Scaling Enterprise AI: AWS, Deloitte, and Accenture on Databricks Partnerships

Summary

MIT research cited in this video found that only 5% of custom enterprise AI tools actually reach production today, and AWS, Deloitte, and Accenture partner with Databricks to help enterprises close that gap through centralized governance and automated pipelines.
AWS customers such as Workday and Mastercard use Databricks and AWS together to build governed, production-ready AI applications, with Workday deploying AI-driven tools that help marketing teams answer questions in real time.
Accenture identifies data readiness — having clean, governed, and accessible data — as the foundational requirement that separates organizations successfully scaling AI from those stuck in the experimentation phase.

Scaling Enterprise AI: AWS, Deloitte, and Accenture on Databricks Partnerships

Enterprise adoption of AI requires more than frontier models. This panel explores how AWS, Deloitte, and Accenture partner with Databricks to help enterprises move AI from experimentation into production at scale. Only 5% of custom enterprise AI tools reach production today. Hear how companies like Workday and Mastercard use Databricks and AWS together to build governed, production-ready AI applications.
The discussion covers unified data governance through Unity Catalog, federated governance at global scale, AI-empowered security operations, and the critical role of data readiness as the foundation for successful AI implementations. Learn what separates scaling enterprises from those stuck in the experimentation phase, and the strategic imperatives for leaders building enterprise AI capabilities.
🤝

Chapters

FAQs

Why do so few enterprise AI projects successfully reach production?

MIT research cited in this video found that only 5% of custom enterprise AI tools actually reach production in an efficient, value-generating way. The gap exists because most organizations lack centralized governance, automated data pipelines, and the data readiness foundation needed to ship trusted AI agents rather than just running experiments.

How do Workday and Mastercard use Databricks and AWS together?

Workday uses the Databricks and AWS combination across a universal data layer to create governance in AI experimentation while building agentic applications, including AI-driven tools that help marketing teams answer questions in real time. Both Workday and Mastercard are cited in this video as joint customers using this partnership to build enterprise AI at scale.

What is Deloitte's focus area in its Databricks partnership?

Deloitte focuses on AI-empowered security operations, helping enterprises build unified governance through LakeWatch and Unity Catalog integration. The partnership supports Security Operations Center use cases and extends into AI-driven security workflows where automated agents assist analysts.

What does Accenture identify as the most critical foundation for enterprise AI success?

Accenture's perspective is that data readiness — having clean, governed, and accessible data — is the critical foundation that separates organizations scaling AI from those stuck in experimentation. Without this foundation in place, even sophisticated models and frameworks will fail to deliver reliable, production-grade business value.

