What happens when an AI agent makes a business decision using information that's already out of date? It's a question I wanted to put to Stephen Deasy, CTO of Confluent, when we met at Current 2026 in San Francisco. After attending more than 20 technology conferences this year, I've heard countless promises about what AI agents will eventually achieve. But there's a growing problem behind those ambitions. Businesses are still struggling to get accurate, current information out of the systems they already own.
Confluent has introduced Streamhouse, a data architecture designed to address that problem. Stephen explains why the company believes existing approaches can leave applications working with delayed or incomplete information, and how the new architecture brings streaming, processing, governance, and querying together. We also discuss Lightning Tables, one of Confluent's major announcements at the event, and what direct access to current Kafka data could mean for developers and their businesses.
I was particularly interested in what this means for companies that have already invested heavily in data warehouses and lakehouses. Stephen makes the distinction between analyzing past performance and running business operations using current information. He explains why Confluent sees these approaches working together, and why existing technology investments can provide the foundation for introducing Streamhouse.
We also discuss the experiences of Instacart and Citizens Bank. Stephen recalls how Instacart faced a sudden increase in demand during the pandemic, when accurate information about orders, inventory, and deliveries became especially important. He also describes how Citizens Bank improved loan fulfillment processes through real-time data streaming. Both examples help explain why the speed at which information moves between business systems can affect customers directly.
The conversation also covers IBM's relationship with Confluent, open technologies including Kafka and Flink, and the announcement that Tableflow is now generally available on Google Cloud. Stephen shares his thoughts on model evaluation, human oversight, and how companies should measure returns from their technology spending.
One of my final questions concerns the misconceptions surrounding Streamhouse. Stephen explains that the architecture brings together approaches many organizations have already been working toward, and that its value lies partly in making those capabilities easier to use together.
With so much money being invested in AI, are businesses paying enough attention to the information those systems depend on? Listen to our conversation and share your thoughts.
Useful Links
For further reading, Confluent has published its Streamhouse vision, a Q4 product bundle overview, and details of the Streamhouse webinar. The Streamhouse Quickstart offers a hands-on starting point, while the Streamhouse Working Group announcement describes the wider industry effort. Confluent is an IBM company, and its platform details are available at confluent.io.
[00:00:04] Welcome back to the Tech Talks Daily Podcast, where this week I'm recording at the current 2026 event in San Francisco. And Confluent has announced Streamhouse here, and essentially it's a new data architecture designed to help businesses work with information as it changes.
[00:00:25] And yes, we've heard plenty around AI agents this year, but what happens when those agents are making decisions using information that is several minutes or even hours out of date? And how much existing technology will businesses need to change before they can put these systems into production? Well, joining me today is the CTO of Confluent. We've stepped away from the show floor to discuss what Streamhouse actually does.
[00:00:54] Explore why lightning tables matter, and how businesses such as Instacart and Citizens Bank are using real-time data to address practical operational problems. And we will also examine that relationship between Confluent and IBM, the role of existing warehouses and lake houses, and what companies should be measuring when deciding whether investment in any infrastructure is worthwhile.
[00:01:20] But enough from me. Buckle up and hold on tight. It's time for me to beam your ears directly to the show floor. So we are recording here today at Current 2026 in San Francisco. A lot of big announcements here, haven't there, today? But to begin with, can you tell everyone listening and watching a little about who you are and what you do? Yeah, thanks, Neil. So I'm Steve DC. I'm CTO at Confluent. So I'm responsible for the engineering, the product operations, everything about Confluent Cloud and Confluent Platform.
[00:01:50] Awesome. Now, one thing we are hearing a lot about at the moment, Steve, is the enormous investment that's going into AI models and infrastructure. But there's another challenge around making sure AI has the right business information at the right time. And of course, I'm building up here to that big announcement today is Streamhouse. So what problem is Confluent trying to solve with Streamhouse? And why is now the right time to talk about the architecture?
[00:02:15] We think that from talking to customers, what we're seeing is that one of the big blockers to taking your agentic workloads into production is the availability of governed fresh data and the approaches that customers have taken, things like MCP-ifying access, things like moving data into lake houses have not been working for various reasons.
[00:02:41] And so what we see is the next evolution is a Streamhouse to really allow that real-time data be available to your business for software to make real-time decisions to move your business forward. And for the non-techies listening, maybe we've got a few business leaders. How would you describe the announcement of Streamhouse today and what it would mean to the people listening? Yeah, so what we think about is use cases.
