make different decisions off of it of how they should engage their customers off of it, too. Just multiply that by
[04:41] hundreds and thousands, you've got a lot of variability and inconsistency in the information that they are processing and using in terms of insights and actions. Using our AI engine, we're able to kind of use that to help drive the insights and the suggestions so they don't have to
[04:58] come up with that themselves, but as we know, they are still in the middle of it. So, it is still up to the territory manager to take that information and see if it applies for the situation they are walking into with their customer. So, they at least have that at their fingertips, they are saving time, it is
[05:15] helping them be more productive and effective having this AI engine there with them as well. So, Todd's going to now talk about the tech platform, how we've gotten to where we are, and then we'll come back and talk a little more about our collaboration. Thanks, Thomas.
[05:32] So, as Thomas mentioned, what the opportunity really was is we wanted a a unified view of field performance across all these various platforms. And every bit of time that one of our territory managers spends looking for information is time that they are not in
[05:48] front of customers, which is where we want them to be. We also wanted to not just have general knowledge, but rather we wanted to have specific Lilly data woven through that as well. And of course it had to be scalable because as we're going to talk about this was not just for territory managers
[06:05] and business directors and AVPs and senior VPs, but rather some of our colleagues are here on the front row, MSLs, our medical community, uh director of strategic accounts, our payer community, our health outcomes related people, our reimbursement people
[06:21] in the field. A lot of different personas that we were going after as well. And we know that we also needed to have a platform that grew with the organization. There, I don't need to tell all of you things change so quickly on the business side and on the tech side. We had to have something that could adapt to that
[06:37] as well. So, with being from Indianapolis, and I saw some hands in here that are from Indianapolis, we thought we would use this platform discussion to compare it to the journey of a race car. If anybody watched the Indianapolis 500 this past year or were there in person, it was
[06:54] fantastic. It was the closest race in the history of the 500. 23 milliseconds is what decided the winner. I mean, we talked about milliseconds all through the keynote this morning. And you think about that, 23 milliseconds is what the winner won by. It was set up by a a yellow that
[07:11] happened I think on lap 197 out of 200, and then they restarted at 199. Armstrong was in the lead and Malukas and Rosenqvist quickly caught him. And then 23 23 milliseconds ahead of Malukas,
[07:28] Rosenqvist came across the line, the yard of bricks, and won. So, we thought what a perfect way to talk about here what we're talking about with the platform for the AI field assistant. What the user sees is really the cockpit. What does that front end look like? What the where the power comes from is
[07:45] really the model, the engine behind it all, and then as well, how do you hang it all together with the chassis? Those are the three things we want to talk through. So, starting with the cockpit, we wanted this to look like any other AI tool. What that means is whether it's
[08:01] Claude or whether it's ChatGPT or something, we wanted there to be suggested prompts. We didn't want it to be where a field user would show up with this the blank chat, and they would have to start from scratch, but instead we have suggested prompts. We also wanted them just to be able to type natural language.
[08:17] It's pretty obvious in today's world. We keep their chat history, so they can see and revisit those things. Some data we don't even want them to look for at all. We say, "Let's surface that in dashboards right as soon as you come in, so that way you can see that data without having to go query
[08:32] anything. We know it's highly sought-after data." Not just the data, but rather how do we surface insights for them that tells them what to do with it. So, for example, instead of just showing prescription data, be able to have data that shows them this account is going down. Here's some
[08:49] insights about why we think that could be and what actions you should take to try to remedy that. And then the last piece around Unity Catalog, that's super important for us. If you think about all the various brands of medicine that Lilly has, you think about all the field personas that
[09:04] I talked about, you think about all the geographies that are spread across the US, the mix of all those things are very important for us to make sure it's the right data for the right person at the right time. So, behind that then what's powering all that? We love the fact that Databricks
[09:20] allows us to have a flexible large language model choice. And if you pair that with the second point around Genie spaces, let me say how that played out for us. As we started with a very deterministic start in this. We had pre-pre-validated queries, tools that we set up inside of
[09:37] Databricks to say, "Hey, we don't want you to answer things that are not in these pre-validated queries. We think precision is super important. Only go after this." And we used Llama because it was more deterministic. We have now switched. We're using Sonnet and we're using Genie to develop those
[09:55] queries for us in real time. Because people have become much more comfortable with you don't have to be We're still trying to be, of course, 100% precise, but depending on what you're surfacing, there can be more tolerance for things that maybe aren't exactly precise, but are very directional in nature.
