Episode Transcript
[00:00:00] Speaker A: Good morning, everybody. I'm very excited to be talking with Roshana Golani, the VP of enablement at Databricks. And there's three things we're going to talk about. We're going to talk about databricks, we're going to talk about enablement, and Roshana is going to talk to us about her career and how she's really pioneering to me a whole new dimension and definition of what L and D is all about. So, Roshana, thank you for joining me. Why don't we start with a little intro? Tell everybody a little bit about your background and your role at databricks and we'll talk about databricks.
[00:00:30] Speaker B: Awesome. Josh. Josh, thank you so much. I feel really honored and privileged to be here talking to you today. A little bit about myself. So I really think of my journey as somebody who thought of skills and did that at the intersection of disruptive technology and the need for transformation. So I've had the privilege of working with multiple disruptive technology. The biggest one that's in front of us right now. But whether it was virtualization in the early 2000s to the next decade being all about cloud and then data and AI, and in every single one of these, humans are the most important part of that journey, people that make that transformation happen. And I've had the privilege to make enablement possible and skills transformation possible.
Amazing organizations like VMware, Amazon, Google, and now databricks to rethink how people should function in this new world.
[00:01:26] Speaker A: So you have a lot of experience with these big technology transformations and evolutions. So you told me a story when I first met you about your job and your reporting structure. And before we get into what databricks does, tell everybody a little bit about where you fit in the organization and what your mission is.
[00:01:42] Speaker B: Yeah, that sounds really great. Somebody asked me this recently too. And you want to be, when you're working in an organization or when you're doing work like I do, where we are focused on helping people, enablement or learning and development capabilities, whatever we are calling the term that we do, you have to really be empowered to do that job. You have to know that the company values that work, but more importantly, it's really needed. So you wouldn't do an enablement job in an organization where the product is super. I don't want to say like small scope, but like enablement doesn't matter. Right. So in this case, databricks is impacting very, very large enterprises to really think about how they transform to become a data and AI company. And so enablement was really, really key. And so you needed to do this from a place of not just I am doing learning. And then you're buried somewhat in the organization where you're just focused on generating a P and L or you're focused on creating the next training content. So I'm privileged again to say that at Databricks I get to work for one of the co founders. I'm part of the field engineering organization, but I work with the co founder. And what that enables us to do is that your org boundaries are no longer boundaries. Right. Whether I sit in field engineering, we're still enabling statistics sellers, we're still thinking of marketing, we're still thinking of product, we are working with product and engineering. So the org boundaries sort of gray out and you're just able to do the work that really matters for our customers.
[00:03:16] Speaker A: So is the traditional L and D leadership development, onboarding stuff not in your domain then or is that also part of your mission?
[00:03:24] Speaker B: It is for the go to. For the teams that support. Yeah, for go to market.
But we do partner up very, very closely with the people team who are an amazing set of individuals and we partner with them for when we are trying to think about leadership and culture development in the organization. We work very, very closely with them.
[00:03:43] Speaker A: Okay, so let me go back to the beginning of Databricks first. So everybody understands what enablement means.
So you validated the thesis we have, which is that most of L and D is lost in the shuffle, in the need to enable people to do their jobs better because of the formal approach to L and D that's been around for so long now, sort of being transformed by AI. Databricks is a very exciting pioneering company that's taken some incredible technology from UC Berkeley and built this whole new data platform that essentially transforms companies. Take a couple minutes and tell everybody about databricks first. Yeah, because I think that story is important because every company in some ways is offering something that's analogous to this, but perhaps in a different industry.
[00:04:30] Speaker B: Yeah, absolutely.
So as you said, databricks is a pioneering. It's an amazing organization and very, very simplistically, Josh. It helps businesses manage massive amounts of data so that they can actually get value out of value out of that data. Right. Like it is the basis of how enterprises do AI because that is what provides the context. So very simplistically, databricks is where businesses organize data to get value out of it. And it is done through this data and AI platform with pioneering technology like GENIE with late base agent bricks. But it all started almost 13 years ago, 2013, where seven researchers from Berkeley, like you said, thought of data as, hey, this is not for a few tech giants. And this should really, we should really democratize data so that everyone can get value out of it. So it's this research in academia that you know and a bunch of open source products that came out of there that, that generated that. And from the very beginning, the idea of open sourcing, democratizing was really critical. So enablement is core because user adoption, the idea of open sourcing meant that you had to reach a large number of people for adoption to happen.