[00:03:07] For example, banks are making loan fulfillment decisions. Consumer companies are making ordering decisions and fulfillment and so on. And so the ability to run those businesses on stale data is no longer feasible. And so in the business world, the ability to move with real-time, with the current information and all of your business context available to those systems
[00:03:33] means that you can move faster, produce better outcomes for your customers. And so it gives you a huge business advantage. And of course, many organizations have already invested heavily in data warehouses, lake houses and analytics platforms. But where does Streamhouse fit into that picture? And what would you say to a CIO who's concerned that this could mean yet another expensive infrastructure overhaul? Yes.
[00:03:58] We think that what's happening in the world, what we see, is we're moving from a world of business intelligence. So in business intelligence, data moves in batch. It gets periodically reviewed in dashboards. A decision gets made by a human and fed back into the business. We're moving from that world of business intelligence to a world of artificial intelligence, where decisions are being made continuously. The data is streaming in real time.
[00:04:28] And that unblocks these types of use cases that previously were maybe not as critical or in an AI world are now expected in the business realm. So the way we think about it is lake houses are not going away. Lake houses help you analyze the business and they tell you what happened. The stream house is helping you run the business. It's helping you make decisions and run the business rather than analyze the business. So they're very much a complementary setup.
[00:04:56] So for CIOs, I would think of this as your real-time applications unlocking the value of the data that you're already collecting through your services, your SaaS applications, your databases, and making that available to these workloads that are now craving it. And without it are blocking your ability to take them into production or see the return on investment that you're making in that infrastructure over time.
[00:05:24] And we're recording this towards the end of 2026. This year, I've been to 20 tech conferences. Predictably, every single one has mentioned agentic AI and agents. And, of course, AI agents promise that they can take action on the behalf of a business. That's everything we've seen this year. But I think we're now moving on to the next phase. And a few people listening might be concerned that an agent-making decision using maybe yesterday's information could create very serious problems. So how does Streamhouse help applications, AI agents, and indeed employees?
[00:05:54] Does it keep them working all from the same current view for the business? Very much what we see is the investment in your data streaming platforms to date is what unlocks this. And the really nice thing about the Streamhouse that we unveiled today is that this is something that is building on that investment already. So you're streaming data, the governance of that data.
[00:06:22] And so we've been saying for quite a while that the ability to shift left that governance to where the data is created unlocks the safe use of this data later. You have forward and backward compatibility. You're able to govern access to the data. You can stream it to where those use cases are necessary. And so we think that that is really the key unblock today with one of our key additionals. One of the other big announcements today was the query side with our lightning table.
[00:06:51] So the introduction of that, which augments where we already have connecting to your data, streaming your data on the best Kafka platform out there, wherever your workloads are, being able to process it now in real time with things like Flink. And then query it through things like lightning tables or land it in open formats like Iceberg and Delta really allow you to unlock your entire ecosystem rather than any one piece of it.
[00:07:19] And for the techies, what excites you about that second announcement? What really excites you then? Yeah, I think there's a couple of things really that stood out to me in our announcements today. Obviously, lightning table is super exciting to me. So we had and we still have a real time context engine, which allows access to your Kafka stream of data in a think of it in a table format.
[00:07:42] Now with lightning queries, you can access a true SQL, true APIs, and you can access the data with very low latency and we're able to govern that schema for you. So as your application teams evolve, your downstream consumers can take advantage of that really, really quickly. A couple of other standouts for me today was very exciting to see our web hook connector. So a big part of our data ecosystem is being able to connect in where that data is.
[00:08:10] And one of the big unlocks with a web hook connector is that producers of events can now put it onto the stream without ever needing to interact with a Kafka cluster. They simply provide the endpoint. And from there, you can stream in that data from first or third parties. And now everybody downstream can take advantage of it. So there were a lot of exciting announcements today. There were a couple that stood out for me.
[00:08:36] And of course, no matter how exciting that tech is, increasingly every tech project has got to go through the ROI test case. There's CFOs will be paying very close attention to that after a lot were burnt on those early AI experiments. So on that side of things, can you walk me through a very real world scenario where this architecture could make a real measurable difference inside an organization?
[00:08:58] Could be something financial services, customer experience, supply chain management, where speed and accuracy of a decision directly affect revenue and cost. Sure. So a couple of great examples come to mind. You know, in our keynote this morning, we had some great demos of, you know, mocking up things like airlines and flights and delays.
