AWS Mid-Year Leadership Summit 2025 | Sarbjeet Johal, StackPane | AWS Summit NYC 2025

Clips
News
More from AWS Mid-Year Leadership Summit 2025

Inside AWS’ halftime playbook: 25 ways agentic AI is redefining enterprise innovation

Agentic workflows are no longer a future promise — they’re the new foundation of enterprise artificial intelligence. As autonomous systems shift from generating content to taking action, organizations are redesigning how they build, operate and scale.This pivot is reshaping everything from infrastructure and developer tooling to team structure and business outcomes. Companies that once experimented with generative AI are now racing to deploy full-stack agentic systems capable of driving decisions and executing tasks. In a candid conversation with theCUBE’s John Furrier (pictured, left), Matt Garman (right), chief executive officer of Amazon Web Services Inc

play_circle_outline Exploring Low-Code Trends and Amazon's Latest Innovations: Implications for Developer Communication and Multi-Agent Workflows

play_circle_outline Exploring Amazon Nova Customization, AgentCore Announcements, and KiiRo IDE for Spec-Driven Team Development

play_circle_outline Amazon Nova offers customization capabilities for agent development in enterprise settings.

play_circle_outline Caution about Amazon's messaging in the rapidly changing tech landscape.

Info
Transcript

Sarbjeet Johal, StackPane | AWS Summit NYC 2025

In this exclusive conversation from AWS Summit NYC, theCUBE’s Dave Vellante welcomes Sarbjeet Johal, founder and principal at StackPane, for an in-depth analysis of Amazon's latest AI and infrastructure announcements. Described as a “halftime report from re:Invent,” this candid discussion goes beyond the product sheet, spotlighting the real implications of AWS’s AgentCore, KiiRo IDE and Amazon Nova for enterprise and startup builders alike.

Johal shares field-level insights from analyst briefings, offering critical perspectives on how AWS’s developer... Read more

explore Keep Exploring

What are the speaker's views on the clarity of Amazon's direction in the industry compared to Salesforce's? add

What is KiiRo and how does it relate to integrated development environments (IDEs) and companies like Google and Amazon? add

What are the details regarding the development and customization of Nova and its internal code name, Nova Titan? add

What advice is given regarding messaging for AWS in relation to emerging trends and new entrants in the tech industry? add

bolt Powered by CUBE AI

Sarbjeet Johal, StackPane | AWS Summit NYC 2025

search

>> Hi, everybody. Welcome back to New York City. My name is Dave Vellante. John Furrier's taking a quick break. He's been going wall-to-wall coverage here at the AWS Summit in New York City. You're watching theCUBE's coverage and I'm here with Sarbjeet Johal, who's the founder and principal at StackPane. Reporter's notebook, we always call you, right? You're out in the field gathering the notes. We've been inside the analyst event all week. So basically, this used to be a one-day event and now they've extended it to three days. It starts Monday night. We had the reception Monday evening. They have deep dives with the analysts all day. They have one-on-ones today during the AWS Summit, which is a one-day summit. Actually it's one day, but they tacked on, I believe an AI day. Partners are here. A lot of local folks, a number of sponsors and partners as well. And Sarbjeet, I was on vacation so I missed it, they gave us a preview of the announcements. They didn't do what they normally do, which is they brief us, but then they brief us deeply in the day before, and then that day we get the Andy Jassy keynote or the Matt Garmin keynote all day, but we're really prepped for it. It was a little bits and pieces here. So there's a lot of announcements that we're going to talk about. I think the ones that stand out that Swami noted, Amazon Nova customization-

>> AgentCore is the top.

>> AgentCore I think is absolutely, and then all the components there that we can go through. The Amazon S3 vectors was kind of-

>> Interesting,...

>> had previously been announced. KiiRo is another big one, so we're going to go through those. But I'll just set it up, Sarbjeet, by saying the following, and this is kind of like the halftime report from re:Invent in a way, halfway through the year, a little more than halfway. Amazon does a tremendous job. Let me say it this way, almost 20 years ago, Amazon turned the data center into an API. So they do a tremendous job of speaking in the developer language and talking to developers, and putting forth capabilities through APIs and services that developers can access and deploy. And they make it really easy for developers to do that. And they're on a mission constantly to simplify the developer's lives. I would say they do a very poor job of laying out the vision for the C-suite, as to what the future of their business can look like. They don't really speak that language, they leave that for the interpretations to the customer. So they work backwards from the customer. That customer is typically the developer or somebody who's technical and they really focus like a laser on solving those problems. Question here, I think they also have a vision, but they don't share that vision openly with the community, with analysts. I don't know if you feel the same way.