The second thing is being born in academia. The value of education or the value of people learning about the technology was ingrained into who the co founders are. They are that today and I continue to work with many of them and they are just like, yeah, we've got to like train more people, we've got to get adoption going. Right. So it's very central to enablement is very central to who, who the company is.
The second thing is that the very first product that DataWorks came up with was also about breaking down these barriers. Like we'd have scientists and engineers and analysts all working in different, different things. And DataWorks created a platform that said we're bringing all of them together so that they can collaborate and they continue to do that. Like that was the first iteration. But even now where we are using GENIE and really democratizing data again for the business users, right? People like me can just go and query data in a way that it can actually help my business. So it's very central to what databricks is. So started off, like I said, with those seven researchers, one open source product and you know, millions of downloads to a massive organization that's growing 80% year over year. Right. And we are now thinking of serving privately held, privately held and over 20,000 customers globally is what we are serving. Right? So it's grown in 13 years to be this giant, but yet in a startup mentality organization that continues on its mission of making sure that people, enterprises can get value out of their data and AI.
[00:07:28] Speaker A: Okay, this is just fascinating. So let me just explain to people listening. I worked in the database industry for almost 10 years. So most companies have these traditional relational or other forms of data in older systems like Oracle or whatever and then they move them into ERP platforms and Databricks comes along with its new technology 10, 15 years ago and says, you don't have to take all that old stuff and Put it in a data warehouse, you can put it in a new structure, you can connect it together, you can add AI to it. And in some ways you were doing AI before we called it AI.
[00:07:58] Speaker B: That's exactly right. Yeah, we said.
[00:08:01] Speaker A: And now you can go to the databricks infrastructure and you can just ask IT questions and you don't have to know where the data is or where it came from or any of that. So in terms of enablement, what goes through my mind, being sort of a little bit older and having been through this, is it the traditional approach to training is. And I, I talked to EMC about this years ago and Cisco and so many companies and I'm sure you've been through this at V, where is. We're going to certify you level 1, 2, 3, 4, 5, you're going to take a bunch of courses, then you're going to take a bunch of tests and then if you pass the test, you're going to get a badge and then voila.
[00:08:36] Speaker B: Exactly, exactly.
[00:08:38] Speaker A: You're going to become the expert. Yeah, I don't think that works. But tell me, you know, what have you done to that model?
[00:08:44] Speaker B: Yeah. What you described is perfectly how most organizations think about training and enablement. And I'm not saying that those things don't have a place, so I want to start off by that. But what we have fundamentally transformed is this idea of push based enablement. Dash. I'll do this course to what I think of now as an AI performance engine. So, and this is something, Josh, I've said this to you before, you talked about this a decade ago, that enablement should happen, personalized enablement should happen in the flow of work. And we all dreamt of it, every learning professional dreamt of it, but we didn't have the means to do it. And databricks actually unlocked that for us. So I'm going to go back to push based enablement to a true performance engine. And courses still have a place for it, but it is not the beginning of it. We don't want to start with what course do we build and what certifications you want to need. Right. You really want to think about what is the job that the person is doing. Can I give them exactly what they need? So it's not just in case training, it is just in time training. Give them exactly what they need when they need it and truly just think about measuring their performance, their productivity in the flow of work. So we've flipped it on its head from just courses and credentials to context and Skills do I know what they're doing and where they're doing it? And that's what we are after. Right. And all of that telemetry based on your example earlier exists. So I know what they're doing in databricks. I know what they're doing in the CRM system. We know what calls customer conversations they need to have and can I prepare them to be successful with that at the time that they need it? Right. So that's fundamentally where we are. A push based enablement system is of the past and we are moving to the idea of measuring skills in the flow of work and providing enablement when people need it in the way they need it.
[00:10:41] Speaker A: Okay, so I mean this is absolutely nirvana. Yes, you, but you didn't buy this off the shelf. Tell everybody a little bit about how you stitch this together. Because this isn't like buy an lms, buy a content development tune, build a bunch of courses, launch them.
