Data You Can Trust Across Your Business | Informatica Keynote, Dreamforce 2026
Informatica launched AI Ready Data Intelligence at Dreamforce 2026, a product that scores data health and helps AI agents retrieve trusted information. A Claude plugin now available on the Claude Marketplace reduces token costs by up to 30× when data is properly prepared, with customers like UKG and Baker Hughes already seeing savings.
Chapters
Woo!
Wow, wow, wow, wow.
This is 2:00 PM on Thursday on the last day of Dreamforce, and this is an absolutely incredible showing. So thank you for being here. Thank you. Along with the thank yous for also supporting Dreamforce and supporting what I know all of you have been doing here, I know there's a lot of you who are interested in many different parts of what Informatica has to offer for Salesforce. But maybe before I begin, I just want to also thank you for being customers and partners for us. I truly mean it when I say that we value that a lot. It means a lot for us to work with you, to learn with you, to learn from you and move along the journey. And challenge us. Please challenge us as we are going through this journey together. It's going to mean a lot for us as we go through this. With that, I can obviously say this is Informatica's first Dreamforce. And it is Informatica's first Dreamforce. And it's probably one of the most incredible conversations we are having because Now, we have the entire data portfolio for Salesforce along with Informatica, and we can power so much more for you than we were able to do together before as just Informatica. It's also my first Dreamforce. I've been now here about 3+ months, and I was here before as a customer. Now, I'm here as part of Salesforce. And I'll tell you, I've had over 50+ conversations with customers such as yourselves, partners, and others worldwide. And I won't bore you with all 50 of those conversations, but I'll tell you a couple of anecdotes which were really, really interesting. Anecdote number 1, I was talking to one of the CIOs of one of the largest aerospace companies in the country. You would know them. And we started chatting about, so what are you doing with agents? You'll, of course, hear the word agents here, right? If you don't, that's a problem. So what are you doing with agents? And they were like, well, you know what? We have all these agents that people have created in different parts of the company. I'm not quite sure where they are, what they're doing, all of that. I'm like, OK, well, what about your data? Where is that at? And he's like, well, there's another team that does data management. It's been doing it for the last 10-plus years. It's creating data quality rules. It's getting the right catalog and everything. I'm like, sounds great. Are they talking to each other? And he said, not really. And that's part of the problem. And he said, what we really need is how we can connect the dots between all of that that's happening on the agent side and all of the stuff that's happening on the data side and bring that together. So that was one anecdote. And I'm giving you a little teaser because this is going to lead to something. Trust me. So that's on one end of the spectrum, that conversation. Another anecdote. I'm talking to the CIO of a manufacturing company based in the Midwest, and he's struggling with just one simple thing. How do I just master my data? Literally, how do I unify my profiles? They're going and acquiring new distribution— distributors. They're in the middle of doing that. So how do I master my data? Just simply, forget agents, forget AI. How do I just do that. So you can see where that journey is all the way from here to there. And the beauty of what you're going to see here in the next 40 to 50 minutes is we're going to cover that entire journey for the enterprise regardless of where you are in your journey. You could be at the front end of being able to deploy a bunch of agents and connect your data and actually have awesome outcomes, or you are still very much in trying to figure out How do you just manage, maintain, scale, and govern your data? Either way, we got you covered. And that's the best part about what we're gonna talk about here. Now let's talk about where the industry is at. The industry is changing rapidly. You all are seeing it. There's a stat out there which says that as humans, we have roughly created about 10 to 15 zettabytes of data. Zettabyte is like a billion terabytes, so it's just big numbers. But we have, as humans, created about 10 to 15 zettabytes of data. Now, agents, which some predictions say are going to be in billions, not hundreds of thousands, but billions over time, they're going to end up creating roughly 10 times that of 150 to 200 zettabytes of data. Take that in for a second. When that happens, What does data management look like? How do you even know where that data is? If an agent takes an action somewhere, how do you know what action was taken? How do you explain that? How do you take that log? How do you know that it had the right data and you're able to show it to an auditor or to a compliance officer or to a CISO? That scale is almost unthinkable. So the industry is changing radically, and the amount of data that's going to be created in the next 2 to 3 years is going to be 10 times more than that was created in the last 40 to 50 years. And it's going to be a combination of both humans and agents. To change with that, we have to change as well. And we are doing that. You knew Informatica as the most comprehensive, modern cloud data platform. That's what you've come to rely on us over the last several years. Now, I am thrilled to showcase the new Informatica. It is going from being the modern cloud data platform to being an AI-ready data platform. That's going to be the way that Informatica is now going to be working on. It's going to be engaging with And it's going to be supporting the entire Salesforce ecosystem through this new introduction. Why is this important? This is important because we know where the world is going. Does that mean that we leave the customers like the CIO of the manufacturing company that I talked about behind? Absolutely not. In fact, if anything, we're going to continue to serve the needs in the most robust way. And you'll hear me say this again. in a uniquely neutral way. One of the questions I hear is, hey, Informatica is now part of Salesforce, so does that mean that it's only going to work in the Salesforce ecosystem? Answer is absolutely not. In fact, we are going to continue, continue to maintain that uniquely neutral position that we have across the entire