Rebuilding Marketing for Agentic AI

So I have with me Liz Dobbs here. And so I'd love for you, Liz, to introduce yourself and just tell us a little bit more about Databricks. She is the AVP of marketing technology, growth, and data at Databricks. But let's get... Kick that off.

Awesome. Hi, everyone. Thanks for having me, Prachi. I am very fortunate to lead the technology, data, and growth team here at Databricks.

And my responsibility is really spanning everything from kind of the back end of marketing. So everything on ingestion of our data, processing of our data, agents, and utilization of the technology for our marketing team, and actually more increasingly GTM as a whole as we're seeing marketing and the context that we're providing become more and more critical to our customer experiences. So very fortunate, and I've been here for about seven years. And I have the opportunity to work with an amazing team every day who is kind of pushing the boundaries of what's possible.

That's great. Thank you. So, Liz, one of the things that I loved about...

We've talked a couple times before doing this. One of the things that I loved was usually when you're talking to, you know, marketing leaders, there's a lot of, like, vision and, like, future state and, like, the story, which is all true.

We both believe in it. Yes. But we also had a lot of real talk, which I loved. And so I'd love for us to not only share the vision, which we should, but also share some of the challenges and opportunities that you see.

So Awesome. We are gonna go go there. But maybe I'll begin by saying that every marketing leader is...

You know, I think there was a time two years ago, maybe three, where everyone was experimenting with AI, and that was okay, and that was fun, and, like, we had budget against it.

The keyword, budget against Budget against it.

Think today, it...

We are past the experimentation phase a little bit

And have found recipes, or at least there is expectation that we are gonna scale agentically, and so marketing leaders are being asked to do more while keeping resources optimal. And yet there is a lot of experimentation that still needs to happen. Yeah. So I'm curious how you think about balancing experimentation, scaling agentic operations, and running the day to day business Yeah. At Databricks.

Yeah. I mean, I I think the... You know, I always actually allude to this story as, like, a hero's journey. If everyone took English lit back in the day, you know, you all, like, you started this journey. Everyone's so excited. They're gonna go change the world. I think that's how we all entered our AI experimentation phase and then like, I think every company hits this is you hit the trough of despair.

I like that trough of despair.

You can't actually like put AI on a bad process. It just like fundamentally doesn't work.

And so I would say like, you know, we, like every other company, hit our trough of despair, which is ironic because we are an AI company. So here we are in the trough of despair in marketing. Yeah. And then I think that we're we're trying to climb out of with using the, you know, trials.

Like, what are the things and the steps of cramming the trough of despair? And I think for us, you know, using AI is obviously...

We're we're big believers in it. We think this is the future. We have a vision of where we need to go in marketing, and we do think it's kind of this complementary human and agent together with agents doing kind of the redundant manual work and service of freeing up creative capacity for our marketers to do all of the things that they can do that agents Absolutely. Can't But, you know, to your point, we can't experiment, you know, freely.

Money doesn't grow on trees anymore.

Like it used to. And so I think where we're really focused in marketing is experimentation and service of something bigger. So when we run pilots, do we...

Maybe we don't know the exact way to measure it. We don't have all the numbers kinda perfect. Yeah. But we know the shape of the problem and we know the shape of the goal.

So when we're doing experimentation, we're at least doing it in service of something bigger. And that's kinda how our approach has been, you know, within marketing because as you said, we're at a very rapidly growing company. And the CEO always told me like, a good leader is always kind of fixing the plane while you keep it flying. And I think that's...

The idea of that has been very much challenged with AI. Yeah. Because what's changed is, like, not only do you have to do your day job and fix the plane while keeping it flying and marketing, you also need to build a spaceship on top of the plane while you're doing all that. And that's where I think, you know, we're seeing so much creativity, but, like, friction and and challenge happen.

Are there any wins that you found thus far where AI is working? You can scale it. You could kinda, like, start the building blocks of your rocket ship.

