Paul Lewis, Pythian | Google Cloud Next ’24
[Savannah Peterson]
Welcome and Introduction
Good afternoon cloud community and welcome back to day one of Google Cloud Next here in Las Vegas, Nevada. My name is Savannah Peterson, joined by my fantastic co-host John Furrier and Rob Stretch, a gentleman looking sharp today, feeling smart.
[Rob Stretch]
Keynote Analysis and Data Celebration
Yes.
[Savannah Peterson]
We got a great keynote analysis, a lot of a precursor of the conversation we’re about to have.
[John Furrier]
Bringing All the Data Out
And all the data, we’re bringing all the data out.
[Savannah Peterson]
It is, it is a data celebration. I feel like, and we couldn’t have anyone better to talk to us about data than Paul.
Welcome to the Show
Paul, welcome to the show.
[Paul Lewis]
Thank you very much. I appreciate it.
[Savannah Peterson]
How’s it starting off for you? How was the keynote? What’s your initial impression?
[Paul Lewis]
Keynote Impressions and Shift to Gen AI Agents
The keynote was amazing. What I really loved about the keynote is this obvious shift from last Google Next, right? The obvious shift was Gen AI was applications. You’re going to build these new applications to create amazing business value. In fact, you might be a first organization shifting that to, you know what?
[00:01:00]
That’s a lot of energy. Maybe it’s agents. Let’s talk about smaller snippets of things that I can create value in much shorter time with technology that’s at hand, right? And you saw a bunch of announcements to say, here’s data agents, here’s security agents, here’s workspace agents, employee agents, customer agents. That’s, that’s kind of the shift, right?
[Rob Stretch]
Use Cases and Data Analysis
Yeah. And it’s really about the use cases and how, how people are looking at this. Cause I, and we even with our partner ETR have been looking at a lot of the data and they’re looking at the different use cases. And a lot of them were talked about this morning, you know, from code development to marketing and content. What are you, what are you seeing from a organizations as they lean in to AI?
Organizations Leaning into AI
Because they’re trying to justify it in a big way.
[Paul Lewis]
Internal Use Cases and Code Assist
Well, it’s much easier to think about the internal use cases first, right? And even much easier if you’re even go outside of it. So the very first set of use cases is code assist, right?
[00:02:01]
They’re already going to stack overflow to get code. Here’s another way to do that where I don’t have to go to stack overflow and it’s already connected to, to their IDE already connected to the code repository. So those are easy out of the box functions, sort of number one, number two is productivity tool workspace. So let me draft an email. Let me draft a document. Let’s do an image. Those are kind of the easy out of the box ones.
[John Furrier]
Paul’s Role and Pythian’s Focus
Talk about your role context before we go further, because you have a unique background in operating side of it, multiple generations of innovation and your company’s in a unique position. Explain what you guys do for your company. And then your role now, as you look back at your previous role and responsibilities and you see the gen AI wave coming at you, want to get, want to dig into like what’s coming and what can we learn from the past or not learn from the past or throw away from the past, but start with what you guys do first.
Pythian’s Data and Cloud Services
And then we get into that.
[Paul Lewis]
Sure. Sure. So Pythian is a data and cloud services company. So we spend a good portion of our time with the primary data sources or our customers, tens of thousands of databases, data warehouse, data lakes, where they’re not just doing their insights, but their operational OLTP database things, right?
[00:03:10]
Their core assets, we’ll call them. And we’ve managed that for 26 years. I’ve been here three years prior to that. I was at Itachi for a decade, very OT centric, right? Nuclear power plants and bullet trains and lights out manufacturing plants, what it means to stop a train and make sure the health and safety of a person. And then 17 years in financial services and actual CIO of 5,000 workloads, highly regulatory environment, you know, a consumer of technology. And that’s in many ways still where I come from.
[John Furrier]
The Next Big Thing
You’ve seen the movie before many times. Yeah. This complexity, the next big thing is coming. Where is it? Is it a really a big deal? I mean, I mean, obviously the gen of AI thing is, what is the real scope, the order of magnitude impact of this next wave?
[Paul Lewis]
The Scope and Impact of Gen AI
It will be big. It’s hard to say it is big now.