>> Somewhat. Not that stark. I don't see that as a stark weakness, but you're right. When you are saying that we always work backwards, that means that implies that customer knows what they're looking for and then we work backwards to fill their needs. But you also have to look forward as a vendor where we are going. So I think as you said, I think rightly so, that they don't paint a cohesive picture of where we are headed and which persona is fit for which of their services. I think that's a go-to-market-strategy sort of narrative. They're grappling with it, just like other vendors are. Another observation made is that other vendors have made their life harder, because the likes of Facebook are jumping in with their own models and trying to entice the developers, even enterprise developers to use their models now. So now these poor guys at AWS, they're forced to talk about why coding, which they don't want to talk about, but they have to, right? So it's convoluting their message. Actually, in one-on-ones, I have said that repeatedly that, "Hey, please do not mention white coding when you are mentioning KiiRo, for example. If you do mention that, have a footnote. You can also do some white coding, because pro-developers hate that term, right?"

>> Citizen developer.

>> Yes. Yeah, it is.

>> It's kind of an insult to them.

>> It's an insult to them. I mean, because I've been to a pro-developer and then I don't like that fact, because okay, you can cook up a utility, a software utility, or a tool by working alone as a single developer, or maybe two or three developers. But when most of the software applications we use at enterprises, they're cooked up by tens and maybe sometimes hundreds of developers. And for that, you need specs, so people can go to the spec. If a developer is hit by a bus, I mean, the way the other folks are showing, if a developer is hit by bus, your application is gone. No, that's not how it works, people. Most of the software, what we use for mission-critical application is cooked up by teams. There are specs behind it. Sometimes we deviate from specs, and then when we deviate from specs, we update the specs most of the time. There are flies here. Everybody's here, guys.

>> This is New York. You're in New York City.

>> It's a spy from Microsoft. So I think they're forced to say certain things, which they don't want to. So that makes the water, it makes the water sort of muddy for them too.

>> I say it this way too. I mean, Swami definitely lays out, he tells a great story, but his keynote today was, "Hey, this is what we're doing to make your life easier in terms of building agents." Which is very important. I'm not trying to trivialize that, but I don't get a sense as to where the industry and the economy is going, where business is going from Amazon. I get that from Salesforce. Maybe Amazon purposefully avoids that, because they don't want to be hype masters like so many in our industry. So they should get some credit for that. But I would like to see a little look. In fairness to Swami, he did a really good job of saying everything has changed in software. And the one term he did use, which I like, I thought George Gilbert coined, but maybe Swami coined it and George used it, but maybe not, is services as software. You know software as service. He talked about services as software. He didn't really describe in detail what he meant by that. But I think what he means is, "Hey, we had all these traditional software applications before the cloud, like ERP and HR, CRM, and then the cloud enabled those to become SaaS, software as service. Now it's like we can generate all these services, all these new services powered by AI in software, so services as software." That's a flip of the switch. Swami did a great job of laying that out. Again, in my view, speaking the language of developers, and that's all fine. It's just an observation, not trying not to make a criticism there, but we want to go through some of the announcements. I'll just say this, a lot of announcements on bringing together operational and some of it's NDA, so I can't go too deep on what exactly is happening. But the direction that we saw at re:Invent, I'll say it that way. Bringing together operational and technical metadata, using SageMaker, whether it's SageMaker Studio or SageMaker Catalog, to really unify Glue and DataZone, we can clearly see the direction that Amazon is headed. Using agentic building capabilities that they've announced today. We're going to talk about Amazon Bedrock AgentCore to simplify the building of agents. Being able to govern that, making sure it's secure, that identity is taken care of, that that's federated.

>> The memory part.

>> The memory, we're going to talk about that. What's missing in my opinion, we heard this yesterday in the analyst meeting. We heard one of the developers talk a day in the life of the developer or the decade in the life of a developer and how her life has changed, which was fascinating. But what's missing in all this is that a really important system of intelligence layer, I'll call it, which harmonizes data, which has a 4D map of the enterprise, which speaks the language of people, because generative AI is the human language. Databases speak in strings and ones and zeros. Humans speak in people, places, things and processes. So a knowledge graph speaks in those terms. And so, that's missing here. The harmonization layer, the knowledge graph, what George Gilbert coined is the system of intelligence. Well, he talks about, maybe he didn't coin. Give somebody else credit, but I'll give him credit, because he sort of introduced me to it. But a system of intelligence layer that connects in the BI layer, the metrics and dimensions and can interpret the business context from that, that has the metadata, the technical metadata, the operational metadata, unifies that along with the process data and then serves it up in a governed fashion to agents. So that middle piece is missing. Now when I push Amazon on it, they clearly understand that. So maybe we'll see something at re:Invent. I am sure they're working on it. They're smarter than most of us, but that was missing. So my point is, a lot of tools to help build agents and make life simpler solve some of the complexities of working in Amazon. But the future requires that layer that doesn't yet exist.