enterprise. And more so, we're going to serve all the use cases all the way from the AI-ready data platform, the CIO of the aerospace company, all the way down to the CIO of the manufacturing company, which is looking to just master the data and make sure that they have all the governance and compliance in place. So that's the AI-ready data platform. Now, another question I get a lot when I'm talking to a lot of customers and partners is, what does AI-ready data mean anyways? There's a lot of opinions. A lot of opinions. So we essentially, being people who are nerds ourselves, we said, forget it. We're going to just define it. We're going to define exactly what data— AI-ready data is. And we defined it in 7 factors and 21 measures. These are the 7 factors. It's things like accessibility, discoverability, explainability, context, quality, all of these factors. And by the way, there's an assessment available on the website right now where you can go and take this assessment yourself and be able to check it out and see where you stand in that journey. And not only that, I'm going to give you another little breadcrumb. Later on in this session, you'll hear about a big announcement that we have, which is now we have converted that entire assessment into a product which we are launching today. So that product will be something that you can now take literally if you are working with Informatica and be able to do that assessment yourself easily. More to come on that. So just keep checking in. There's going to be an amazing demo that comes into that as well. And this becomes important because as a customer, I know I was taking another example. They created a massive deployment of agents for customer support. And then they ran into issues with deflection rates. Customer support agents, as you all know, are one of the most popular agents out there. But they ran into issues with deflection rates. Deflection rates, instead of going down, they went up. So there were more deflections happening to humans versus less. And they said, what just happened? When we went in and we did a full analysis using this AI-ready data assessment, one of the things we found was they hadn't connected 2 data sources. The discoverability, which is one of the factors there, wasn't covered. Solved that problem, the deflection rates went down. So this is a real, this is a real story. This is not theoretical. This is all very practical. So that's what AI-ready data is. Now, you know us as the industry leader in all of the segments. This is where I told you that we cover all sides of the journey. This is the foundational journey. Anytime you're doing any kind of data management, master data management, governance, analytics, any of that, Informatica has got you covered. And then as we move forward and we go next into the journey of agents all the way on this side, you can see the kind of customers we have who are already on that journey. And I want to give you 2 examples of those customers. Example number 1, UKG. Some of you may have heard them speak earlier in the session as well. UKG use Informatica to do an internal data audit, the cloud, the data catalog, And in doing so, they were able to save almost 20x in tokens.
How?
You know what they were doing before? They were literally taking an XML parser for all of their data objects in all of their data stores and putting that into a, a Dynamics DB database and being able to cover that with a little structure of a model. And that's it. The amount of tokens they were burning in the process were ridiculously high. They changed all of that, grounded it with the catalog, and suddenly that 20x token reduction happened. So this is real. This is a real story, and the cost savings are also real. Vestcom did that for their customer service agent, and they grounded it with our MDM. So the unique profile that was being used was so helpful that they were able to reduce the call hold times. You can now identify who the person is calling in versus having to wait for 2 minutes or 3 minutes in the process. So these are real tangible examples. Continuing on, I mentioned uniquely neutral. This is one of the most important characteristics. We're not going to change that. We're going to work across the entire ecosystem. Salesforce, of course, we're going to deeply integrate with it, but we're going to continue to do that across the entire ecosystem of partners. Now, what's coming up next? What does the actual AI data platform include? We look at it in 2 worlds. This is one of the most exciting slides for me because we talk about exactly what we are announcing right now and then what's coming up in the future. If you look at the left side, it's data for AI. Simple. What does that mean? How we are taking all the different datasets and making it tangible, applicable for AI agents. So today we are announcing the AI Ready Data Intelligence. That's the product I mentioned where you can now go in, and if you're using any of the Informatica products, you can test where you are in your AI data readiness. Number one. And number two, And this is super exciting. You're going to hear more about it later on. We are announcing an extension to our Anthropic partnership by enabling a Claude plugin for our Data Catalog that is available now starting today. And what that does is if you are an existing Anthropic customer, you can now use that to ground your Claude co-work and Claude code and be able to leverage it. More to come on this as well. And then in addition to structured data, we are enabling unstructured data for our catalog and being able to combine all those assets together. And one of the biggest things is that we are announcing the general availability for our Informatica headless, which is coming up in October, that enables you now to capture the best of all of Informatica in your own IDE. You could be using Claude, Slack, or anything else. You can capture that. So that's data for AI. A lot of very cool things happening there. Which ultimately lead to your outcome for reduced time for any work you're doing with data. Then AI for data. How, if you are a data steward, if you are a data engineer, if you're a data scientist, how do you reduce the amount of time you take on common tasks? How do I master my data? How do I get my right bulk editing done? How do I get the enrichment done for these things? These are common tasks that have happened for the last 10+ years. How do I reduce that time? That's what we've done with our Clair family of agents. Some of you're familiar with them. We're doubling down on that. So in fact, you're going to see more and more of those agents over