[10:10] So, we are very much in that journey now, um, and working through those things. Not all of our data is structured. Some is unstructured. Especially for our non-sales roles, and that's where vector search and rag is very important for us to be able to surface that data back as
[10:26] well. And then agentic workflows, we're going to talk a little bit about this on some coming slides, but in this case, think of this more of like supervisor agent calling other data-related agents to go grab what you need to do. And then thirdly, then we've got the
[10:41] chassis. So, what holds all this together? I mentioned Unity Catalog earlier. Again, very important for us because of all the different field personas, brands, geographies, all the different pieces to be able to make sure that we can have that role-based access. We have about 50 different types of data
[10:56] that we're bringing together. Some of that's purchased external data like claims data, something like that. Some of it's alignment data, master HCP data, it's activity data. All these different data pieces have to come together in the semantic layer to be able to help us with this.
[11:13] And then, of course, when you're bringing in all that various data, you need to make sure that you're monitoring it. Make sure it's high quality. Doesn't make sense to bring it in if it's not high quality, and you're monitoring that and, uh, having all the appropriate data checks on it.
[11:30] So, if we talk about collaboration really between Thomas's team and my team, I'm going to start here in the middle. So, Tech at Lilly, that's just what we call our internal IT group. So, in Tech at Lilly, we're really responsible for all the Databricks platform architecture, all the AI large
[11:46] language model engineering pieces, all the data foundation I was speaking about. Security is super important in our world. Scale is really important in our world. Have to make sure this thing performs well, and it's got to be reliable. Operational excellence is what we start with. If we don't have that, then we know we won't be able to do
[12:01] anything else that's cool. So, Thomas, why don't you go ahead and take the other two personas? Talk about those. But, back to the Indy 500 for a second. Um if you attend or however, we sometimes what we do, you pull names for winners, and I pulled the guy that came in second, and it doesn't feel good
[12:18] losing by milliseconds. So, um that was a little sore for me, but that's okay. It was a good race. Um from the field operation standpoint, our role we we face the business. I mean, we all do, but our primary job is to partner with our um different business partners. We
[12:33] help drive, um understand their strategy, the use cases, make sure we understand, like we talked about those important moments, but what else is it? What else are they trying to solve that we could bring solutions to to help them be more productive and effective. Um change management, we're going to talk
[12:49] about that in a little bit. That is a big part of my team working with our business partners as well. And then, nothing hap- we know nothing happens in a silo. So, between my team's leadership, the tech team's leadership, and our uh business teams, their senior
[13:05] leaders, um we kind of take that role of like going across function, uh bringing all those together to get to the decisions that are needed to move along our capabilities for our uh field teams. And then now, not on stage, but introducing a core group to our team are
[13:20] our field users. So, we're very fortunate. I don't know how it works in a lot of places, but we've got a dedicated group of field team members that are our partners. So, like it's a role that it's above and beyond their day job, but they sign up for it, they're interested in it, and they want
[13:36] to be partners with us to help build. So, they uh validate a lot of the use cases. So, hey, what are we talking about and is it really going to solve something that's happening on the ground? And more importantly, they do a lot of the um user acceptance testing for us. And then also um they'll help
[13:54] drive adoption. So, when when we talk about OCM and we go to launch, they're kind of the forefront in front of their peers helping us drive adoption. But, just having this on a slide for us is an example of collaboration, cuz you got to know your lane. And I think we
[14:11] know what each of us brings to this collaboration, and we know what we have to do extremely well to make it successful across the board. So, talking a little bit more about our field users, we engage them a number of different ways, but just like two
[14:26] examples uh we've done. Um this like a first look Friday, we got to have a catchy name, right? But, we we've we're in the middle right now of a pretty big enhancement deliverable that's coming up in the next few weeks. So, we actually had one of these Friday.