[00:10:56] Speaker B: Yeah, it was not easy is the thing to say. But databricks made it possible. So I'll start off with that. And the second thing that I think again back to the idea of you want to work in an organization and in a culture that allows you to innovate, that pushes you to innovate versus continue to do what what we were doing. So four years ago when I joined databricks, that was what we started with. We built amazing courses. Our customers loved it, our partners loved it. And we had requirements in the program for part to get certified, just like every other company does. And it's working and we still have that. I'm not saying like we've thrown away everything, but it's my boss, Arsalan, who also leads the field engineering organization, who happens to be one of my audiences that I'm serving. And every time I met him, he would ask me the question, how do you know that these certifications actually lead to my solution architects being better? And I never had a good answer. And I did all this causal impact analysis and I'm studying the data to understand are they meeting more customers. And I've done this very, very successfully.
[00:12:01] Speaker A: I wrote a whole book on that.
[00:12:03] Speaker B: Exactly. So I did all of that and I could never feel good about the answer I was giving. And he was very gently pushing and saying, well then we should find something better and we should find something better. And it was not until last year that it finally clicked that we cannot just evolve what we were doing. We had to revolutionize what we were doing and flip it on its head. Like I said. Because we needed to understand is it that the solution architects were doing, where were they doing it and can I bring it to them in a way that makes sense. So we started off with first understanding what are the skills, right? They need to be builders, they need to be good customer facing individuals, they need to be good industry professionals. And then we said how do you know that they are good at this?
And so we first understood the assessment criteria while they were working and then we took that and we said, okay, if those are the good activities that are going on and I can see that, can I now give them the enablement exactly that they need? Because then I'd have causal impact to say that they were getting better at it. So we launched, we built this entirely on databricks, right? A tool that internally we call Skills Nav, Skills Navigator. And the idea is to say that it's just there, it's doing what it needs to do a skills inference engine based on this entire data lakehouse and it has the data of all of the solution architects. They can go in there, they can log in, they can see their profile and they can see what skills they have. This allows them to have the right conversation with the manager to say, hey, I haven't really worked on this tech, can I do that so that my proficiency goes up on that? But if they want to start with a training course or a certification, that is still an option. But we're not saying go to their
[00:13:54] Speaker A: that stuff is there system is inferring their skills from their activities.
[00:13:58] Speaker B: Exactly.
[00:13:58] Speaker A: And did you have to hook in things like Salesforce or other sources of data?
[00:14:02] Speaker B: Yeah, absolutely.
[00:14:03] Speaker A: So the benefit is you had the data for BRICS infrastructure to consolidate all these data sources. Yes, but I think, I don't know if you're ever going to go to market with this thing. But you know, just as a theory, theoretically, if you're Starbucks, if you're Boeing, if you're whatever company you are, you have these sources of information, but they're sitting around unintegrated. And if you could bring them together, you would probably know pretty well what are some of the skills and capabilities people needed.
[00:14:31] Speaker B: Exactly. And with products that databricks has, like GENIE ontology, a lot of this tagging happens automatically too. Like again, this is information that is sitting in your organization, in people's systems, in people's heads. And we're just pulling it all together using databricks, doing the inferencing on top of that and then providing learning.
[00:14:52] Speaker A: Let me touch a little subject that's been A big problem for me for a while. You didn't have to build a skills taxonomy to do this, right? The system or did you?
[00:15:01] Speaker B: We started with it and then refined it with the, with data.
[00:15:05] Speaker A: The system's telling you more and more what the actual skills look like.
[00:15:09] Speaker B: That's right. That's right.
[00:15:11] Speaker A: What do you think happens? Well, first of all, what you did is kind of unique to you guys, but there's a pattern here that many, many other companies could follow and I hope you do go to market with because frankly, everybody needs it. What do you think happens to content? Content companies, content curricula, the off the shelf stuff From Coursera or LinkedIn or whoever. Is that becoming irrelevant or just very small compliment? Where does that fit? Because in your case, I assume the R and D is going so fast that any course you build kind of has to get updated every month.
[00:15:44] Speaker B: That's absolutely right. There is a place for. So I would say complimentary is where I would start that courses still have a place, certific still have a place for job proficiency, entry level. But that's not making you productive and amazing at the job that you're trying to do.