[00:04:00]
There’s a reason why it’s at the peak of inflated expectations, right? There are some bad news forthcoming. However, I believe it will be revolutionary beyond cloud in terms of its actual impact in the business. Now, will it be application by application? No. I think the answer here is mostly feature-based embedded, right? I’m going to build things. I’m going to buy from the marketplace and I’m going to consume from my big technologies, the sales forces, the Atlassians, the service desk. They’re going to come with embedded JNI features. Because they got the data. Yeah. They have the data. It’s already controlled. It’s already highly secure. It’s already private.
Data Ducks in a Row
Those are good things.
[John Furrier]
All right. So we were talking before we came on camera about data, the reality of where we are. Are things in place? And you said the ducks got to be lined up. I said, that’s not even on the pond yet. So what ducks need to be lined up for an organization to truly start reaping some of the rewards from pouring more cash to building out embedded AI, AI enabled things, AI for this, things helping AI.
[00:05:05]
Prerequisites for AI Adoption
What’s the ducks that need to be lined up?
[Paul Lewis]
It’s definitely the crux of the issue, right? And which is why it’s going to take a little bit longer. The reality is there’s a bunch of prerequisites still there. I don’t have all of my knowledge bases in a single entity, or at least not accessible by a gen AI model. I have databases at the edge in multiple data centers, in multiple clouds, in multiple SaaS products to which I can’t even get the data out of. That’s a problem. And we’re measuring these things in petabytes, if not exabytes of data. I can’t just migrate that data somewhere, right? I can’t just make it available for use, right? And therefore getting my sort of data ducks in a row, not just accessibility, but security, privacy, when I can use it, how I can use it. And really what are the architectural and economic limitations to that, right?
[John Furrier]
Baby Steps and Best Practices
Well, what are people doing now? What do you see as like the baby steps, or as they say, three feet in the cloud of dust?
[00:06:01]
What are people doing to like get moving? We heard from Jensen that GTC for NVIDIA, accelerated computing, everything’s being accelerated. What are some of the best practices? What are people doing?
[Paul Lewis]
Enterprise Search and Data Dictionary
They’re starting enterprise search, right? The unstructured knowledge bases at least already exist. They might not be well-tuned, but they exist. I can get to my drive, I can get to my wikis, I can get to SharePoint, I can get to these things. And therefore I can have a better enterprise search, both for my internal IT staff, or even for my customers. And that’s where I’m going to start because I have that data. Once I need to start grounding, right? Once I need to start augmenting it with my own data dictionary and my own customer information, that’s where it gets a little bit more complex, right? Because I now have to have a data dictionary and metadata of my company to say, when I say price, I mean this specifically.
[Rob Stretch]
Metadata and the Evolving Role of DBAs
Yeah. And I think that to me is one of the things we’ve been talking about a lot is the metadata and having that global metadata across there.
[00:07:02]
The role of the DBA is, and having started my life on that side and financial services, when you look at the role of the DBA today, it has become uber complicated. Now they’re called data engineers and now they’re trying to figure, how do you see that evolving in this market given where data is just spread out all over the place, but you need the right data in the right place to action these AIs?
[Paul Lewis]
The Evolution of DBA Skillsets
Well, we have 500 DBAs and we see it real time, this evolution of skillset. What DBAs used to be was alter table, right? Alter table, and then they managed the actual database service itself, right? They patched Oracle as an example. But that’s now not good enough. They need to have multiple roles. They need to be worried about getting data in. So they need to be data engineers. How do I migrate data in? How do I get data in? They need to be data platform people. It’s not just the OLAP, it’s the OLTP as a single device.
[00:08:03]
So the database engineer is not just Oracle database, but it’s also Snowflake. It’s also BigQuery. It’s also Databricks. They have to have a much more foundational knowledge of security. In fact, they might be a security analyst or a security officer because privacy and security of that data, that’s the nuggets of gold, right? That’s where those bad actors are trying to get to. So now they have this convergence of all of those other IT skills. Alter table simply doesn’t cut it anymore.
[Savannah Peterson]
Regulatory Differences between Canada and the US
I’m curious, so you’re based in Canada, correct?
[Paul Lewis]
I live in Toronto.
[Savannah Peterson]
Yes. So when we’re talking about regulatory environment, we’re talking about privacy, we’re talking about governance and collaboration here, you’re dealing with some of the biggest companies in the world, hands-on. Is there a difference, or I’m curious to even hear if you have any observations between Canada and the United States in terms of our approach?