Yeah. So we... You know, I think for for us, fortunately, we're a data company. And so one of those key things we have in marketing is when, you know, when we're talking about what to do, how do we wanna do something, it all starts with what does the data say. And as a small marketing team, we're a very... Like, we have, know, a little bit more than 200 people at a 12,000 person company. It's a very lean team.

But as part of that, like, the asks of getting the data for all of our decisions across the team was very slow. And we realized that was kind of a big bottleneck to our agility in marketing.

That is true, by the way, for every marketing team. So you're not unique there.

And so we actually, you know, have the advantage of being a Databricks, we have a lot of great products that are not real yet. Like, we don't sell customers. And one of those was actually Genie two years ago. So now it is a product that we do sell, but back in the day, it wasn't, and it was kind of a dog food thing. Yeah. That someone on our our data team was like, hey, I think this actually... There's there's a... They're there.

They're there. Yeah.

And let's figure it out because we're not gonna get more analysts. We're not gonna get more resources. And my six and a half, almost seven years working at Databricks, I've never built a dashboard my CMO can remember the URL of. And so we needed a different way to get the data in the hands of our marketers.

And so I think that was our first foray is building kind of a a a data smart agent that had the business context of marketing, all the slang all our teams used that we really instilled trust in Yeah.

To help kind of change using agents the way our marketers were interacting with our data. Nice. And then the goal of that was kind of exceeding up decision making, making sure our teams could get the answers when they needed it in the moment. I think so that was kind of a big bet for us. And I think what that's freed up is a lot of capacity that was spent building decks Yeah. And, you know, emails for tickets and waiting and pinging.

Not asking questions, answering questions.

And, you know, we have a big builder culture, and I think it's freed up a lot of time for our marketers to start tackling agents and things outside of the direct data part. How do we tackle customer references?

We built something ourselves because we now have the capacity for experimenting on those Interesting.

Yeah. Yeah. Yeah. No. That's amazing. You know, one thing I will say on the dashboards thing Yeah.

In Asana...

Do you remember your URL? I'm sorry. I...

No. I don't remember my URLs, but we have a a task in Asana assigned to me that I've, like, started. So all the links in that place. My team was just as tired.

There you go.

And we do use data genie now.

Oh, that's... See, there you go. Yeah.

CMOs love the marketing genie.

So one of the things, you know, we've talked about AI in terms of, you know, workflows and operate... Operating efficiency. Yeah. But there's also an external world for marketing that is changing in terms of how buyers are buying and where the research has happened, and we've all been spoiled by search. Yeah. You know? Search was like, you get the click, you get the person. Today, it's... There's so many acronyms, but let's just call it G E L.

Yep. Yep.

Has has has that been a challenge or an opportunity for Databricks that you all have done anything about?

I think it... I mean, for me, as... I think it's actually one of the most exciting new frontiers of marketing is tackling agent search. And for us, it is... Well, actually, biggest marketing priority in the second half is us figuring this out. So definitely a priority for the company.

But I think tying into the future of work thing, it was actually quite challenging for us because we...

First, it's new. And it kinda reminds me back in the SEO days when I was working on a web platform, it's like, you remember all the white People used to hide all the text on their pages to like show them inserts. People were just hacking it and like, it was really fun. Like so...

So many hacks. I think it like that's kind of the era where it went to you. It's like kind of fun and hacky and then like one day Reddit's in and then OpenAI's Reddit's out. And then you're like, what happened to our Reddit program?

So I think it's it's it's a lot of fun. And I think that's what...

You know, the fun of marketing is keeping it interesting. And so I'm always like very drawn to projects where we're figuring it out. And I think GEO is a really interesting one because first thing we need to figure out is the data because we didn't...

We we knew it was important, but we didn't know were we showing up well or where are our opportunities. Like, we don't know. So we, you know, work with some partners and ingested all of the data from the AI engine so we could figure out against the prompts we know. That's the second level question we haven't answered, which is what are the prompts we don't know?

But is that all in Databricks for you?

All of it. Yeah. We're pulling all in Databricks. And then so the prompts we do know, how are we showing up? And so we kind of not, like, super quickly, but pretty quickly figured out where the... The shape of the opportunity. And then we're like, great. Okay.