Full transcript

[00:20] Hey everyone, welcome back. Ari Kaplan here with Summit Live. Super happy to be here with Mona with AWS. And the funny thing is you can't see behind the cameras, but the AWS booth all day and all tomorrow I love it. I love it. customer success with Databricks on AWS. So, Mona, welcome and can you walk us
[00:37] through the core values that customers gain from choosing Databricks when they work with a customer? Absolutely. Well, really the core values come down to three things, which is consolidation, speed, and economics. And that's And really the best way to illustrate that is to see like what
[00:54] we're actually seeing with customers. And part of what we're seeing is that Look, and I think everyone's experiencing this now, which is while everyone is using AI in some way, shape, or form, very little is actually in production, and that's actually in production in a efficient way, right? In
[01:11] a way that's really garnering value for companies and enterprises. And so, what, you know, we've seen and there's been, you know, stats on this across all the different, you know, you know, analysts and, you know, even MIT most recently reported that only 5%
[01:27] of custom enterprise AI tools actually reach production today. So, that's really where the value of Databricks and AWS comes in, and it really helps customers close that gap. So, and what that What does that mean? That means centralizing governance, automating data
[01:43] pipelines, and really giving teams the tools they need to ship trusted AI agents, and not just experimenting with them. And so, the proof is really what we're seeing with customers and helping them build enterprise AI at scale. So, Workday is one of our, you know, really
[01:59] powerful joint customers. They're using Databricks and AWS together across their universal data layer. And what they're trying to do is land AI experimentation, um you know, creating sandbox environments and creating these agentic applications. But what it really also
[02:16] highlights with Workday is that they've created a governance in AI experimentation that goes hand in hand without compromising their user experience. So, what they've done is they've taken their marketing organization and their team is accountable for running, you know,
[02:32] conferences just like what we have today at the Data AI um Summit. And they're they've deployed AI-driven applications that really help their marketing teams answer questions at an accelerated rapid rate and at real time. So, getting that
[02:48] giving the ability for like, you know, uh line of business and um knowledge workers to have information at their fingertips is really critical. Um another great example that we have together as like a joint customer is Mastercard. And they face sort of like
[03:03] that classic multi-catalog challenge where they had data scattered across and different engines. They had metadata that was fragmented, governance was really inconsistent. And so, what was their solution? They federated Databricks Unity Catalog to AWS Glue,
[03:20] which enabled um you know, we're going to geek out here for a little, enabling Apache Spark engines to really consume data in Iceberg tables that registered their Unity Catalogs and sync uh sync seamlessly across AWS services. And what
[03:35] was the outcome? So, the outcome was a single source of truth for governance that lets their domain teams choose the optimal compute engine for each job. And, you know, they really built a sustainable future-proof um architecture that supports machine learning,
[03:51] streaming, governed data um sharing across their entire entire enterprise. So, just a massive, you know, I think, you know, those are two really good proof points. And on top of that, MasterCard was actually in the keynote today and talked about how they use Unity Catalog
[04:08] and how they're really building agents in real time to help them, you know, build out their business. So, and helping them move from experimentation to production. So, again, just a really excellent proof point. Yeah, and first we love geeking out. A lot of customers on the call. And yeah,
[04:24] Workday, MasterCard, epic companies, epic partnership. So, speaking of the partnership, like how is the better together narrative, like it's it's really resonating. So, how do you see our partnership evolving to accelerate all of this customer innovation?
[04:39] Yeah, you know, I So, this is a over a decade-long partnership and the trajectory is very clear. More of our, you know, as we look at our partnership, we have more joint investments. We have even more tighter integrations together. You know, for example, Databricks and is
[04:55] integrated with their Unity Catalogs are all integrated with Bedrock and Bedrock Agent Core. So, that delivers sort of faster paths to AI production. And, you know, the other thing is that Databricks eclipsed $2 billion on AWS Marketplace in 2025. And their
[05:13] trajectory and the trajectory with Databricks is just even like way higher than the $2 billion as we sort of think about 2026 and beyond. And, you know, look, at the end of the day, agents are only as good as the data and runtime behind them, which is really
[05:29] why AWS and Databricks have built this deep, you know, deeply integrated agentic stack. So, Amazon Bedrock provides access to manage frontier models like Anthropic Claude, Amazon Nova, Meta Llama, OpenAI's ChatGPT is
[05:45] now available as well. We made that announcement with that partnership just a few weeks ago. And you know, in addition to that, we built these guardrails in place. So, when you pair it with Databricks, these models are able to reason directly over the governed high-quality data rather than relying on like any
[06:01] sort of stale copies. And you know, Amazon Bedrock Agent Core supplies that enterprise runtime, um you know, the agent so that they run and they are optimized to run at scale handling more memory, identity, code Yes, exactly.