here. And then in the future, you will see that we'll give you the ability to create your own agents. We're not the ones that we should be creating the agents. We'll give you the ability to create those agents. So this is a really exciting roadmap. It also ties in really well with Salesforce. And speaking of that, this is what a better together with Salesforce looks like and Informatica. Some of you may ask the question, how does it relate to Data 360? What do I do with Data 360 if I'm a Salesforce customer? Well, we now have MDM integration with Data 360 that is enabled today, and that is something that you can use as of right now. We also have our catalog, the Data Catalog, integrated inside of Data Fabric from MuleSoft. So those of you who were there in the MuleSoft keynote, if you saw the Agent Fabric, which is incredible product, you can now ground the Agent Fabric with the Data Catalog as well. And then there's going to be MuleSoft and Data360 connectors that are going to be coming up as well. And then finally, all of the headless capabilities that we have, all the player agents that we have enabled through Slack and Teams and all the other interfaces. So this is really where we are going. This is what the future of data looks like. This is how we are going to enable the ecosystem for the entire enterprise and for Salesforce together. Now we're going to switch into the 3 big areas that we're going to talk about. One is going to be around our Trusted Enterprise AI Harness. So Trusted Enterprise AI Harness, for those of you who were there at the keynote that Rohan had on Tuesday, we had— that's where we announced this Enterprise AI Harness. And Informatica's role in the Trusted AI Harness is so important because we are providing the context and the governance. And today we're going to be focused much more on the outcome. the business outcome of this trusted enterprise AI harness. And to talk more about that, I want to invite Des to the stage. Des.
Thank you, Savinay. So if the— if GenTech Enterprise is organized around customers and the business outcomes you want to deliver on, the question you need to ask, how does AI make that possible? Because models are incredibly intelligent. They can reason, they can plan, they can act, but they're also probabilistic, which means that they're giving you answers that are likely to be the answer. But in the enterprise, you need determinism. You need answers that are grounded in truth and that you want answers that come back every single time. Because models don't know your business. They don't know your data, your customers, your workflows, your policies. And no matter how intelligent the model becomes, intelligence alone can't drive your enterprise. What you need is something around it, and that is something that we're calling the AI harness. Now, the AI harness is everything around the model that makes your enterprise work with artificial intelligence today. So AI connects to all of your customer data, your workflows, your policies, your guardrails, so it can act with trust. And the best way to think about it, the model brings the intelligence, and then the harness brings your business. But let's actually talk about what's bringing this transformation. And you'll see here that it starts with the model. And as you can see, a lot of customers have a lot of choice with this. In fact, my model choice is Claude Code. I use it every single day. And as I said, models are incredibly intelligent. They can analyze, they can plan, and they can create. You should see what my kids are creating these days. It's insane. But models still don't inherently know your business. And your business is everything that you see here. It's all of your customers, your data, your workflows, your applications, your governance, your institutional knowledge, your history. Because a model will know what an order is, but a model will not know that the order is associated to the customer. And a model won't know that that customer has called in 3 times because they haven't got their order. And then the model will also not know that that inventory is actually stuck in Italy right now. Because at the end of the day, the model needs something connected to your business. And that is where the AI harness comes in. This is where you're starting to now connect all that business understanding, all that business context, all of your customer data so that it can actually have relevant answers. It also can be connected to your systems, your agents, your APIs. So it can act on your behalf, but it also has all the governance, the security, the guardrails, so you can control AI with all the cost controls that you have across your organization as well too. And when these things come together, when you have your, all the power of the model of intelligence with your business, it can also now start to deliver all of your business outcomes. It can start to drive revenue, it can start to drive customer delight, it could also do reduced costs. It can also improve productivity and ROI, but it allows you to do it so much faster. And so we're really excited because now Salesforce is introducing our own enterprise AI harness. And the idea of it is that we're taking the best capabilities across our entire platform today from Data360, Informatica, Tableau, MuleSoft, and bringing into one system with 6 composable capabilities. And as you see here, we have it with trusted models where you get model choice. You also have trusted context, which is where informatics is gonna fit in, trusted agency so that you can actually reason and plan and act, get that determinism and probabilism with the agents that you build, trusted action with the workflows that are connected to your entire business, trusted governance, which is not just across data but also AI, and then trusted security because agent identity is something that's super important now in the enterprise. And it all sits on the AI control plane. This is actually driven with Agent Fabric today. This is actually available today with MuleSoft. This is where you have your observability, your governance. This is where you can do cost controls. And this is really what we call our enterprise AI harness. But as you see here, Informatica has a big role in here. We, we wanna make sure that you see that all the capabilities when it comes to MDM and governance and lineage, all the things when it comes to discovery, all of that's gonna be built into this harness. And I would love to actually introduce someone who's actually doing this today. With their AI-ready data, Deloitte, can we invite Mark Cameron to the stage? Welcome, Mark.