So now let's build content to address these prompts Yeah.

And so we can improve our rankings.

But you know what?

That's really hard. Yes. Content development and marketing was spanning across like 12 teams. And so, you know, it come from product marketing, and we had technical marketing, and we had the web team. Each of them, the create, you know, the content team, the creative, like, it just had to get so many steps. And when you realize the velocity and the speed of the prompts and the changing, like, that traditional distributed workflow of content development just fundamentally does not work if you wanna move the needle in GEO.

Was it too slow when you said... Slow.

But it was also because we were building content for people. Yeah. And so Yeah. We kinda had to challenge ourselves to be like, hey, we now have to build contents for agents. This is a new audience for us.

And so what does that mean? And so we actually, in marketing, have a new team under our corporate marketing team called our agent marketing team.

Oh, interesting.

And so we're pulling in pieces. It's kinda like a formalized pilot really of pulling in pieces of people from who were in distributed teams together Yeah.

Okay.

To kind of rethink the content life cycle because now it's a new audience. And if we're building content for an agent, does it have to go through our copywriting team? Probably not because people don't really read them.

They just, you know, it shows up as a Yeah.

A snippet in OpenAI or JITPT. So I think that was the the challenge for us is, okay, so we had to figure out the shape of a new problem. We we knew the opportunity. We had to rethink the...

How...

The content we wanted to build. Okay. And then ultimately rethink the team building that content. Okay.

And so what that's kinda led to for us is actually a total rebuild of the content factory as a whole. And so now we're we're much more kind of trusting our AI, building in the brain, the prompts, the guardrails, all the context we have in marketing to deliver at scale hundreds of pages that we're publishing directly every week. Interesting. So we're trying to build in the agility.

And it's kind of the thing. It's like, it's not gonna be perfect every time. It will make mistakes. AI always does.

Yeah. But it's kind of the the risk, the tolerance that we have for this is is much higher, I think, than traditional marketing. Yep. We're trying...

We we think it's such an opportunity and we're trying to figure it out.

That's great. I love that. So we at Asana also have AEO, GEO written content kinda established in that way. Yeah.

One of the things that I'm curious is like, what about video content and influencer content? Because in the current theme of Yeah. You know, AI search, those things seem to be super critical. Curious if that's something you all are tackling in the back half, if you have thoughts.

Yeah. I mean, it's definitely something. So we... It's funny. We...

I think have started to realize that we as a b to b company need to take a lot of cues from b to c. Because I think b to c has been really winning for a long time, the brand war, the influencer war. I think b to b, we've been very like, how do we get really great emails? How do we do really great like e books? And those in like the grand scheme of things where, you know, these LLMs are not really rewarding building this written gated content anymore.

Who's figured this out? It's b to c companies.

And I think exactly to your point, it's like, yes, they've been doing influencers for Years.

Years.

They are full agencies who do influencer marketing, and it's kinda crazy.

I know. And it's been a new frontier for me to learn a lot about and I've worked very closely with Jenna who's our, you know, head of the corporate marketing department to kinda go on this journey with her because even a lot of these agencies, they still operate in a very traditional agency model. Yeah. They're like, hey, we'll give you the data.

Yeah. And I'm like, Jenna, no. Like, we need the data when it comes in. So how do we rethink, like, what success in inter...

Influencer marketing is in b to b? Yeah. Is kind of a very new frontier for us. That's great.

So exactly to your question, we're we're trying to expand the channels Yeah. We are making Databricks known on, and we're doing that. Influencer marketing is a big bet for us. Brand, we're moving a lot more towards brand.

Shout out to the Cal Bears if anyone saw the football stadium. That was new for us this year.

It's fun.

But I think, yeah, this is a this is a definitely a new bet for us, and it's all because we think GEO and the influencer and the way buyers are buying is changing, and we're trying to figure out an experiment to change with that.

I I love that. I love the idea of creating a separate team that is focused because one of the... Otherwise, it's hard to get someone to focus. I also like that.