[14:41] I I was off, but um my team, Todd and team, met with one of our um senior business partners and and some of her team, and we were rolling out the updates that we've had and what progress we've made so far. So, we're able able to keep them on track. Hey, is this
[14:57] meeting the needs of what you're looking for? Is this kind of getting the look and feel? So, we kept them up to date, and I think fortunately we're on track. So, we're going to keep moving, but we kind of keep that on like what we call that first look Friday. The other thing, something another example we've done, and this is more
[15:13] probably with our field users, is like the testing Tuesday. Um like they're kind enough to take time out of their day job to partner with us. So, we don't want to surprise them or throw stuff their way. So, if we're in the middle of a big build getting into the UAT section, what we've done is hey,
[15:30] if it's over 1 to 4 weeks, whatever, however long that takes, we'll kind of try to spread it out, get a little bit of time, and just get 2 hours of their time so that they could work around it. But, just a couple ways we try to keep our users engaged. But, I'll tell you there's so many other examples. These
[15:45] were just two we thought we'd bring on how our kind of field users interact with us. And maybe surprisingly or not, I don't know, we have an opinion that governance is good. So, I we believe governance is a driver of collaboration because it's a
[16:01] it's cross-functional efforts between the business, field ops, and the tech team. And if you don't have a governance, if you don't have a hey, here's how we work, here's how we make decisions, our experience, you just swirl and spin, and then nothing moves
[16:16] forward. Um what we want and where we've seen success is when our team members, our core team members, are enabled and know that they could drive and get the work done they need to do. And then, they know, hey, we've got example on
[16:31] this project, bi-weekly leadership meetings that we come together, and it's like, hey, any status progress, anything happening, any barriers, any decisions, any stakeholders we need kind of need to work through, and that's with our senior leadership. So, we've seen great success
[16:48] in terms of driving governance through our build of these capabilities across ops and the tech team. So, I guess in in summary, um governance is good. Field users, getting them like early on board is great. And then, how
[17:04] do you just make sure everybody knows their roles, partner together, um have a little fun along the way as well, too, from a um collaboration standpoint. Um I'll start with a few of our lessons learned, and then Todd will close it out on some of the tech lessons learned.
[17:22] I think pretty general, but when you talk about people in process, over the years what we have seen and what we've learned is we all want to be fast. We want to We want to deliver with speed and we want to be agile.
[17:37] You've got to have the process in place to do that. You can't do a square peg in a round hole and expect to get the results. We actually spend time creating the process for speed. We've redefined several processes. Let's say we were on like a monthly sprint. Like we would get
[17:54] the group together and say, "What does it take to get this down to 2 weeks?" Our teams would make adjustments. They'd figure out how we would need to adjust work so that we could hit more of a 2-week sprint versus a monthly sprint. And then ad hocs, we always get ad hocs that come in. Hey, if they're We've got
[18:10] a way to prioritize if they're high priority, it's extra work. So, we've just got to move with that and try to meet those needs. In terms of establishing a cadence, um I I love having a good ways of working and how to get work done. Um in our in our
[18:26] role and like with this project, a a typical week what it might look like is hey, Mondays we get together, the core team. Here's the priorities we need to accomplish for the week. Um we reserve meeting time, whether it's a Tuesday or Wednesday, say a 2- to 3-hour block, and
[18:44] then that Monday decides what goes into that meeting. Um you can't just slam calendars every day. So, we like, "Hey, these are the groups. We've got the core team. Out of this core team, who needs who to move something forward?" It goes into that Tuesday or Wednesday meeting. Friday, Todd and I meet with the team in
[18:59] the morning, give us some quick updates, and then usually typically like that Friday afternoon, we'll send out an update to our leaders. One thing we've learned, and don't tell our bosses, we just shifted and we do that on Monday mornings because we were getting
[19:15] peppered with questions over the weekend, and our team felt compelled to answer them even though we did not want them to answer anything over the weekend. So, those go out Monday morning now, and they haven't even noticed, I don't think. So, it worked out pretty well. But, that's the cadence in addition to the governance meeting. So,
[19:30] we try to keep everybody coming along, keep everybody informed, keep everybody aware. Here's our progress, here's our barriers, here's how we need to work through things. The next thing, everybody has to speak up. We've got some high performers on our team, and we
[19:45] do not have time for people sitting in the corner with what we are being asked to deliver and accomplish. So, we count on people, our core team, it is up to them to bring recommendations back to the group, and it's a collective group of recommendations.