Hi. Hi, Jess.
How are you? I'm good. I lost my voice last night. I couldn't stop yelling. I loved Gwen Stefani last night, so you can hear my voice is really bad.
What a big crowd, eh?
It's a good— it's an amazing crowd. Standing room. All right, so Mark, why don't you tell me what Deloitte's been doing and what was the business problem that you were trying to solve?
So Deloitte, very complex organization, 150 different countries we operate in. We've got a huge product range from tax audits to, you know, O&M, which is operate. We want to make sure that our sales teams, our client managers, have actually got information at the time of conversation with our clients. So that's the problem we were trying to solve. So we were given this way back and say, how are you going to fix this? How are you going to pull this together?
That's awesome. So then once you had this, you know, oh my gosh, I love this little cool guy right now. I see him on stage. How did you make sure that the data was reliable for this experience that you were trying to build?
Right. So it's been spoke about simple things like master data management, knowing your data, knowing the context, quality of data. That was a huge big initiative we drew on to actually make sure when we actually did Agentic and AI, our data was in good shape. And there was a long drawn-out exercise to make sure the context of client information, the context of the client was all pulled together ready for our then initiative to actually go headless and start building out capability like Dottie, which we'll talk about.
What did you call this? Because it sounds like you took your customer information with business outcomes. Did you have a name for this?
Yeah, yeah.
What was it called?
It's Relationship 360. Dottie was kind of the code name and thing we gave to it. give it a life.
It's a little guy, isn't he cute?
Yeah.
Yeah.
I love it.
She's a she.
Oh, it's a she. Sorry.
That's who you're talking to.
Sorry, buddy.
Sorry.
Sorry.
Yeah.
Yeah.
Yeah.
So, so, and we wanted to brand this and we wanted the whole thing about bringing capability to our sales team is actually brand it and they understand it. Now underneath that, there's a lot of, a lot of plumbing.
That's amazing. And now what did you see the results?
I'd love to hear what was the, like, outcomes that you were seeing from this So firstly, experimenting and understanding our data and what's the right information, what sales intelligence, what you have to give to our end users was quite key. And a lot of that was a learning curve both with your team as well as our own people to try and pull together something that's going to be really effective, works first time, gives the right information.
And what's next for you guys? What are you guys working on now that you're like looking forward to?
There's just so many things. I mean, this is amazing. We're actually picking up ideas just from the audience. just as we're actually being at this event. So we've got a huge big pipeline. Agent Fabric's in there, all the MuleSoft, Informatica driving that hard. So yeah, we're going to be busy.
That's awesome. That's wonderful. Thank you so much, Mark. I really appreciate it.
All right.
Thank you, everybody.
Thank you. All right. Well, now you heard from Savanay. We would love to actually bring Sumit on the stage to talk about your AI-ready data intelligence. Welcome, Sumit.