So one content has to get updated very, very quickly. But the second important thing is the personalization element of it. So let's go back to my solutions architect example. If I am a solutions architect serving the, I don't know, manufacturing industry and I want to do a demo to my customer, I should get training that teaches me how to do a demo to a manufacturing customer using databricks. Right. So a genetic course only goes so far and then the human has to then translate all of that and do all of this work for making that real in their environment.
Today with AI, we have the capability to reimagine that you can create content on the fly that is tailored to the human being that we are trying to enable. So we imagine a world where content has a place, but we're not serving that monolith to any individual in its entirety. We think we're always customizing it, we're always tailoring it to exactly what the person needs in the format that they need it. So many of our audiences, your AI
[00:17:07] Speaker A: is doing a little bit like what we do in Galileo. You're dynamically creating custom content for each person.
[00:17:12] Speaker B: That's absolutely right.
[00:17:13] Speaker A: That's one of the things when you get into this whole idea of dynamic enablement and the source of content becomes a really interesting conversation. So you've got, in your case, engineers building new things, product managers, how do you get that content into your learning infrastructure? Do you do interviews of those people? Do they write it down? Do you capture their documentation and just import it?
[00:17:36] Speaker B: We take whatever we can get in most organizations documents. So starting with PRDs of documents, that's a great input source. But also our product teams also do these webinars. They'll share. They'll just do knowledge transfers. Whether it is for the support team, whether it is for field engineering. That becomes an input source into what we are trying to create. So any and all media or document asset becomes an input to how we're creating. Course, we still have humans in the loop. Obviously, we've built something that makes content really instructionally sound like that's still important. The value that an instructional designer brought to courses was great because that's how human learn. And we don't want to take that away. But we are saying that that's an input into how the content gets created, into how it gets localized. So we didn't want to lose any of that. But we're not trying to do that manually. All of that gets done through the system.
[00:18:35] Speaker A: Let me tell you a funny story that I think a lot of people on the podcast might relate to. Many years ago, I interviewed the head of.
I think it was the head of L and D. I forgot her job title at Verizon. I think it was Verizon. And we were talking about all of the products and packages that the cell phone companies create and how they create training for all of the salespeople in the stores on all the packages. And she said basically what they discovered was there was no way for the L and D department or whatever it was called, to keep up. We had to embed it into the product manager's job, that the product manager had to build the training before they launched the product. Yeah, it was part of the product. I mean, that was 20 years ago or 10 years ago. But what you're saying is that can be automated.
[00:19:19] Speaker B: That can be automated, yeah. And what if we could just put it into the product? So we also got the product telemetry. You used this feature, but you're not able to use these. And we've seen you struggle with this. Can I just.
[00:19:31] Speaker A: Then can you imagine in a company that's not a software company, I don't know if this. If you can think this way, but I bet you can. You're a services company or a product company or a phone company or a coffee company. Could you do the same thing?
[00:19:45] Speaker B: Absolutely. Absolutely. Because what we are trying to because data is still available in that organization. And with the power of databricks, what we're trying to say is surface, the learning for the human in whatever context they are doing, like I am today, maybe putting it out in Slack or putting it out in the CRM system. But for somebody who's making coffee or whatever, it could be in the portal that they are using to take an order.
So it doesn't matter. It's like. It's like help, just better.
[00:20:17] Speaker A: Yeah. I mean, one of the customers I talked to about this, because we've been doing all this work with Sana was Polestar, the automobile company, Electric car company. And what they told me was the way salespeople used to get trained is they would take courses, and then the customer would walk in the store and say, my car's doing this. And the sales guy would go, hmm, never seen that before. Well, you're on the new release. Well, we haven't been trained on that. So they would get. So the customer knows more about the car than the salesperson.
[00:20:43] Speaker B: The salesperson, yeah.