Haven't done that. Okay. So we we do... It is a top priority for us, but we haven't carved out a team that just does that. So I like that idea.

We do the content production. Yeah. One of the things I'm curious, so it's like, you're right. Like, you're just generating content for AI to read, like, that's great.

It could be anything.

Yep. But today, a lot of people could also land on that same content because it's a... It's still a combination of search and AE. And as content production becomes...

I don't know. It's like everyone... Like, the bar for, you know, everyone's gonna use agentic content production. It's... Where do you see competitive advantages moving? Like, right now, you're early, like, you'll still have the competitive advantage. But what's, like, the next, you know Yeah.

State of that?

You know, I I think it's it's... Comes back to what we're just talking about because I I think in... At least in my job and probably a lot of your guys' jobs in marketing, we've been chasing the white whale for twenty years. And the white whale for me as a technology person is like, right content, right person, right channel, right time.

And that is like always been this thing that how many tech vendors have promised, hey, figured it out. Yeah. And I think actually for the first time with agents, we're actually...

It feels like that is the future. Yeah. And I think for so long, we've anchored on our technology roadmap as the thing that's gonna differentiate us. It's like, hey, we get technology right, our marketing's really gonna shine.

Yeah. And I think with agents incoming, I think that's actually becoming...

We have to assume everyone's gonna figure it out. Or now all these companies are gonna be able to get there more quickly than ever. And I think good marketing now, if you see good marketing happen, it's so easy to imitate. And so you'll see a message that you're like, that's a great message.

I love that. And then you'll see that next message in twenty four or forty eight hours Yeah. On three other companies Yes. That they're using.

Yes. So I think good marketing is now becoming very easy to imitate. And then I think what we see on the technology side is these capabilities that used to be so hard for IT and marketing tech company...

Companies to pull together are becoming...

Up.

So I think that kinda challenges us to say, okay. So what is that differentiation road map? Where is that gonna come from? And I think what we're we're gambling on is it's gonna be the creativity marketers are putting into the message and the brand and the influencers and trusted sources.

So I think where we we really think the future is gonna be is actually gonna be a great content roadmap, not just a great technical roadmap. Yeah. Because that content and rethinking the channels that we're getting our message out on to ones that are not as saturated Yeah. I think is gonna Yeah.

Be the Because, you know, we were talking last night. We...

I...

Someone could have cracked the white whale for marketing and be sending me the world's best email. But honestly, get so many emails, I don't read them. And so so it's just kinda like this chicken and the egg thing of like, even if we do char...

Like, it's this How we get the customer. Emerging messaging.

Like, that's the kind of stuff we need to test and try.

No. I love that. And I think, you know, I was I was chatting with somebody else yesterday and I was telling them, like, the... Some of the fundamentals of what works still don't change. It's like, are we providing customer value? Like, at the end of the day, if you... Whatever content you produce, whatever... You know?

If that's providing customer value, it's moving the needle Yeah.

Then it's gonna work. Agreed. Distribution channels are different and we gotta figure that out. But...

Yeah. And I I would say the same with AI is like it comes back to you under the three fundamentals of of been true forever. It's people, process, and technology. Just because it's AI doesn't mean there are shortcuts in any of those.

That's true. That's true. I love that. Okay. So since you're talking about people, process, technology, I'm gonna take us in that direction. I know that is a project runway that is something that you've designed working with the Asana team.

I was very intrigued when you were talking about it last night. So tell us a little bit more. What is this project? What's the vision state? And, you know, where are you at right now?

So I think I think our Databricks team will commit to that we do not have it all figured out when it comes to workflows.

And one of the big bets we saw with agents and the big opportunities is like, hey, how do we take this manual work off of our end team? Yep. And then we realized, hey, it's actually not in a complex enterprise environment. It's not one team that owns this.

It's every operations team and every function actually owns their step on their journey....

Flows. Yeah. And then it wasn't good. We weren't gonna be able to make material progress because as we probably learned in you guys, slapping AI on a bad process doesn't get you a lot of automation benefits.