[20:01] Um examples, uh on my and Todd's teams, we have we have a number of individuals that work with us, our associates, they're demoing to the senior VP of a business unit, they're presenting to our senior VPs, they're getting pinged by the they they are fully engaged across
[20:18] every level and speaking up and giving their ideas and thoughts. Um another example, how we feel on I was off on Friday, um had a afternoon off with my daughter, and then So, Todd, after we had that first look Friday with our senior uh VP, he was on with my team and
[20:34] his team. So, they were going through everything, kind of getting everything put together. What do we need to get ready to send out a summary? How do we get ready for next Monday with the what we came up with maybe four big items we had to keep moving forward from that call that we had to drive focus on. So, he was filling in for me on that one.
[20:51] I've talked about like the lesson of co-creating. The In this project, we've we've brought our our team has been on from day one. I think the first workshop we had in Indy was the week before a Christmas, and you know, you start winding down
[21:07] then, they still came. It was cold, it was frigid the week before Christmas, and we had our kickoff workshop with our field users in Indy. So, they've been there from day one on this journey. We've had other projects though, where you had to move with speed, or maybe our field users in the middle of a
[21:24] product launch, so hey, we don't really have time with them, that we've not had a chance to co-create. We've launched some other smaller capabilities. They're still involved. The only thing is it just shifts. Instead of getting it right up front, you're just going to have to fix it and do enhancements on the back end. So, it just kind of shifts where
[21:40] the work is, but they were still involved with that as well. But, these are the a lot of the learnings from people and process. I don't know if other companies which you use, but from a change management perspective, we use ADKAR. It's kind of how we use from OCM.
[21:57] See a head nod over there. Our team leads that my FieldOps team leads the change management with the business. What I say is one and three are pretty easy. So, what we'll do working with the field teams, awareness, knowledge, you know,
[22:13] if you need it training courses, videos, resource guides, communications, meeting with different lead teams. Like, we could do all that, and typically we could build up a pretty good buzz with something new coming out with a general sense of excitement.
[22:30] But, that's it. It really starts, and the rubber meets the road for the racing metaphors, on launch day. So, when you actually roll out something new, and now the users are expected to oh, now I need to start using this, incorporating this into my workflow.
[22:46] That's where OCM really starts. And for us, that's really like with number two and four, that desire and ability, there's only so much we could do sitting in Indianapolis. So, that's why we count on two groups, our field users, as I mentioned, to help drive the adoption
[23:01] and the use. And then the local sales managers as well. They're obviously with their teams every day. They see the opportunities. They see the challenges. And they know how to coach and position this to show them it's all about showing them how they get
[23:16] value out of these tools. That's the goal and that's what we want to drive from an OCM perspective. And then reinforcement's pretty easy. But this is a model we've used for many years on driving OCM pretty successfully across several big capability updates.
[23:33] Um my last slide for now is just over the years again, uh my opinion or our opinion, like in our field teams, there's kind of like three groups of users. You've got the early adopters. They're going to do anything new. They're excited. Let's get after it. Uh probably
[23:50] the biggest group, we've got those that are just a little curious. Um they want to. They've probably tried something. They've tested it. They probably just need like a what does good look like or show me an idea or give me a prompt that works, right? Just kind of you got to give them a little nudge. And then you've got those that are slow to never
[24:08] adopt probably. And I That's okay, right? Cuz I mean, they've been doing their job a certain way and they probably still can. Like they could probably still do their job. But I do know and believe because the tools and like this AI assistant, we've proven it to have add value. Like we can show how
[24:25] it can save you time. We can show how it gives you better insights. And remember our field users being there from day one, they've told us it's going to add value, too. So we feel pretty confident in the value that it's adding. And I think over time, those that just don't jump on board, you're going to start to see a pretty clear separation in
[24:42] performance in people that are driving things forward. It'll be pretty evident. But the the takeaway is move with the movers. Don't worry too much about the people that aren't ready to move. Just go with the movers. And I love I'm probably a curious one. On we're all going through an AI journey. Like if you
[24:57] show me a a way and it clicks with me if okay, now I got it. Like, help me get it. Those are the people we really want to work with and get them moving with some of our resources like this AI assistant. Todd here's got a few key learnings from the technical side as well and then he'll wrap this up.