[00:20:45] Speaker A: So they're using Sana to fix that. But. But you would posit that you could do this in virtually any industry if you had the right technology from a
[00:20:53] Speaker B: data point, skills inference. Absolutely. Absolutely right. Like, let me give you maybe an example of a technology that we are extremely excited about. And even a user like me, who's not a data engineer or a data scientist, is using this. And this is how I imagine learning to happen as well. But I am a learning leader. I have a P and l. I'm doing customer education as well. And I have to always be looking at what's happening to my training classes, what's happening to skills development, who's where. And these are questions I have to think about every single day. And I just. I could depend on a dashboard, but then it's static. And every time my curiosity takes me to the next question, I'm waiting another week for somebody to refresh the dashboard. That's not how databricks works anymore. You're like, you go to Jimmy, I ask a question, and it sees a pattern that I come back with the same question, and I could create an agent for it. And now it's just doing this for me whenever I need it. But I'll give you an even more complicated example, because it's not just about pulling and querying data. It's actually calculating and giving me insights into how I want to think about it. So recently, we had to think about. We had to take a new Target. A target that was slightly more than what I had. It happens all the time. And I had a bigger revenue target. And I'm like, where do I go? And I could pull in my delivery team, my ops team, my curriculum team, my regional leaders, and I could pull everybody into a meeting to say, where do we tackle this revenue from? Right. Instead, I just sat down with Jeanne for half an hour and I said, what are we going to do? I need to take in this much revenue and what's happening with my business overall. And I came back and said, here's the deal strategy. Here's the attach rate you have. Here's how much you're attaching.
Potentially there is room in this business unit in the Americas if the attach rate went up. Here are the deals that are upcoming. And if you did this, this is possible. And so I took that at face value and I'm like, is this even true? Like, am I like, is it making it up? And so I sat down with my StratOps team, and then I'm like, should we just give them an extra headcount so that they can attach to all of these deals? And there is a potential here. And we didn't have to increase our class sizes or anything like that. And that was the go to for
[00:23:07] Speaker A: the customer revenue generator training. You're making me want to get databricks for our company. We're a little bit too small, but I think we probably need it.
[00:23:14] Speaker B: You need it. You need databricks. It's magical. It really is what it can do for us.
[00:23:19] Speaker A: So do customers. So customers learn through this GENIE stuff, right? So you sell this as a part of your service to customers.
[00:23:27] Speaker B: That's exactly what it is.
[00:23:28] Speaker A: What's their reaction when they look at this and then they compare it to the junky stuff they have in their own companies?
Do they. Do they say, hey, can we buy this thing? Are you. Are they asking you to get it? I would assume yes.
[00:23:40] Speaker B: So genie is an actual product. GENIE is.
[00:23:43] Speaker A: Okay, so you can buy it.
[00:23:45] Speaker B: Yeah, you can buy it. And this is what every business user, in my opinion, should be using it here and now. Right? Like, this is. You just get to results faster. And it's not just depending on data from the Internet. It's depending on your data company's data. And it has the context and it has the ontology to make this real. Right. Like, that's why it knew exactly which bu. Which deals where I could make the revenue.
[00:24:13] Speaker A: Okay, well, this is another way for you guys to sell into the HR departments. Of companies.
[00:24:18] Speaker B: Exactly, exactly. Yeah. I think as an HR leader this is again really gold. Our HR team uses this all the time as well.
[00:24:25] Speaker A: So one more question. I'm sure you get thinking about this a lot. I talk to a lot of L and D people and I've been surprised at at how slowly they're reimagining their roles. And they're a little bit afraid of it, I think maybe intimidated by it or skeptical of whether AI could do the kind of work that they do. What has been your experience in the traditional L and D person, instructional designer or whatever they may be called getting this. I'm sure there's many, many new things for them to do. But what's been your experience with that transformation of the people?
[00:24:57] Speaker B: I want to say that that getting started is the hard bit and once they are with it, it is easy on from there on. So everybody, it's not just the HR folks, it's people in finance, it's people in marketing, it's people in the field, everybody, when they come up with this first technology, a large chunk of people. Right. Let's say greater than 60% of the people are first. Like oh my gosh, what is this? And this is going to take over my job. Right. As an enablement professional, I think we focus a lot of energy first disarming that. Right. Everybody's going to be better when they're using this technology. And you are still the smart person in the room, you are still the human in the loop that is critical to this job. So we don't fear AI, we fear the person who's going to be using AI. So how can we be the person who's going to be using AI? So I think enabling everybody in the organization and again, databricks has done a tremendous job. I've worked with a lot of customers as well who we are helping them.
[00:25:56] Speaker A: Do you tend to look for people who are dyed in the wool? Instructional designers?
[00:26:01] Speaker B: Absolutely.
[00:26:02] Speaker A: Or do you look for people who are technologists?
[00:26:04] Speaker B: No, actually business users, non technical users. In all honesty, Josh, this is like really interesting, but I think the technologies tend to overthink the AI because they're wanting to engineer it, but the business users use it much more.