Thank you, Des. All right, can you do a quick show of hand? How many of you here are eager to move your agents in production? A lot of you. No wonder why you guys are here, the last session of Dreamforce. So today, I have a different question. How will you know the health of your data which is feeding to this agent? Now remember, what I mean— what I meant here is now what you think is the health of your data, not what your team thinks is the health of your data. I'm looking for actual factual health of your data. And I've spoken to many of you. For many of you, it's all done through manual audits. Somebody sitting next to you will say, hey, this data is good, this data is bad, this data needs work. Now imagine putting such kind of agents in production. And that's why we have launched this new product called AI Ready Data Intelligence product. It will connect directly to your catalog, extract the right metadata, data lineage, everything you need to give you the right factual data score of your— which is basically the health of your data. So what we did here was very unique. We have taken these 7 factors which are measured across these 21 dimensions, and then towards the end you'll get a score. And here we have not taken the average score. We have taken the score which is a minimum out of all these 7 gates. And there is a reason for that, because your agent will hallucinate by just having one data quality problem or that one lineage which you have not fixed or that one missing data source which you have not discovered. That's all it takes for your agent to hallucinate. And that's why we have taken the minimum of this score. So when you run this utility, you will get the score like this. Along with that, you will also see how you can improve this score tomorrow, and then what you can do, how you can improve your roadmap for future. Now, you can run this against any use case of your choice, whether you are running support agents, autonomous agents, or you're building your chatbot, you can run this utility today. So what are you waiting for? Take control of your data, find that missing link, and make sure that you have the right data ready for your AI before the agent finds that missing link. Now, if you guys are as excited as I am about this product, I'm sure you want to see this product in action. So for that, I'm going to go to my demo laptop and give you guys a quick overview of the demo of this product. All right, so in this use case, I'm going to play a role of an AI engineer in Meridian Corporation, and I've been tasked to build this support agent use case. And before I do that, I do want to check the health of my data. So what I've done here is I've already given this Claude the access of my AI-ready Data Intelligence product. It has connected to my catalog, which is sitting in Informatica, and it seems that it has already registered 530+ sources. So if I go to Informatica and show you how it looks like, this is how the Informatica catalog looks like. It has 530+ sources. I can see all the data classification rule. I can also see how the lineage coverage is, all the data policies, and everything around this. Now, if I go under Browse, I can also see where are my data sources. It seems that I have data sources sitting in S3. These are some unstructured files which are sitting on my S3 table. These are all the KB articles. Some of my data is sitting on Oracle, Google, and SAP. So these are all the data sources which is already being cataloged for my support agent. Now, when I go back to my Claude, I see that just by giving that simple prompt, Claude was able to scope down the scope from 530 to these 8 sources which are needed for my support agent. Now, once I agree to the scope of this, what will happen now is Claude will directly connect to my catalog, and it will start accessing the health of my data across those 7 factors which we have discussed before. Now, I'm using Claude here. You guys can use any harness of your choice, even if you're using OpenAI, AWS, Azure, Google, whatever AI harness you have, this will work as it is. Now, as it's going on, you can see that it has started accessing the health of my data across those 7 dimensions and 21 factors. And then as it goes on, it can find the issues around my data. Like in my case here, it is already picking up some of the, some of the data sources like Salesforce Service Cloud, SAP. And now it is trying to set up the whole measures and it started checking the health of my data. And as it's doing on, it is also finding the issues around my metadata completeness, which is not looking good. There is some issues around my glossary coverage and all of that. Now I'll let the Claude run and I'll go and show you how the actual report will look like. So this will be your end report. Once you complete this whole product, this is the kind of report it'll generate. It'll give you a scorecard. In my case, it was just 42. Now, if I go deeper, I will find out that the main reason why I got the score as 42 is because I have this problem around discoverability. And here, the big issue is that my unstructured index is sitting at 12.5. What does that mean? It means that my agent has access to only 1 out of the 8 possible KB articles. Now imagine if you are a human trying to solve a support case, you only have access to 1 KB article versus 8. So this is the kind of problem you will face if you would have put this agent in production using this data. Now along with that, you will also get the heat map Of the other dimensions. Then you will also get all the detailed scores. And then this was very interesting to me where you will see that what happens if you would've put this agent in production, how will it hallucinate? So you will see how the problem will happen here. And then towards the end, we give you what are the 5 opportunities, how you can fix the health of this agent. Now here is the kicker. We don't just give you the scorecard and give you these opportunities. We even take this product one step further where we can even help you to fix some of these problems. So what I've done is I have fixed the first 2 problems using the same product, and I have reran the report, and now my scorecard looks way better. It's 73 versus what it was before, 31 points better. So to summarize what you just seen in this whole demo, what I showed you is using AI-ready Data Intelligence product, it can help you to solve 3 things. Find the right data. You can also find the health of your data, and you can also really immediately solve the issues around your data. Now, if you guys are building your agents today or you have plans to do that in future, Please scan this QR code to get access of this product. If you are an Informatica customer, I can give you the access of this Claude skills, which I've shown, which I'm using my Claude. If you are not an Informatica customer, no issues. We have using the same QR code, you can take the same product in an assessment way. So before I call on Gaurav, I'll just leave you with this comment. Your agent is as good as the data it's fed to. I repeat, your agent is as good as the data it's fed to. And that trusted data, ladies and gentlemen, is what Informatica provides. So with that, I'm going to bring up Gaurav Pathak to talk about the next exciting announcement.
