Streaming Real-Time Data into Databricks for Enterprise AI with Confluent and IBM

Summary

Data silos: Stale data across clouds and on-premises prevent enterprise AI value.
Real-time stack: Confluent streams hybrid data, Databricks analyzes it, IBM watsonx governs.
Kafka and Flink: Apache streaming solutions solve stale-data for enterprise AI assistants.

Streaming Real-Time Data into Databricks for Enterprise AI with Confluent and IBM

Talk by Denny O'Brien, Managing Executive, TD SYNNEX, IBM
Enterprise AI stalls on stale, siloed data, which makes the streaming layer as important as the model. This video shows how Confluent, Databricks and IBM fit together: Confluent captures and streams events from across clouds and on-premises systems using Apache Kafka and Flink, Databricks turns that real-time data into analytics and AI, and IBM watsonx operationalizes and governs the outputs across a hybrid environment.
Denny O'Brien of IBM lays out why AI value stalls: data silos across hybrid environments, multiple model providers, point solutions that add complexity, and stale data, illustrated by an AI assistant recommending restaurants that had closed. He covers IBM's hybrid cloud and AI strategy, including AI productivity, data management, and security and governance, and explains Confluent's foundation of Apache Kafka, Apache Flink and open table formats. Confluent streams data into Databricks, wherever your Databricks estate lives, with watsonx AI on top. He then explains how TD SYNNEX helps partners move from one-off projects to repeatable, services-led revenue.
Sponsored by TD Synnex.

Chapters

FAQs

Why does enterprise AI stall on silos and stale data?

This video describes how AI applications fail when data is siloed across multiple clouds and on-premises systems or too stale to be useful. It illustrates the problem with an AI assistant recommending restaurants that had already closed.

What role do Apache Kafka and Flink play in enterprise AI?

Confluent builds on Apache Kafka and Flink to capture and stream events from across clouds and on-premises systems. This video explains that this streaming foundation lets Databricks turn real-time data into AI and analytics.

How does Confluent stream data into Databricks?

Confluent captures events and streams them into Databricks across any cloud environment where your Databricks estate lives. This video shows that IBM watsonx then operationalizes and governs the outputs across hybrid environments.

Why are point solutions adding complexity to AI stacks?

Multiple disconnected tools for streaming, analytics and AI governance increase operational overhead. This video argues that integrating Confluent, Databricks and IBM creates a unified stack that reduces complexity.

Full transcript

`00:00` Introduction

You guys, you drew the short straw to get to the last session of the entire conference, right? Or maybe I drew the short straw to present at the last session of the entire conference. Hey, but thanks everyone for joining. My name is Denny O'Brien and I'm with IBM. I will not tell you how long I've been with IBM, but it's more than one year, but less than 29 years. So pick your number between those two. I've done a number of different roles.
Almost every one of them has been in software. And today, I'm the managing executive for all of our business through TD Cinex, which is our largest distributor of IBM technology worldwide. So today, I just want to kind of close out the conference by talking about, you know, how IBM and TD Cynics and IBM Confluent and Databricks all partner together to deliver AI outcomes to our clients. Let's start first with the challenge.

`00:53` Why enterprise AI stalls: silos across clouds and on-premises

You know who was not ready for the AI explosion? We weren't. Nobody was ready for the AI explosion when it came up upon us. When you think about the number of apps that we're going to be looking at by 2028, we're looking at over 1 billion new apps. Not apps, new apps on top of the apps that we already have out there running on multiple clouds, on-premises, pick your infrastructure of choice there. The problem with that is, right, we did not prepare and build the infrastructure to support this, right?
We don't have the cooling systems and the power to support the explosion of apps that are coming our way. So plain and simple, we weren't ready. And then you look at some of the other challenges such as fragmented data. The data that we have on premises, the data that we have running in our public cloud. By the way, raise your hands. How many of you are running on one public cloud only? What's wrong with you guys? Most enterprises are running, most organizations are running on multiple clouds today, which unintentionally creates the problem of having fragmented data running in multiple silos.

`02:30` Why AI applications need real-time data, not stale data

And then finally, the last challenge I want to talk about here is that in the world of AI and building AI applications, because many times we're not running on real-time data, these applications are lagging in slope. So in my personal life, when I travel for work, I use AI heavily to plan, I'll just say, my after work hours. I'm a big fan of craft IPAs. This is not being recorded, so I feel like I can say that in mixed company here.
So I like to go online. I like to use Copilot and say, all right, I'm planning to eat dinner here, but I would like to have a couple of craft IPAs before I have dinner. Can you recommend a few spots, Copilot? And it does. The problem I often run into, as a matter of fact, I ran into it today, three of the spots it recommended are permanently closed, right, because of old, stale, non-real-time data. So many challenges that we're facing, right, in this AI revolution, for lack of a better term.
All right. And many vendors, and by the way, every logo you see on this page is an IBM partner. Ten years ago, that would not have been the case. It would have been strictly IBM competitors. But all of these logos that you see are IBM partners. And each one of them has tried to solve part of the challenge, but has unintentionally added to the problem. Let's talk about our hyperscalers. What is their goal? They want you to run all of your workloads on one cloud, right?