Like, it seems like good on paper, but, like, in practice, you're not actually saving that much time. So earlier this year, we kicked off a cross functional project with Asana Services and our marketing team, especially our operational teams, to really rethink kind of our end to end process all the way from campaign ideation to closed loop reporting at the end and what are all the steps in between. And as just part of that, we realized we've been using Asana...

You know, I've been Databricks for six and a seven years, probably for ten years now maybe we've been a customer of Asana.

But we've never actually had an owner.

And so we realized that every team in marketing, even within marketing, was operating out of tasks or sub tasks or portfolios, and there was no consistency in how we were making Okay. The set And so we realized to get kinda some of the benefits we wanted and to create...

Take advantage of all these great announcements you guys were making today is, like, we needed to, like, actually go down to first principles and say, what was the work that needs to be done? How do we wanna get it done? How do we track it? And how do we really make sure that we're all operating out of the same playbook to then be able to kind of rethink the work within the playbook?

And so that's kinda how Project Runway started is it's a multi tenant purpose of a bunch of teams coming together, documenting our work, looking at the most complex things we do in marketing end to end, step by step, team by team. So we, one, can actually start operating at the same altitude in Yeah. In Asana. And then secondly is, like, let's look into the work and all those steps and say, hey, are they still necessary?

Yeah. Or is there an opportunity for us to re swizzle and not be so functionally siloed and think more on project milestones? Can we parallel path the work where we had been very linear before? So those are all of the things that are going into Project Runway.

I mean, we haven't...

It's not figured out. Still a journey. It's a big me. Yeah. We're we're going on it.

But I think I feel really good about the fact that, like, the hard thing is understanding the work first. The easier thing, I think, then is Yeah.

Changing the You know, as you kinda embark this journey and we're trying to figure out, like, okay, what are the, like, trickiest things or the easiest things?

How did you figure out what is your, like, long list of, like, different workflows, I would say, to solve for, to even prioritize for a project like Project Runway. Was there a method to the madness?

Squeaky wheel was the technical term. No.

I mean, I think we knew where...

We knew the things that took us the longest time, and those are the things that are probably like any other marketing team is like us running a large scale webinar in multiple languages and simulcasting. Like, those things are like the things that we take the most prep all the way through execution. Yeah. So I think for us that's really we started it.

And then like where we saw the teams just really feeling like this was super onerous and they were spending so much of their time and very vocal about that. Okay. So we just kinda wanted, I think as part of this too with AI, it's a lot of hearts and minds. You wanna get some wins.

You want people to feel the benefit. Yeah. And they want...

You want them to feel like it's different and it's working for them. So I think that's technically how we started, but I think Yeah. What our goal is now is...

Well, we're starting with these. It gives us an opportunity to go all the way horizontal to all the work. So now that we've done, like, hey, we have to send an email. That email step is applies to so many campaigns.

So That's true.

I think it's easy to know once you've done the hard thing to then extrapolate into the less complex Yeah.

Okay. I like that.

We talked earlier about, hey, you know, while we have some things that we know will scale, will work, we are still gonna have to experiment quite a bit with AI. And that is an investment of time, an investment. Yeah. It's an investment at the end of the day. Yeah. How do you decide where do you invest, how much you invest to get to that, you know, end state?

Yeah. I mean, it's it's a great question.

I would say with us, I think they're... The the goal of us, as I said earlier, is in in AI, I think we have to celebrate the failures as much as successes. Because I think we, as a very fast growing company, it's easy to celebrate the wins. And then if everyone thinks that all you get recognized for is a win, you're almost hesitant to experiment because you're like, I don't wanna be that person who like messes up and blows all my tokens on something. I have to... We're sitting at dinner last night. Yeah. I don't wanna ask my CMO for more tokens when they are gonna realize that I blew them on, something doesn't work. But like, I think the fact that we're learning something and we're trying something new, like, we have to celebrate the fact that someone found a problem they thought was so meaningful and tried to solve Yeah.

That's great it didn't work, but then ask for help.