[00:09:19] But maybe a real world example that came up that we're all very familiar with is customer of ours, Instacart, who during COVID, all of a sudden, the idea for listeners that are not familiar with Instacart, think of it as grocery delivery, fulfillment, things like that. And so during COVID, you can imagine you had a massive spike in demand pretty much overnight.
[00:09:43] And in there, you have to manage everything from orders to inventory to fulfillment to shipping and returns, purchasing all. And that all has to happen in real time because if I'm purchasing my groceries now, I don't want to find out later that you didn't have it in stock. Or the delivery guy left and I won't be arriving for 24 hours, things like that.
[00:10:08] So real world examples of needing that real time data connected up across these different systems. So that's one example that stands out of really bringing this to bear in a real meaningful return way. And I guess we should also mention there's also that relationship between Confluent and IBM now. Now, many large enterprises operate across public clouds, private infrastructure and decades of existing technology. Dare I say a fair amount of legacy tech as well.
[00:10:36] But how does an open hybrid approach help the businesses listening modernize without adding another layer of complexity or vendor dependency? I think there's a couple of things in there, but I think you really hit it. When we think about our approach around open standards. So we have Kafka, which is the open standard for data streaming. We have Apache Flink, which is the open standard for processing.
[00:11:03] And now with products like Tableflow, which today we announced the GA of Tableflow on GCP. So it's across all of the clouds now. Producing that in things like Iceberg and Delta format and integrating with the catalogs you already have mean you don't need to go through an entire technology refresh to take advantage of these real time use cases. We already plug in where you are. And so that's very much the philosophy that we've had and will continue with. Love it.
[00:11:33] And there's also a lot of excitement around AI agents, what they will eventually be able to do. But businesses still need to justify that tech spending today. Anything else a company could achieve with Streamhouse right now? What would it struggle to achieve with its existing data architecture? Maybe just to help people listening and watching understand what that before and after picture might look like. Yeah, maybe a good example of this that can make it real would be another great customer of ours, Citizens Bank, for example.
[00:12:03] One of the, you know, loan fulfillment processing, for example. You know, they were able to move to real-time processing and real-time stream on what is now the Streamhouse. And in there are able to reduce loan fulfillment times, increase customer satisfaction, decrease loan reprocessing and things like that. And so what we see is that by bringing these capabilities together, you can have very real meaningful impact for your business, for your customers.
[00:12:31] And by having all of that business context current and available, it also unlocks the business, the use cases maybe you haven't thought of. So now it's available in a catalog where somebody comes up with a great idea, wants to connect into some AI use cases. You already have that governed fresh data available to you, which reduces the time from idea to production very significantly.
[00:12:57] And when we're talking about AI agents making decisions in real time, anything you can add on what human oversight sit in that process, particularly when an automated decision could have financial or even regulatory consequences? I think in the agentic world, we think of it as a new type of model of operating.
[00:13:20] We're moving from worlds where we built software and we would test in regression, which we're still doing, to more probabilistic models where you have to continuously evaluate and look at the data. And it might be that a new model comes out. It might be that you have some additional data points. So if you think about the combination of your models and your tools and your context, bringing these together allows you to continuously evaluate the quality and the outputs.
[00:13:48] And so something like the stream house allows you to do that both in real time, but then also reprocess some of your historical data with maybe some of this newer context or tools. And so they're very common use cases that we see in a stream house that allow you to have confidence that you can make those business decisions in real time. And again, going back to the return on investment, sorry to go back to that, but I know it's such a big topic right now. When people are measuring it, I think it's very difficult to know what metrics to measure.
[00:14:18] So is it faster decision making, lower operating costs, better customer experience, all of these things or something else? How do you measure that ROI? I think it is all of those things and it varies. So what we see with our customers with stream house architectures is operational use cases like observability. We see financial use cases like fraud detection.
[00:14:45] So another exciting announcement today was the managed model inference. So now on Confluent Cloud, you can connect up to models that we will run for you. We introduced Google and IBM time series for things like anomaly detection. So whether it's an observability use case or a fraud detection use case, things like that, we're now able to provide the plugins for those in addition to your own proprietary capabilities within that stream house architecture.
[00:15:14] That goes back then to unlocking the business value quickly so that you're not spending time standing up and building infrastructure. Instead, you're solving real business problems with the data you have in a secure governed environment. And one of the things I always try and do on this podcast every day is bust a few myths and misconceptions. Now, as someone that must spend a lot of time not only in a technical environment but also doom scrolling down your phone, you probably see a lot of information on LinkedIn, etc.