`03:50` Why point solutions add complexity to AI stacks

Their cloud. The reality is most of us in the room, except for the two or three individuals who raised their hand, and I'm not going to pick on you guys anymore, right? Run on multiple clouds and also run on-prem or in a private cloud there. And that generates silos. And what we haven't solved in many cases is how we bridge all of that data and info together. Then we have many companies who are developing platforms like Databricks and ServiceNow and Adobe and SAP.
They build a single platform. They gain domain expertise in one area, and then they expand from there. What's lacking in many cases is the integrations across those domains. Then we have the multiple model providers. Then where does governance come into play and security come into play around that? And then we have the open source providers, which helps speed development and create agility. But the challenge around that is, where's the built-in security and support that helps us scale there?
So unintentionally, we have many point solutions that are targeting to help resolve the solution, but adding to the challenges there. So what do we need? What is the right approach and strategy to make AI more useful, real-time, and accurate? Number one, and this is the IBM point of view around this, is we have to embrace that we're hybrid first. By the way, IBM has its own public cloud, but we partner with the other public clouds.
We build integrations to the other public clouds. We provide value to the other public clouds because the reality is every one of our customers uses somebody else's cloud. So if we're not integrating with, if we're not providing value and observability to those clouds, we're going to lose those clients there. Another key to our strategy is open, open source. We're a huge champion to open source. Almost every acquisition that we've made, and I've got a slide that will show you that in a couple of minutes, has been to a company that has built its foundation on open source.
And then finally, to do AI correctly, it's got to be responsible. You've got to have the proper governance and the proper security in place integrated within your platform, not an add-on to your platform. And so the way that IBM is approaching this, given the challenges that we talked about and the strategy and the approach that we are targeting towards, IBM has modernized over the last five years into a hybrid cloud and AI software company.

`06:39` IBM's hybrid cloud and AI software strategy

I just want to show our hands real quick. When you think of IBM, how many of you think hybrid cloud and AI? How many of you guys think mainframes? There we go. I thought I'd get some hands on that one, right? I mean, that is the legacy of IBM. It was mainframes. It was laptops. It was power servers, big old boxes that it takes 15 men in a white van to unload and get into your data center. But that's not what we do anymore, right?
We're hyper-focused on helping our clients accelerate their hybrid cloud journeys and their AI outcomes, right? And so what we've done is we've modernized as a software company, built on five key motions or use cases, right? One is around AI productivity, all right? And helping our clients drive outcomes by leveraging AI to be more productive, right? And by the way, we're a client zero user, right, of that notion. And I'll just give you an example.

`07:39` Five motions: AI productivity, data management, security and governance

So, agentic AI and orchestration, we have within IBM, right, we have embedded it everywhere. So, how many of you like to use, raise your hand, your internal HR tools? And you're going, really? You like doing that? I will not have a craft IPA with you, sir, after the show. But in all seriousness, right, so we use one of the leading HR platform providers. which, you know, the four times a month I use it, I always forget even to how to log in.
So what we build internally at IBM is using our WatsonX Orchestrate portfolio, an HR chatbot that basically I go in and say, can you transfer the guy who likes using our HR tools over to this manager over there? And it automatically does it for me. I don't have to go into SAP or Workday to do it. But something that normally would take me, I don't know, an hour to figure out how to do, I'm able to complete that task in two minutes, right, by simply leveraging agentic AI, right?
Then there's data management, right? If you don't have good data management, you're not going to have very good AI. And what I like about the third piece is in the center, right, we have security and governance. I actually came from, you know, my previous role within IBM is I ran our security software product portfolio, or at least part of it. And the fact that we're no longer thinking about security and governance as an add-on, we're thinking about as a centerpiece of every piece of software that we develop and deliver.
The next part is all around integration and resilience, right? We talked about the challenge of having applications running on-prem, data running off-prem. Being able to integrate across hybrid environments is critical. And then finally having that infrastructure management across that, being able to automate your infrastructure. So that's the IBM layer of it. But the fact is, you know, back when I started at IBM, we had many customers who were IBM-only clients. there are about 0.
1 of those guys left, right? And that's why we have an open, all of our platforms are open so that we can integrate with the Databricks of the world, the SAPs of the world, the sales forces of the world, et cetera, and help our clients, I would just say, build on the investments that they've already made so that they're not wasted there. And just real quickly, in the essence of time, we continue to invest here. Not only are we building organically, we are investing in acquisitions that complement and extend our portfolio to drive those hybrid cloud and AI outcomes there for us.
And so you can see web methods at BIO Hachicor. And then finally, our most recent acquisition, Confluent, which I'm going to spend a few minutes just talking about this afternoon. How many of you in the room are familiar with Confluent? All right, we got a few. I got to be real careful here. I'm sure all of you guys know more about Confluent than me. So, oh, you're not a confident guy. You're welcome to take the mic, sir.