[Paul Lewis]
There’s definitely large regulatory differences between the two. Break it down for us. And it’s hard to say whether either one of them are advanced in fairness, comparatively, but if you look at the last executive order from the White House, they are setting up an obligation of the federal government, each one of the agencies to say, thou shalt have and implement a set of privacy requirements, a set of security requirements.
[00:09:19]
Each one of them has to have a chief AI officer, a chief data officer, a chief security officer. They now are as focused on sort of the security regulation and legislative boundaries as any other private entity. That’s an assumption that’s going to move from just the government to private entities. There’s now an expectation that every organization from 50% up will have that same structure.
[Rob Stretch]
Working with Partners and Google
Yeah. And I think one of the interesting things, and I think we were chatting at the last Google Next, which feels like a year ago, but it was literally only like eight months ago. I think one of the big things that also comes around that is how you work with your partners and people like Google and how that comes together, because security, privacy, and all of these things, how do you work with your customers to help them understand that Google is a good place for them and to work with you on that?
[00:10:18]
[Paul Lewis]
Deeper Relationship with Google
It’s very helpful by having a deeper relationship with Google as an example. So Pythian is an MSP. We’re also a reseller. We also help on the engineering side of some of their technologies. We’ll test LRDB as an example. We will be involved in the product marketing stuff. My team is actually a pretty big consumer of GCP. So we get to use our own selves as an example of how to deploy complex architectures into that cloud environment. And we can say, here’s the framework we used. This is what makes sense for our 25,000 databases and 400 customers. It probably works for you too.
[John Furrier]
Digital Transformation Impact
Talk about the digital transformation impact. We’ve been talking about that, Rob, for decades.
[00:11:01]
It’s become cliche that there’s a pre-gen AI hype and or changeover, which we’ve been talking about. And I think it’s legit. I think the bubble will burst, but it’s still revolutionary. Totally agree. But all that talk about digital transformation pre-gen AI was about the data lake, snowflake, data bricks, maybe big queries thrown in there a few times, Spanner. So, okay, cool. In comes gen AI. What has that changed in terms of the role of the CISO, CIO, CXO, and the development teams? Because now you’ve got more stuff going on that’s generative, not so much pre-programmed or pre-staged. How does the digital transformation equation or journey change, get killed or rebooted, reset?
The Changing Role of CIOs
What’s your take on this?
[Paul Lewis]
Excellent question. I have the opportunity to talk to at least 100 CIOs and CTOs every single quarter. I’m on a few boards, but I do a lot of round tables. And I definitely have a deep appreciation for the change in the CIO’s role for digital transformation over the last five years.
[00:12:01]
The CIO is now at the table, the CEO, they’re making the growth business decisions. They’re responsible for the digital transformation program, which meant the rest of the IT was focusing operating IT. And now there’s this gap between the VPs and the CIO because they didn’t get to experience that. But now the CIO is saying… Didn’t get to experience what? The leadership table. They weren’t part of the growth program. They were just operating IT. So the next set of CIOs might be 10 years away instead of five years away, as an example. But this digital transformation program, wow, what is it? 76% of companies have a program, but only 8% are successful. Very small amount. That’s dramatic. Very small amount.
The Goal of Digital Transformation
Because the goal is saying, I need to change the way I’m selling my services to how the consumers wish to actually buy. That’s a customer journey question. That’s not enable data. That’s not even just monetized data. That’s saying, I have a certain set of customers that are going to some other competitor.
[00:13:00]
What do I have to change holistically to make that happen? And some of those holistic changes are simply the difference between more and better. Most companies are more features, more functions, more products, more things, more stuff, where customers are actually acting for chaotic environments. I don’t want a 30-year relationship. I want a one-month relationship. And if I don’t like your app, I’m going to delete the app and download a new one. That’s a very different customer journey, which means I have to rethink how I’m going to implement technology.
Monetizing Data and Creating New Interactions
One side, I have to monetize data. How do I get a better insight to my customers based on the data that I have? And then how do I create using Gen AI or new talk with my data or new enterprise search or new creative means to create a different journey for that new expectation, that new segment that’s going to some other cloud-native environment? That’s the big difference between the two. Digital transformation used to be just monetize data to get that journey. Now it’s monetize data, actually create new interaction.
[00:14:04]
[John Furrier]
Gen AI and Product Development
That requires Gen AI. That’s a product conversation. That’s not an operational thing so much as it is both operational and product development. Exactly. And by the way, generative is runtime product development. So if you’re generating a product on the fly, you got to enable an entire new data marketplace.