[06:18] There's like world No, I like to think that they're, you know, applauding us in the partnership. We're not AI. We're real people in a real conference. Exactly, exactly. So, you know, I would just say that, you know, all of the agents are handling memory, identity, code interpretation,
[06:34] um you know, observability across the entire stack. So, through a governed MC connection in Databricks um to Databricks via Databricks apps and Agent Core um agent can really securely can securely query Unity Catalogs and can
[06:49] govern that data, interact with AI BI Genie um for really natural um language analytics and and low and read low latencies uh you know, state from lake base. And all doing this while honoring the existing access permissions. And so,
[07:06] having that security layer is really critical. And that's really the the uh you know, one of the key benefits for Bedrock and Databricks and being able to access those models um via Bedrock. And so, for builders creating these experiences, whether they're in Kiro or AWS's Agentic IDE for um spec driven uh
[07:23] driven um deployment, it really helps turn intents into working code for applications that run across Databricks and AWS. I love that better together story. It is a great story, Okay, so I mentioned we have a lot of practitioners out there and data people
[07:40] out there as well. Welcome, welcome from around the world. What are the key takeaways and what should they do next? Yeah. Well, look, I think here are the you know, takeaways. To really succeed with a genetic AI enterprises need a single secure location to gain insights
[07:56] from their most valuable data and Data Bricks on AWS is really a direct path to get there. And you know, we talked about customers like MasterCard you know, achieving federated govern governance at global scale, worked at powering AI experimentation and a
[08:12] genetic application without compromising governance. And you know, they acted. They didn't they didn't wait. They acted and they worked with AWS and Data Bricks. So, I'll just say this. The first thing that you know, you should do is really explore your sort of use case and what that is. And what we're finding
[08:28] is most customers are falling into sort of these areas which is you know, accelerating their AI builds, you know, unifying their data governance, you know, trying to figure out their data state modernization and you know, trying to improve their scale performance and TCO.
[08:44] So, figure out what that use case is. And then second, start a conversation with your AWS account team or you know, and Data Bricks representative. Having those conversations together is really important and coming to the table together at accounts is super important. And then third, I would say you know, we
[09:01] talked about AWS marketplace and you know, Data Bricks eclipsing the $2 billion. So, take advantage of AWS marketplace. And look, onboarding takes seconds at this point. So, deploy and you know, experience what that looks like with Agent Bricks and and AWS. So,
[09:18] look, just the combination of Data Bricks and AWS is just you know, technical you know, it's not just a technical choice. It's a strategic decision to accelerate AI. Awesome. Well, thank you so much. The partnership very much appreciated. It's
[09:35] epic, it's monumental and appreciate you taking the time to come on. And now we are going to I had the chance actually to sit down with Steve with Deloitte. So, let's go and roll that video. All right, Steve. Welcome to Summit Live.
[09:51] Thank you for having me, Ari. I appreciate it. Yeah, I'm excited of our conversation, but first let's have you introduce yourself. Sure. Sure. So, Steve Mahar, I'm a managing director with Deloitte Cyber Practice. Been with the firm for 13 years, and my
[10:06] focus primarily is on helping our clients with cybersecurity operations. Great. Well, why don't we jump in and tell us a little bit about how your team's working with our Lake Watch team? Sure. Yeah. Yeah, so our our work with your Lake Watch team has been really
[10:22] highly collaborative. Um we're focused on helping our clients with their security operations journeys. Um again, it's it's always a journey and never a destination with security operations, but um we're really helping them to incorporate
[10:37] uh security data lake, uh analytics, next-generation security information and event management, or what we call SIM, in in SecOps, and hyperautomation, um things like agent chaining
[10:53] uh of traditional SecOps functions like uh common detections, level one analysts, uh response, threat hunting, and also detection engineering. Um we see the current road map uh you know, current and road map
[11:09] capabilities of Lake Watch really as an enabler for the AI-empowered SOC, not necessarily of the future, but of the not-too-distant future. And uh our clients are on this journey to the AI-empowered SOC, and our work with the
[11:26] Lake Watch team is enabling that journey from SIM augmentation all the way to genetic security operations. Yeah, awesome. And then those are all super important. Like every company, if you have a CISO, they have to be looking at this. Um so for me, you know, in
[11:43] addition to Lake Watch, we have Unity Catalog that helps, you know, your clients with security and compliance reporting. So talk a little bit about that. Yeah, so um you know, again, we have number of teams that help clients in different ways all the way from SecOps to data and privacy,
[12:00] identity and access management. So our teams really are leveraging a combination of Lake Watch and Unity Catalog really to onboard and merge uh raw IT and security uh telemetry. Um this is a process of onboarding that
[12:16] type of data that traditionally can take weeks, even longer, with traditional SIM technologies. Now we can accomplish it in hours. Uh as a matter of fact, we'll be showing a demonstration later this week um where we cut that process, you