And so I think back to to your point where we invest is like, I think the more we can free up of time of people in the manual things, the more they can creatively start solving those problems. And they're gonna sell us the problems that maybe on paper we don't know we have. Because they're...

It's almost like not like suffering in silence, but like they're doing this manual work all the time. They're probably not complaining about it. But those are where the hidden context of work Yeah. I think is where we can really move the needle.

So back to your question about where we invest, I think we need to follow the pain a little bit and make sure we're the pain, and the pilots are in service of the North Star. Yeah. So I think it's a little bit of the bottoms up innovation that the teams hopefully will have capacity to start tackling themselves, mirrored with the tops down transformation that we as a marketing org know we need to go on. Yeah.

And so, I mean, it's a little bit messy in the middle, but I think that's where kind of we need to continue to really surgically bridge that gap.

Yeah. I like that.

You know, you you talked about creating a separate team for just the content factory and, you know, agentic agentic content, I would say. Are you thinking differently about who you hire? Like, the profile of people? Like, are you seeing a difference on the team? Just, like, curious. Is that... Is there any change there or not much?

Yeah. I mean, I think in our in our operations team, we we definitely... We went through the full exercise of going through our job letters and saying, how does AI change the expectations of, you know, a manager, a senior manager? What are we expecting them to be able to do and contribute using AI? And really building AI fluency is very native as part of our jobs. Yeah. You have to be able know how to use Cloud Code or GCode for us. Like, that's just part of the job now.

And I think what we're figuring out, and I don't... If anyone has a great way, let me know.

But how do you interview for that? Because I think there's the critical thinking that has always been, I think, really important to interview for is, like, how do you think critically? How do you show your judgment in making hard decisions?

But then it's like how do you...

For AI, you know, not just, hey, I have a great skill that sends me a daily report of leads in the morning. It's how do I build a system? How do you show the systems thinking? Yeah.

And I think everyone in marketing in the future will have to be a systems thinker. Yeah. You can't be Design thinking. A deep specialist anymore.

That is...

It's it's like the the jobs are getting wider. Yeah. The skills are getting wider.

Yeah.

And I think the ability for people to think in systems is gonna be, at least for me, like where I think AI is really gonna shine.

And so that's kinda how we're thinking about it, but I think No.

I I love that. Yourself? Yeah. I mean, I think for us, we...

There are two clear things that I'm seeing. Yeah. One is, to your point, when when AI first came on the scene, like, this is, I would say, '24, early twenty twenty four, we started experimenting. And the default was, like, let's start with the marketing ops team.

And, like, they can figure this out for the entire team. I...

You know? And the reality is that builders can live in many different places Yeah. In your entire team. And so we started seeing pockets of success through the PMM team and other teams in addition to mocks Yeah.

Yeah.

You know, doing their stuff.

And so, you know, it's like, how do you look at the builder mindset? Like, who is builder? And I think that's like curiosity, willingness to try and fail, you know, and then you have to create a culture for it.

But also systems thinking, design thinking, like, have to think about the end to end process to really get AI to work. So those were some of the things that we did. And then on the on the other side, it's like you have to create capacity for everyone else to skill, to understand, you know, how to use. I think once you create that, people can adopt.

Mostly in most teams, if you don't inherently have that or at least that's what I saw when when AI was new. Now it's kinda more everyone's using, but it's still true. If you don't have the inherent, you know, drive Yeah. Then you don't you don't create capacity.

And nobody has capacity. Everyone Yes. Has a full plate. And so I think like, you know, that's one of the things we talk about actively is like, do we keep creating the space for the team to keep experimenting?

So I don't know if you've, you know, you have ideas for capacity creation because there is a business to run. Yeah. But that's definitely something that......

Say we have too much capacity. We have a...

No. I mean, I...

Yeah. I mean, I think it it is it is a challenge. I mean, we...

I think we have a...

As a team and our leadership team have tried for years to we have it like focused Fridays, like no meeting mornings on Fridays. Yeah. But of course, like, that is respected by nobody. But it's on the calendar, we can say we checked the box.