`11:07` What is Confluent? Apache Kafka, Flink and data streaming

All right. All right, well, let's keep going here. So Confluent, why did we acquire or announce our intention to acquire Confluent, right? Let's go back to the problem we had with me winding up at a place that I was hoping I was going to get a nice West Coast IPA, but find out that it was permanently closed because of lagging and stale data. Confluent solves that problem for us, right? Confluent is our enterprise data streaming platform.
It was actually started by the guys who created Apache Kafka, right? Apache Kafka is a data streaming service and it is the cornerstone of, or the centerpiece of IBM Confluent. But Confluent is more than just Kafka, right? Getting the data from point A to point B is just part of the challenge, right? We also need to process those data streams and that's where Apache Flink comes in. And then also we need the modern, right, standard data table formats so that we can perform analytics on top of that.
So that is kind of the high level of what Confluent is. The other part of it and key to our strategy of being hybrid and open is you can run Confluent on-premises in your own data center. You can run it on one of the public clouds such as AWS or Microsoft. or you can bring your own cloud and manage it yourself there. So it gives you options in terms of where to run it. The other thing about Confluent I want to mention as we move to this next slide is out of the box, over 120 connectors that enable you to connect Confluent to, for example, Databricks or other data sources that are out there.
So another way to describe the value that Confluent brings to our customers is how many of you guys remember, so for me it was in the 10th grade, learning about the industrial revolution in your social studies class, right? Betty P. Myers was my teacher in 10th grade. She taught about the Industrial Revolution, and she essentially assured me that I would not be a history major by taking that class. But the great thing about the Industrial Revolution is, right, it was the explosion of all these factories, right, that were creating, you know, furniture and clothing, which is all great.
You needed the factories to build the goods, but then you also had to get the goods to the people. That's where the railway comes in. Without the railroads, the Industrial Revolution would not have been the success that it is, and it would not have transformed the world the way that it did. Confluent is very similar. It is the rails for the AI revolution that we're in today because it gets the data from the producers that you see to the left to the consumers and the use cases that you see to the right.
And the value that Confluent brings here is that not only are we taking the data and we're streaming it, we're processing it and delivering it in formats so that it can be reused and provide value across multiple use cases that are out there. So at a very high level, that is Confluent. All right. And then finally, you know, I mentioned, right, Databricks is a key partner of IBM. and I'm going to just go ahead and tell you a little secret here.

`15:05` Streaming Confluent data into Databricks with IBM watsonx on top

This was a slide that we use internally at IBM, but that middle column did not say Databricks. It said WatsonX. Databricks does the same thing, WatsonX. data does for IBM customers. Because of the connectors we have, insert Databricks, and we're able to stream the data into Databricks, wherever your Databricks estate lies, and then be able to sit our AI platform on top of it to deliver outcomes for our customers. So that is a high-level Confluent, IBM, how we all work together in terms of bringing outcomes to our customers there.
But now I want to just talk about the fact that when I was in security, we often said security is a team sport. It's not a one person or a one vendor show. It's the exact same thing with AI. And TDCinex as our largest distributor, in fact, take that word away. Distributor is a really bad word in this case. Think of TD Cinex as a solution aggregator that works with other partners, right? Think of the CDWs and logic haluses of the world.

`16:19` How TD SYNNEX enables partners to scale AI outcomes

So basically build out practices so that our partners can drive those outcomes to our customers. So if you are a business partner of technology vendors like IBM or pick one of your vendors around here and potentially interested in partnering to deliver outcomes with IBM, TD Cinex and IBM have together built a methodology that accelerates the onboarding of our capabilities and drives go-to-market with our partners. Starting everywhere from strategically aligning at the top levels of both of our organizations to getting our partners ready, we enable soup to nuts both technically and in terms of sales enablement.

`17:16` Partner enablement and go-to-market execution

And over the past year alone, or especially last year and a half alone, we've built out 100 data and AI practices across the United States built on our WatsonX technology there. So if you're not an IBM partner and you're an end user, but you're looking for a partner maybe to help you drive outcomes, TDCinex can also help you find the right solution provider to deliver outcomes for you. The point I just want you to see here is that we've partnered to develop a methodology so that our business partners can go and drive outcomes for our customers around AI.
And not only have we done that, we've built in a RevTech stack that is based off of WatsonX. orchestrate, which helps our partners identify target customers and the target products to deliver value with on that. And with all of that said, right, I just want to leave you this. If you want more information around what IBM or TD Cinex can do for you as either a business partner or as an end user, reach out to Patricia Rotundo.
She is with TD Cinex. And then Alan Glosser, by the way, who is in the back of the room, is on my team. Raise your hand, Alan. Thank you very much, sir. Your paycheck depends on it. You got to cooperate, buddy. But feel free to reach out to either of them for more information on IBM, Confluent, TD Cinex, or any of the products in our portfolio. And with that said, thank you guys for sticking around and coming and learning about how I use AI to find a good beer spot.
And with that said, enjoy the rest of your day and the rest of the last 10 minutes of the show. Thanks, guys.

Learn more about the Databricks Data and AI platform.

The information provided herein is for general informational purposes only and may not reflect the most current product capabilities or configurations.