[12:31] know, by weeks. Wow. Um and we're pretty excited about it. Um so we're able to help clients to merge that data and do it in a single governed data environment. Um we also with fine-grain access controls, so we're not
[12:47] duplicating data um or data environments. And this makes it much easier to streamline the reporting uh and act, you know, as well as the access for our data and privacy teams and our identity and access management teams. So really what we help clients achieve is digital trust and
[13:05] privacy um in a much more streamlined manner and also at a much lower total cost of ownership. Yeah, for sure. Um yeah, so what's exciting you about the future of our partnership? Uh um gosh, what's not exciting me, right? Um
[13:21] so really, so when I think about Databricks, it's already one of of fastest-growing alliances in the firm. Um as we expand how we're going to market together with our cyber practice, we expect that growth to really accelerate, you know, and if you think
[13:37] about all the news in frontier AI and AI-enabled cyber adversaries, what they're doing is really driving the market and our clients um to process, analyze, and operationalize
[13:52] massive volumes of disparate data, right? Data that most security teams have not traditionally logged into their SIMs. And really to defend against uh modern cyber attacks did but to do so against, you know, at machine speed. This is what we're all
[14:09] moving toward. And really, you know, Databricks, you know, the data intelligence engine combined with Lake Watch, and you know, this is where the magic happens. Combine that with Deloitte's leading practices across AI, data, and cyber, uh we're
[14:27] really looking to help clients transform the way that their businesses work and the way that they defend their businesses businesses against cyber adversaries. Awesome. Well, that's a great summary and everyone who's watching, if you're in the cyber
[14:42] security practice, absolutely need needful. And if you're tangential to it, you know, please bring this to your attention of your security department since it's more important now than ever. And Deloitte is right here as our partner to help make it happen at your company. So, thanks for coming on.
[14:59] Appreciate it, Ari. Thank you. All right. All right, and thanks again, Steve, for meeting with me uh prior. Also, yesterday I had the opportunity to meet with Arun from Accenture. So, let's roll
[15:14] that tape. All right, we are here with Arun from Accenture. Welcome to Summit Live. for having me. Oh, well, we're talking about like all the momentum in AI, uh but there's also data readiness that could be a real bottleneck. So, want to understand what
[15:30] separates companies that are actually scaling AI from those that are being stuck. You know, Ari, we're living through interesting times. We talk about AI, but one of the things we tend to forget, if not downplay it,
[15:47] is the value of data. Uh we have all the technology for AI for the most part. Yes, it's going to evolve. Yes, it's going to grow. But what gets lost is the data importance of data. Organizations
[16:03] who want to deploy AI at scale typically view data strategically. Mhm. They look at data strategically. They talk about bringing AI to data. They talk about driving business outcomes to data. They talk about building a unified data foundation.
[16:18] Data is, you know, I'm going to say one of the most key black goals if you want to get to the AI. And that's what excites me the most being in this conference, looking forward to hearing about how we bring AI to the data,
[16:34] but I think data is absolutely strategic to any successful AI. Yeah, 100%. So, where do you see, you know, your clients, like where are they struggling most when it comes to getting that data ready for AI or AI ready? Most clients struggle when they get
[16:50] excited about the AI and the business outcomes they can get, but they're not prepared with the data. For example, they don't have the right con- contextual data, structured data, unstructured data, data foundation. Some of those, they don't have the right assets
[17:06] to actually go deploy AI to get meaningful, unique insights for their enterprise. That's where they get stuck with, and then it becomes a bottleneck where number of our clients they start with business outcomes but they soon realize they have to fix
[17:21] the AI and the data part of it before they can actually go back and get back to the AI basics. Yeah, for sure. So, when leaders they think about next steps that they do, like what should they be doing differently than they are now? I think the one
[17:37] you got to have a data strategy. And I'm what I mean by that is a data foundation. What exactly data, what data you need. What AI you want to do. That's pretty important. B, how do you get there? What's the journey to get that data? C is what kind of
[17:54] you know, contextual semantic layer of data you're going to build because AI goes multiple places, but you need that semantic layer to actually make that happen. What kind of semantic layer you need. And four, what kind of powerful tools you have like Genie which Databricks all that I can
[18:10] I'm so excited to hearing about some of those. How do you use AI built into the data? All right. Because those will give you instant uh AI benefits and will help you also optimize your tokenomics of the cost.
[18:26] Because that is a very powerful tool and I'm so looking forward for that. Yeah, well, so you know, great summarization, you know, everyone's racing ahead with AI. People still need to be data and AI ready together and it's big and big honor for having you on
[18:41] the show since you have like such a global scope and Accenture is like one of the leading brands to help companies get there. So, thank you for coming on. Thank you so much. Thank you.

Learn more about the Databricks Data and AI platform.

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