[00:15:42] Are there any misconceptions that enterprises have about what they need to change to make what we're talking about work today? That's a good question. Yeah, there can be so many different frustrations. And I just wondered if there are any trends in the kind of questions that people are asking all the kind of things that they have in their head. Well, I think one of the interesting conversations that I'm having at current today and with customers is that, and I don't think it's a misconception.
[00:16:11] I think it's actually what we hear from customers is almost an affirmation that what we're describing in Streamhouse and how it works in the various components actually resonate not because it's new or somehow revolutionary. It's actually putting a name on things that people have already been doing.
[00:16:30] Streaming the data, governing the data, processing it, and now querying that in real time is something that people have been doing and are feeling even more opportunity or challenge to do because AI is pushing everything so real time.
[00:16:46] And that maybe it's a good thing to do because AI is a good thing to do because AI is a good thing.
[00:17:16] stream house to you and so when we talk to customers about that there's a lot of oh that helps me frame a thing I've been seeing and feeling and now I can use. Love it and if I had a virtual crystal ball here I ask you to look into if we look ahead to the next one two three years into the future how do you see that relationship between AI agents and enterprise data how do you see that changing or evolving and what what should business and tech leaders
[00:17:42] be doing today to prepare for that future in your mind? Yeah what what what we see with you know we have a data streaming report that we've put out and there's other data that are out there and and what we see is a number of of things number one you know less than a third of customers have brought these agentic work you know workflows into production and nearly two-thirds of enterprises
[00:18:06] report that data is one of the key blockers to bringing those in and so what I see happening in the next year or two is that as you unlock these use cases and realize the value of the data you already have you start to crest that wave if you will and so I think what we see and what we're hearing again is that this helps unblock those so my prediction would be that those rates go up
[00:18:32] significantly in terms of production workloads coming in because we see solutions like the stream house unblocking or are working against some of the major blockers that we hear from customers that are preventing them getting there and there are so many people here at the moment I'm conscious you must have back-to-back meetings meetings with customers meetings with people like me not only that there's a lot of keynotes and conversations on the show floor if you were to put all those conversations into I don't know
[00:19:00] what would you be taking away from from the event what are you going to be thinking about on that plane right home this might this might sound trite but I I really do believe that the energy that we're feeling from people around this stream house capability and like explaining how it flows together really has clicked for people and so whether it's the practitioners in our customers whether it's the CIOs whether it's the business side what we're hearing across the board is yes this describes
[00:19:30] what I'm feeling yes this shows me how I can achieve this and I'm excited because again it's something that I can tap into with what I already have I'm not starting from scratch and we're not proposing that you throw out the investments you've made quite the opposite we're saying they actually allow you to build into what we've got and now with these additional capabilities these you know these capabilities now unblock workloads for you so I would say that that's probably the most exciting part when you come
[00:19:59] to an event like this you always have some anxiety about will this land is it is it what people are needing and wanting and I think we've we've seen really positive um you know excitement about that and I think that is a powerful uh moment to end on I'll include links to everything you mentioned including the reports the announcements and everything in between and for everybody watching this I'd love to hear how any of this has resonated with you what you're going to be working on
[00:20:26] in the future but more than anything just thank you for joining me today thanks so much this was fun massive thank you to Stephen for joining me here at the event I think his distinction between using data to analyze a business and using it to run a business was particularly interesting and it's worth remembering that many organizations already have substantial investments in their data infrastructure
[00:20:49] but Stephen believes that those investments can continue to play an important role but the big talking point here is around the announcement of stream house the new data architecture designed to help businesses work with information as it changes and just going back to the point that we've made several times today I would implore you to think about what happens when all your AI agents are making
[00:21:13] decisions using information that is several minutes or worse still potentially hours out of date I'll include links to all things confluent stream house and the announcements coming from current 2026 in the show notes but I'd love to hear your experiences too are outdated systems and delayed information holding back projects inside your organization or have you found other practical ways to solve
[00:21:38] these problems please share your thoughts pop over to techtalksnetwork.com send me a dm send me an audio message I'm nice and easy to find or connect with me on linkedin instagram etc just at neil c hughes the leading issue of agentic ai in businesses right now is ensuring agents act with compliance guidelines and Denodo applies guardrails across your entire data estate by aligning your company's data
[00:22:07] infrastructure under one system these guardrails perform consistently across your platform so start scaling your business and start with denodo simply visit denodo.com to learn more but that's it for today so thanks for listening as always bye for now