Yeah. I know. But I mean, I know. I think it's really hard and I think it's funny.

I see myself probably modeling not healthy behavior where it's like, I kinda have a minute. Like, I gotta leave, you know, to pick up my kids from school, put them down. They're young. I am online, and I'm watching my favorite trash TV show.

I'm also the founder of our trash TV Slack channel, Databricks. But you know, I'll have a glass of wine and I'll be watching some mindless TV and I'll be like, oh, what was that idea I had? And then I'll like just start Yeah. You know, going in it.

Yeah. And it's like, you know, it's like, so I think carving out part of my day I haven't been successful with and I need to newborn of an effort. But I have found that like sometimes just having space Yeah.

And like something that's space outside of work Yeah.

Is where I'm most creative.

Yeah.

And I will Slack things to my team, which they will attest. White wild hours with some crazy thing I've built. Yeah. And not that it's like always a great thing, but it's I think it's the ability to prototype ideas using AI in new ways has been such a differentiation for me to I think better crystallize Your own Concepts.

Yeah. Concepts.

And I see that across our marketing team is like it's so much easier now for someone to say, I had an idea and I pulled this thing together so you can actually see my idea too Yeah. Versus us writing a three page business requirements document which, you know what I mean? Like, someone reads and more things on and then they build like two things of that.

I know. Or completely misunderstand.

Exactly. So I I think it's it's been a great for us to kind of have a new way to kinda communicate our thinking.

I love it. Okay. We are almost getting to time. I will ask one additional question Yeah. Which is, if you had to think about, you know, where to get started, what is the easiest way to, like, take all this that we are talking about today and put it into action, is there, you know, a blueprint or a recommendation that you would make for our audience here?

So I I think something that I've always loved about Databricks is one of our our corporate cultural principles is this idea of, like, first principles. So why are we doing something to begin with? And I think with AI, it's so easy to, like, skip first principles thinking and just jump straight to a solution. And so a thought exercise our CEO, Ali, has always kind of instilled on us when...

Especially when something happens, it's not good at the company, is is this idea of a five whys exercise. So I don't know if anyone's familiar with this idea. But essentially, it...

It's...

Ask why five times. So say, like, you know, we had an email was sent to with the wrong content to an audience. Yeah. Why did that happen?

Oh, it didn't get caught in QA. Why didn't it get caught in QA? Because, you know, that person had just skipped the task, didn't actually do it. Why did they skip the task?

Oh, they're...

They had a really big workload. You know, so you actually follow Exactly....

The problem. And so I think the five whys exercise is really helpful, especially in AI, because you kinda come down to three buckets of work and the answers are kind of, I think, in three things.

One is it might not be necessary anymore because agents can do it. And Yeah. And, like, so you should automate it. Great. It might not be necessary anymore because it doesn't need to happen at all, and it's just a remnant of bad process Yeah. Because our company's grown so fast, we, you know Yeah. It's all bubblegum and duct tape.

And we realize that, oh man, this is just a manual step that maybe doesn't need to be there. Yeah. So that's kind of bucket two. I think bucket three is like you realize this is strategic valuable work.

And we haven't made it clear to people that like, we actually need you to pay attention.

Yeah.

So I think that's why in AI and systems thinking, just asking why five times gets you to an answer where you're like, I think it's a little bit clearer. How do you solve that problem? And then you can put AI or not put AI on top of the problem. Yeah. That's kind of something I would recommend.

I love that. I love that. What I'm hearing, if I had to like think of like the things that we went through, it's that agents don't fix broken processes. So I would say remember that, like work on the process first and then the agents can come or not come later.

I also think there's value in creating capacity before you drive adoption. So like we have to work on figuring out how we would do that. And maybe a third one is like, you know, I think we talked about this this at dinner, you mentioned this right now. It's transformations at your scale, especially, don't come with a finished playbook.

Yeah.

So it's okay to not be perfect. It's okay to start where you're, you know, starting and, like, just get the ball rolling. So those were my takeaways. Again, Liz, thank you so much for coming here or doing this. Appreciate this.