[00:26:18] Speaker A: You need to have a business context around it. What is the problem we're trying to solve here? Not like how, how snazzy can we make this?
[00:26:24] Speaker B: Yeah, exactly. So I see the non technical users tend to use AI much more disarmed and do amazing things. I've had program managers on my team build apps for Managing events all on their own. And I'm just like, this is genius.
[00:26:41] Speaker A: It's such a big topic in our company. Most of the people working on Galileo don't have technical backgrounds at all, but they understand our business really, really well.
[00:26:49] Speaker B: That's right.
[00:26:49] Speaker A: Yeah.
[00:26:50] Speaker B: So we should enable them on databricks and then they can be like, building for their own stuff.
[00:26:54] Speaker A: All right. We can turn this into a sales webinar very quickly. One more quick thing that I really want to poke a little bit on. So one of the sources of knowledge and information is subject matter experts out doing things.
[00:27:07] Speaker B: Yeah.
[00:27:08] Speaker A: How do you capture that information? Or do you in this learning infrastructure?
[00:27:13] Speaker B: So subject matter experts continue to be really important. Like I said, we make it easy for them. You were giving an example a little while ago of how it became the product manager's job to build training. Well, they're not going to be very good at it. Right. But what they are very good at is understanding their product. So like I said, we are happy to take. If they just want to come and present a webinar, they do that. And then the app that we have, the infrastructure that we have for learning, converts it into a learning material that works for the user. It could be a podcast, it could be, you know, it could be a. A document.
[00:27:47] Speaker A: You have a very active subject matter expert sharing program to capture that content. Yeah, exactly. Yeah. I mean, I think that process has been so difficult in the past that companies don't have enough formal ways to do it, but now it's so easy. You can just have so easy lunch and learn types of things online and you get all that content. Okay.
[00:28:04] Speaker B: Yeah. We just want to capture what's in that head into any kind of format and then we'll take it. And the human in the loop, like the final review then becomes so much easier because we're like, record it at your own time. You know, don't even worry about the quality.
[00:28:18] Speaker A: One more question and then we'll wrap up. Imagine you were in the leadership development side or the soft skills development side. How would you apply this technology and this approach to that?
[00:28:29] Speaker B: Exactly the same and slightly different. Exactly the same in the sense that I still want to provide it in the context of what they're trying to do and where they're trying to do. So take an example of a manager. And we've done this again in our world is imagine a manager who has to give Curtis feedback to their employee. This happens during perf. It doesn't matter when you took the training. It's when it's happening is when it's the hardest. So we trained all of our managers on my team right before Perf with a real live role play. And the character in this situation actually was part of the team, understood the career ladders, so was trained on that. And we were trained doing feedback on giving them critical feedback on performance.
So whether it is soft skills, leadership development executives, we're doing a lot of. And not saying everything is replaced by AI. Again, humans are still critical in participation and building confidence, but it doesn't like. I think leadership development training is actually one of the best that we can now scale with AI better than we've done before.
[00:29:35] Speaker A: Well, let me run. I have a hypothesis that what happens when you do that is you start to encode the cultural values and the behaviors of your company that are unique to you.
[00:29:46] Speaker B: That's exactly right.
[00:29:46] Speaker A: Is that correct? Do you agree with that?
[00:29:49] Speaker B: That's correct.
[00:29:50] Speaker A: I mean, that may be one of the biggest ROIs of all of this stuff sometime in the future.
And we have a management culture embedded into the AI in some sense.
[00:29:59] Speaker B: Yeah. No, that's absolutely right. Like, everybody's using AI and you're training them in the context of your company with the cultural values of your organization. Yeah.
[00:30:09] Speaker A: All right, Rochana, thank you so much for your time today.
I would love you to come to our conference next year and we'll do some more things together with you guys. And we'll look at databricks for us.
[00:30:19] Speaker B: Yeah, yeah.
[00:30:20] Speaker A: And we'll tell more people about it. Congratulations on all that you've done, and I'm just so excited to have been able to work with you. Thank you again.
[00:30:27] Speaker B: Thank you, Josh, for inspiring us to do this AI performance engine or for setting the vision and for databricks to actually making it happen. So we've loved reading your work. So continue to inspire everybody to do better. Thank you, Josh.