[Paul Lewis]
Prerequisites and Costs
And back to those prerequisites, that sounds expensive. That is an expensive adventure to run and manage your own models, tune and retrain your models. Or hire someone to architect it.
[John Furrier]
Hiring for Gen AI Projects
Who do you hire to do this? I mean, obviously, you guys are in this business, but this is what I find people are struggling with. Okay, I get it. It’s mind-blowing. I can see the leap of faith, that bridge to the future. But what do I do now?
[Paul Lewis]
The Biggest Error CIOs Make
Well, I can tell you the biggest error CIOs have made in this Gen AI period of time is they gave those projects to the chief data officer or to the BI team. The reality is Gen AI is a software engineering project.
[00:15:01]
I have to have pre-processing and post-processing, and I have to make sure it’s guided by the principles of my organization and that hallucinations don’t make errors. Prompt engineering is a great example. Prompt engineering in the consumer sense is asking a bunch of questions to hone my answer. Prompt engineering in the enterprise sense is asking a bunch of questions that gives me one answer, because there’s only one right answer.
[John Furrier]
Well, Bob and I are talking about this, Paul. It’s a great point. It’s a systems problem, not a, I got to analyze the data and do some schema work, unstructured data, the database, we can query it, it’s pre-programmed, prompt response, no reasoning, really. So, I mean, Rob, this is what platform engineering, all this we’ve been talking about.
[Savannah Peterson]
Yeah, yeah, yeah. It’s all those data ducks in a row, like we were talking about.
[Rob Stretch]
Treating Gen AI as a Product
And I think another piece of it is, and I think you just hit on it, is how do you treat this as a product, not just a project? And I think that’s the big difference is that companies that are doing AI as their product get it because that’s all they’re doing.
[00:16:03]
But when you’re dealing with organizations that that’s not their primary role, right? You’re going in there and trying to help them understand how do you monetize your data? What’s the data you want to do, use as part of that? I would assume it’s a much different conversation with those customers about how do you really leverage and what is the right cloud to be leveraging for that matter as well.
[Paul Lewis]
Educational Conversations and Use Case Mapping
Yeah, 100%. Most of our conversations, especially with our clients, that is, let’s go back to the drawing board. Let’s have a educational conversation. Here’s what Gen AI is and isn’t. Here’s what you can or cannot do. And then let’s look at all your potential use cases, and let’s put it in the two by two matrix. The difference though is in the old days, we would never have gone to a customer and said, here’s cloud. Give me all the use cases to deploy to cloud because it doesn’t make logical sense. Here’s a database. Give me all the use case implemented. But we do that with Gen AI. Here’s Gen AI. Tell me all the things we could do to create value for your data or create a new customer journey or all of those things, which they’re much more naturally in tune to giving you an answer to.
[00:17:08]
[John Furrier]
Baseball Metaphor and Early Value Points
Well, great insights. And this is a great interview as we explore and riff on this. We were using a baseball metaphor earlier, as we always do. Early innings, later innings, the value will come in. But in one case, Rob and I were talking about people are so starved for value to prove the value of Gen AI that you see RAG, Retrieval Augmentation Generation, become the hottest app. That’s just data wrangling. It’s a search problem. And so that is, I won’t say desperation, but that’s an easy way to go for proof points with my own data. So what are going to be the value points that you see people doing out of the gate besides RAG that’s going to be value-based, where they can get a win, a single, lay down a bunt, get on base, to use our baseball analogy?
Value Points Beyond RAG
Because that’s what people are trying to do. Hey, boss, lookit, we got some movement here. We moved the needle a little bit.
Internal Use Cases and Code Assist
We got on base.
[Paul Lewis]
It’s mostly going to be internal use cases. So they’re going to start with enterprise search with their knowledge bases.
[00:18:02]
They’re going to start with code assist, helping the development staff. They’re going to do auto commenting on some of their code. They’re going to say, evaluate this old legacy technology, legacy SQL, legacy code, legacy infrastructure, and say, tell me about this because this person no longer works here. Document the things that I have. Oh, by the way, I have a hundred different documents that I’ve produced by Sally. Sally no longer exists. What did she do for a living again? Because I’m trying to hire somebody else. That’s the real win use cases now. And it doesn’t require sort of the privacy and security and governance, all of those extra functions that I need to worry about for external use cases.