Awesome. Thank you. Thank you, Sumit. That was an amazing demo. He showed us how we can assess the data that we have in our organizations, find the health, and see how AI-ready that data is. Another big round of applause for Sumit, please. Now, all through this week, you've heard about how Anthropic and Salesforce have joined forces to deliver Claude models within Salesforce's workflows, data, and governance. And I'm thrilled to announce the new Informatica plugin for the Claude platform. Now, what does this plugin do? Claude is already doing amazing things for you. You know, these models are becoming so intelligent, I sometimes get intimidated just talking to them.
Right.
So with Sonnet, it feels like I'm talking to someone on my level. But with Opus, it already feels like, you know, I'm talking to an engineer who's waiting for me to come to the point. And with Fable, it feels like I should have prepared for this conversation a little bit better. With enterprise use cases, it's not just intelligence that you need. You need to give it enterprise data, and that's where the Informatica plugin comes in. We bring highly contextual, relevant, trusted, quality data to Claude, and let me show you how we make that happen. So for that, I'll go to my demo laptop, and we will begin by showing you first, if you don't have the plugin, what happens. Now, here is a report that was created by Claude. We just pointed Claude to a repository of spreadsheets. We gave it a database MCP, and we asked it to figure out what was the Q2 revenue. And it comes back with this beautiful dashboard. This is an HTML local file. It got all of that data, and it created this visualization for me. Now, this is great, but the question is, can you trust it? And now here, there are many things that have gone wrong. First, because the models don't understand what is the meaning, the organization-specific meaning of Q2 or revenue, the context behind it, what are the policies in which we should look at all of these numbers, It gets those wrong. It does not understand what are the right datasets to pick up. It found 40 different tables in the warehouse with revenue in them, and it still picked up a spreadsheet, which was not the certified source of getting this data right. In this case, the numbers don't even add up to the number that we were looking at earlier. And there was no check on quality itself. The data that was picked up, whether that is of high quality, whether the pipeline that was putting this information in, in the dataset, whether that ran last night successfully. None of that grounding ever happened. So those are the kind of mistakes that the model can make if it's not provided the right, relevant, trusted data that it needs. Now, this is where Informatica comes in. First, you already saw the Informatica data governance and catalog that Sumit was showing. This is Informatica. Here I'm showing you those cataloging and governance capabilities as well. We gather very detailed metadata about every single important dataset within the enterprise. We have automated scanners, hundreds of them built, that look at each of these data sources, extract very deep metadata, Like what are the key data elements behind it? What is the lineage? Where is this data coming from and where is it going? We extract relationships from it. What are the joins that are possible? What are the right business meanings that are associated with it? And we assess as well as remediate data quality so that this, the dataset that users are looking at, are relevant and trusted as well. Now, what we have done with the Informatica plugin, which by the way is now available on the Claude Marketplace. So that's really the big announcement here. So you can search for Informatica now and get this plugin. If you are an Informatica user, you can now grab this plugin and connect it to your Claude platform, whether it is Claude Code or Claude Cowork or Claude Chat, all of this is becoming one. You can now connect to your enterprise data using this plugin. So here within my Claude desktop, I have already configured the Informatica plugin, and I ask it the same question again. Can you help me build a comprehensive HTML dashboard for the EMEA sales in Q2? And look at what it does right now. It now goes and connects to this Informatica plugin. It brings in the right definitions. It brings in the right certified data sources for answering that question. And then it goes ahead and picks up the right datasets with the right quality, looking at lineage. All of that rich metadata that was earlier going to humans in the catalog, is now available to agents as well. And that allows us to bring trust indicators, trusted data, not wrong data, to the users who are making these decisions as well. So if we go back to our report that we were looking at earlier, now these numbers are trusted. No problems with definitions, no problems with certified datasets, The right pipeline that ran yesterday is the dataset that we picked up for that, and with high quality and lineage as well. So that really is trusted, relevant data that we can provide to agent all through Informatica. Thank you very much. Now we have done all of that. There is also one point about you know, money costs. And nowadays we measure money in tokens. So let me invite Kaiser here, who's going to talk about all about economics of tokens. Kaiser, please come on stage.
Thank you.