[25:13] Over the course of time we've had these huge systems we build out that are one size for everybody and I think those days are gone. We are now trying to get more in this idea of a right-sized approach. As I mentioned earlier, we have very different personas, different needs of
[25:28] what people have. And so, the idea that they all see these different data, they may want to consume the data differently, they may want some sort of a cloud front end, they may want some sort of a react front end, but if we keep enough of the system the same, then we need to have the flexibility to
[25:44] be able to move with them wherever they want to go. We also know that if you make it too general, they don't like it. It's just underwhelming. If you make it too deep, it's overwhelming. It's too much. How do you really get the data and the insights that go with it for them?
[26:01] Uh I've talked about the one UX can't fit every workflow. The value pitch changes for the different audiences. So, that's something for us to always keep in mind as well is that what might resonate with the associate VPs of a field group would be way different to what might resonate
[26:17] with some of our medical team or what might resonate with a territory manager that's out in a specific territory. And the tools and expectations shift monthly. We no longer think about things in terms of like, oh, this is what we'll do for the next year or the next 6 months. We
[26:33] can see about 3 months out and that's about it. And so, we are trying to make this to where we can pivot as quickly as we can. We've already pivoted It's mid-year. We've already pivoted a couple times this year about what we're going after and where we're what we're doing. So, that's one key thing. Go from uh one
[26:49] size to right sized. And then this one's not as much as of a learning other than just maybe a vision of where we're wanting to go. We're wanting to take this field assistant to be sometimes I think Data Bricks uses this term like single pane
[27:04] of glass. That's what we want for this field assistant. We want them to be able to do all their actions that they need to do through this. Through other agentic things. Not just bring back data, but complete tasks on their behalf. Embedded into the flow that they're
[27:19] already in. The tools they're already using, the places they already are. And of course it's got to be customized to them with their access. And you've got to have that sort of auditability and and approvals as needed on that. And then if for some reason those
[27:35] upstream systems aren't available, you want it to be graceful. You want it to say something like, "Oh, it's not available right now." versus just crashing on them. Cuz that's going to be a horrible experience. So really in summary I would say for this whole thing, there's really
[27:52] three things that I think are super key. One is partnership. It's very intentional that Tom and I are here on the stage together. It would have seemed really odd if him or his team would have done this without us, and it would seem really odd if we would have done this without them. Partnership is very key
[28:07] in this thing. Second I would say is speed. We get pushed to go faster than we've ever gotten pushed to go. We're uncomfortable every single day. But I also understand if we don't move at that pace, what the business wants us to move or what the tech is innovating
[28:23] at, then we become irrelevant. Even as quickly as we feel like we push ourselves to move, we've had a couple cases already where we very easily could have become irrelevant with what we were doing if we didn't have the ability to pivot and listen to what it is. So partnership, speed, and third is impact.
[28:39] If what you're doing doesn't really matter and you're not going to drive an impact, then don't do it. Really you've got to have the value. So that's why we look at this and say, "Okay, we've kind of renamed MVP from minimum viable product to minimum valuable product." I mean, viability
[28:54] just means it works. Value is really where you're making the impact. So, it's kind of more like instead of just it working, can you get a spot in that person's day? That's where the value statement comes in. And in our world, that means you're bringing you're driving productivity and
[29:11] you're driving effectiveness for these different field personas. We are supposed to remind you to complete your surveys for this session. And then Thomas and I will stick around afterwards for anybody that's got any additional questions. Thank you guys for your attention. Thank you.

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.