[John Furrier]
Building Momentum and Skillsets
So momentum, get the momentum. Hey, boss, can you imagine if we did this, what we could do for this project or build this product?
[Paul Lewis]
Get the skill set. Once you have the skill set, engineer new projects, then you can start to worry about the hard ones.
[Savannah Peterson]
Advice for CIOs and Avoiding Mistakes
And get a win. What’s some more advice?
[00:19:00]
I love your advice for CIOs right now. And I mean, you’re touching a lot of, like you said, a hundred a quarter. What are some of the big mistakes or avoidable things that maybe haven’t happened in the past, but as someone’s evaluating their strategy for this year and looking forward for the next three or so, what would you tell them? Or what are you telling them?
[Paul Lewis]
Avoiding the “Cloud First” Mistake
Most CIOs made the mistake, especially in the cloud world, of saying, I’m cloud first. I’m creating a technology principle that didn’t make a lot of long-term sense, right? Because they put all their eggs in a single basket.
Avoiding a Single-Basket Approach in Gen AI
Gen A, I mistake, will be putting all your eggs in a single basket, right?
[John Furrier]
And in terms of a foundation model or technology or all the above?
[Paul Lewis]
Technology suite, right? You want to go where there’s options. So go to a model garden that has 30, 50 models. Because I want to be able to use different models over time. They will get better over time. They will get cheaper and more expensive over time, right? Presume you’re going to have a bunch of technology tooling that you might want to displace and replace all the time.
Presuming Data Division and Analytics
And also presume that your data will be divided and conquered across a variety of different opportunities to exist.
[00:20:08]
So know that some of those analytics need to stay there versus central. If I have analytics in a bank branch, it probably needs to stay in the bank branch. So that’s not really a opportunity.
[John Furrier]
Yeah, yeah. It’s architecture. It’s a system. You’re building a new operating system, operating model for the corporation. So it’s a whole reset.
[Paul Lewis]
A New Operating Model for the Corporation
And you kind of have to think of infrastructure, applications, data, analytics. It’s almost a fourth pillar now, right? Because it’s distinctly different than those other three.
[John Furrier]
And I think that’s why people are so hyped up. That’s why I think there’s going to be a big bubble pop like the dot-com bubble, because it’s obvious we talk about it. I can see it happening, but then it’s like, shoot, how do we do it?
The Infrastructure Bubble
Yeah.
[Paul Lewis]
Infrastructure is the bubble, right? There are very rich infrastructure companies in Gen AI. It’s going to hit a peak at some point, right? We got to raise a round now, guys.
[John Furrier]
Let’s go.
[Paul Lewis]
Our new CUBE AI funding.
[00:21:01]
Let’s go.
[Savannah Peterson]
Pillars of Gen AI Adoption
Speaking of pillars, what are the three pillars of Gen AI adoption?
[Paul Lewis]
So we want to make sure they’re accessible. We want to make sure that they’re enabled and we want to make sure that they’re secure, governed. Those are the sort of the big threes of making sure that that data is all the prerequisites are checkmarked.
[Savannah Peterson]
Accessibility, Enablement, and Security
Okay. So this has been fascinating and you obviously have quite the lay of the land and the pond and the ducks on that pond to continue our favorite metaphors of the week.
Looking Ahead to Next Year
What do you hope that you can say sitting next to us next year that you cannot say yet this year?
[Paul Lewis]
Real-World Examples and ROI
I’m going to, I believe that we will have real world major examples within our organization with our customers that have an obvious ROI, right? But not big projects, not big multi-million dollar things, small hundred thousand dollar quick wins that have obvious in-quarter ROI. That’s the goal. Let’s have a thousand of those.
[00:22:01]
[Savannah Peterson]
In-Quarter MVP ROI for Gen AI
In-quarter MVP ROI for Gen AI and some of these projects.
[Paul Lewis]
Thank You and Closing Remarks
Love it.
[Savannah Peterson]
Can’t wait to be chatting about it with you in 2025. Paul, thank you so much for joining us for this fascinating conversation. John and Rob, always a pleasure. Your insights and your questions were fantastic. I thank all of you for tuning in to day one of coverage here at Google Cloud Next in Las Vegas, Nevada. My name is Savannah Peterson. You’re watching theCUBE, the leading source for enterprise tech news.