Thanks so much, Gaurav. Thanks so much. Love this guy's energy. I get to talk about the money. That's the funnest part, right? So this is my first Dreamforce. Guys, I think we've all been to so many conferences. I've never been to a, like, technical business conference where I feel like I'm at a concert and the audiences are so engaged. So I want to hear— we have a whole— we have standing room only. How has your Dreamforce been so far? I feel like I'm like Justin Bieber up here. I get to have the crowd so engaged. It's incredible. I've been so excited about this session, honestly, the entire week. I've loved the demos. I've loved what we've announced, the new capabilities, the new solutions. And it's all about what? What have you heard? Delivering you AI-ready data, right? And from cataloging all your data to governing your data to changing data management tasks that used to take days and doing them in minutes. We saw first from Sumit an incredible demo. It showed what? What's so important? You first have to understand and have a complete diagnostic of your data. Is it ready for AI? And then Gaurav took it to the next step. We're bringing your AI-ready data to you and your favorite tools that you like to use, like Claude. Now, I want to talk about, as Gaurav kind of gave you a little bit of a teaser, about tokenomics. I think it's something that we can all relate to. But first, indulge me a little bit. What do you think these numbers represent?
200 million.
Up to $500 billion. For me, I would love— my mind immediately goes, that's my current 401 and then my expected retirement portfolio. I would love that. I would be in Hawaii going full ohana, having a great time. But sadly, that's not it. And I'm here, go at a conference again, but it's been amazing. It's been amazing. What do these numbers actually represent? This is really interesting. This was the token spend and usage of a major rideshare company over 2 years, starting from $200 million to $500 billion. You take a step back, that's an insane amount of increase in such a short amount of time, right? Now, how does this happen? It's all about, did I have AI-ready data? Now, some of you might have already experienced this, tons of token usage, Right? Maybe you're going through it right now, and if you haven't gone through it, you're hoping, can I avoid this? And that's exactly what I want to talk about. Now, it's so important as we've talked about the 7 factors of what constitutes AI-ready data, and it's not a slogan. It's something that you can measure, you can understand, and you can constantly improve on. So that's exactly what we decided to do. We decided let's measure the impact that AI-ready data has on token usage. We partnered with a third-party independent, and we did a benchmark study to see what's the impact of starting with AI-ready data versus not. And we had a few insights I want to share with you. One, we saw that you could reduce your token costs up to 30x. 30x. That's incredible. Now, why is that? We know that agents are built to reason and reason and analyze and discover and search and execute. So if your agent doesn't know what data to look at, what is it going to do? Naturally, it's going to reason and reason and search and discover and analyze and execute over and over and over again. But if your agent doesn't have to do that because you've pointed it to the right data, that's when you start seeing those cost savings. And we found a few capabilities that really make this happen. We've talked about them today. A trusted catalog. That allows your agents to discover the right data, understand it, trust it, and be able to execute on it without searching through millions of raw data objects. Also, you have to have metadata that's classified. And then lastly, of course, you need that context that we provide and not just do tons of context stuffing. The second insight that you see here in the middle, this is almost more important. We saw that even if a data estate Grew 50 times over. And Sabine talked about that at the top, right? We're gonna see more and more and more data that agents are creating. Even if a data state grew 50 times over, the token usage and spend increased less than 1.5x. What does that mean? The agents were efficient. The agents were efficient at scale. So even as your data grows, your token spend won't grow as parallel. And that's exactly where you get those cost savings and you can turn token spend into business value. Now, you can download this. That's the QR code right behind me, and you can get this benchmark study. You can see the methodology that we used and the capabilities that we implemented to make this change so you can realize the same benefits as well. I see a few people taking a picture. I love that it worked. Now, I want to go and talk to someone that's gone through this journey of AI-ready data. So please join, join me, James Brady, Chief Digital Officer of Baker Hughes. Thanks so much for joining me, my friend.
My pleasure.
Thank you so much. So let's start with, tell us a little about Baker Hughes and your role.
Okay, so I'm the Chief Digital Officer of Baker Hughes Oilfield Services Equipment. So Baker Hughes is about a $30 billion energy technology company that spans both sides of the energy equation, right? We go from finding either the hydrocarbon molecules or oil or hydrogen, you know, gas, geothermal discontinuities. So the source, and we help our customers find extract. And then we also provide energy industrial outcomes, power generation equipment, equipment that's used to make fertilizers, etc. So we play on both sides of that. I'm firmly rooted right now on the oilfield services and equipment, but we work together to create solutions across that value chain.
Thank you. Thank you. Now, you mentioned outcomes. A lot of people in this room, you probably have a remit and your charter is to Use data and AI to improve business processes and business outcomes. How have you done that at Baker Hughes?
Yeah, so I've actually got 2 remits. So one is I actually am in charge of all, pretty much all the software that's developed, both the external solutions, but then also the systems. So systems efficiency, enterprise systems. So like, and so the outcomes will almost seem kind of mundane from what you've been talking about, right? We did a large central finance, so CFIN by SAP. That's not an advertisement, we just did it, right? And in order to do that, you find very quickly that you need to have very, very good data control because it does transaction replications. And when your transactions don't replicate well, you know, when you start looking at more than 99.9% accuracy, you end up going the wrong way on support. The whole point of the central finance was to be able to close in a month. I'm sorry, close a month in a day.
Yeah.
And then close a quarter, I'll just say faster than we do now, right? So one outcome, that we look at with data is just our ability to run these complex, semi-automated closed processes. Another would be just an example of our billing and collections efficiency. All of the above really require a certain amount of data integrity to do any of that stuff.
Totally. Tell me a little bit more about that because I know you have a huge remit and a lot of different technologies that you use. But me and you have discussed the importance of mastering that data first and how it impacts everything downstream. So how has master data management helped you with other initiatives that require it?
Yeah, so it was interesting. So I think the central finance integration that we did was probably the best thing that ever happened to us in that it actually forced us to get our data ready for AI. And so we had to go in, we had to actually master our customer data, that was probably the most tricky. We had to master all of our financial entity or financial legal entities, financial cost centers, profit centers, and then how they're spread across our systems. So we ended up doing a big reference 360 approach in order to align all of our data across, yes, all of our ERPs. Probably anybody that's with a size of company like mine realizes that all is more than 2. It's a lot, right?
Yes.
So there was a certain survival aspect of having to get our data ready for CFN, but then there was the opportunity. So we have a big lakehouse like probably everybody does, and we started cataloging it and building natural language queries into it. What happens? Of course, you see that it's giving you the wrong answer, right? Until you start going in and hacking. Well, shell is sometimes spelled with one L, and sometimes with 2 Ls and all that, right? So we actually had to go in and leverage a lot of the mastering of the core data just in order to get a lot of our lakehouse answers and then agents working well.
Thank you.
Thank you so much. Now, as you get your data ready for AI, what's next for Baker Hughes? What are you excited about what's coming next on the roadmap?
There's a lot of cool things that we're doing on that, right? Like I said, we're at the start of agentic workflows across systems, and that's a lot of what we had to do, right? There's some opportunities around automation of our billing processes and some other things. So now that we've helped lay that foundation, right? And that foundation is not built on just cleaning up your master data store. It actually— we like to say you clean up the lake, but you have to clean the rivers, right? And so a lot of this now is the propagation of data quality rules and data updates via the event broker, via Address Doctor, and some of these things. in order to keep the data pristine. So I'll say that while the initial challenge was just getting it decent enough, the next challenge is then assuring that we can keep it clean and that we can get these downstream value, downstream value, which is largely business process execution across multiple systems.
Right. Thank you so much, James. Thanks so much for joining us.
Thanks.
I can take that too if you want. Awesome. I love that. Okay, well, we're almost done. Stay here. We're going to close it up real quickly. Sabine, back over to you.
All right. All right. Okay. Thanks. That was awesome, Kaiser. Okay, we're back to the summary right now. So you saw a little whistle-stop tour of the one thing which I think we heard more about in this session than the word agents. What do you think that is?
Data.
Data.
But what kind of data? AI-ready data. Yeah, you're paying attention. Okay, so AI-ready data. This is a real thing. If when you walk away from here, when you ask your own team members what they're doing for enabling their own agents, ask them, is your data AI-ready? Think about that. And when you think about that, you should think about what Informatica and Salesforce are doing for you. That's a great starting point to have this conversation and go from there. On the other hand, I go back to what I started with, which is in some cases you may not be at that journey yet. You may still be at the journey where you are just enabling your existing datasets and mastering your existing data. And that's okay. That's perfectly fine. We're there for you for that journey as well. But the one thing that I want you to sort of think about when you think Informatica now is we are going to be that AI-ready data platform. We are going to be the companion for you as you're going around this journey from enabling that data platform, modern data platform, to the AI-ready data platform. We're going to be enabling that as part of the trusted enterprise AI harness. Des talked about how important that's going to be for us in the future as Salesforce. We think that's going to be one of the de facto enterprise AI platforms for the future, and Informatica will play a big role in that. Keep a lookout for that, the Trusted Enterprise AI Harness. You also heard about what actually AI-ready data means, the assessment, the product. Try it out. It's super simple. Try it out today. Try it out when you go back and ask your administrators, ask your teams to do that. It's very simple and it gets you started, something very tangible that you can do. And then finally, the money piece. The money piece. Yes, it's real. There is a lot of value to be had in terms of actual token savings. when you get the right AI-ready data.
Yeah.
So thank you all. Thank you all for being here. Thank you for being at Dreamforce, and we look forward to partnering with you in the future. Thank you so much.