>> So, the fabric, yes.

>> So your thoughts, and then let's get into some of the announcements. In particular, like you said, the top one here is Amazon Bedrock AgentCore.

>> Yeah, yeah.

>> So what is AgentCore? Why is it important?

>> All right, let me address what you just said first. So I think having a map of your enterprise process behavior and people, places, things, you usually talk about with Gilbert, right? Most of the time. It's easier said than done, because every company behaves differently. Every company calls certain things differently. Purchase or a P.O. they name it differently. So there's a language aspect to that. And language is fluid. Language is very fluid. I mean, most of the time we use English terms in the software world, but you go to China or other countries, they call those things differently. And as the software leaves English-only world, with the help of gen AI, the things will become even more complex.

>> That's what I'm trying to say.

>> So long story short, it is a dream goal of us to have that fluidity of, okay, the BI is the add-on. It was always an add-on. Like, okay, you have transactional systems and BI will do the business intelligence on top. It was like-

>> There you go.

>> Yeah, it's separate, right? Separate database and data bearers, and all this stuff. Now we are trying to merge it. I think just the merging part is so complex. Forget about having that layer on top of that. So that will take maybe 10 plus years to get there. Maybe we'll be striving for it forever, but that's that. But talking about what AWS is doing now, even the AgentCore, what they have done is that they have brought in sort of the pro, well, not only pro code, but the pro applications, like professional application development constructs under one umbrella. So they talk about memory, they talk about identity, they talk about observability and also the gateway. I mean, gateway, the API gateway was very key glue piece actually for AWS, which made all other services fly. Everything goes through the gateway. So they are doing that for the agents now, and they have the runtime. And I think the idea behind runtime is that you can manage state in a more cohesive manner, because not everything is async. And I have been saying this for a while now, for two years or so, after the gen AI stuff came in, that even in this agentic age or AI-based application age, we need a facility like enterprise services bus kind of arrangement, which is missing, because everything is new. So they're missing pieces and they're trying to fill in those gaps for professional applications. I think they're doing a good job of that. I like the KiiRo, they get their native IDE.

>> Before you go to KiiRo, I just want to frame AgentCore. It's six new services. They're managed services, such that any developer can build an agent with open source frameworks if they want to. And then, once that's built, you can deploy it in a secure environment. They take care of all that. And then there's AgentCore runtime, Amazon make sure there's no data leaks. So that's the suite of services as you said. And I'll just give the quick rundown if I can. There's the AgentCore runtime, AgentCore memory, AgentCore identity, AgentCore code interpreter, AgentCore browser tool, AgentCore gateway, and AgentCore observability. So again, this is like here you go. Here's all these APIs that you can invoke. AgentCore runtime, it's a secure serverless runtime. AgentCore memory, they have short and long-term memory. AgentCore identity, you can integrate with whether Okta or Intra, or whatever. The interpreter is a secure sandbox. So for code and JavaScript, or TypeScript or Python, I mean, on and on and on. Then at the end of it all is AgentCore observability, because agents are going to do stuff and there's going to be problems. So you need to have the telemetry and the observability to visualize where those problems are. And that's a set of six services that they basically say, "Here you go, developers. Build." And that's the Amazon way.

>> Yeah, okay. Amazon is a startup-friendly company and they are ISV-friendly company. Some of the ISVs are the biggest companies or their SAP is an ISV. That's how I see them. So they will be using all these basic constructs. All these companies like Intuit is one of the big ISVs as well. So they'll be using all these primitives to build applications. I think where Amazon lacks that connection is the enterprise developer, which is developing applications for in-house use. They don't sell software or they don't sell SaaS, for example. They build it for their own use. And that's where of Microsoft actually they would do a good job, but they do a poor job, those guys do a poor job with ISVs and pro cellular developer side. I think. That's how I see it.

>> So the other big announcement is-

>> Google is in between....

>> you mentioned KiiRo, it's an agentic integrated development environment that integrates natural language specifications for a lot of code focus, code gen, debugging, testing. That's what KiiRo is. What are your thoughts on that?

>> The main thing is a spec. It's a spec. KiiRo is their native IDE now. So it will hook up into their core sources, which is good. It's based on VS Code. Everybody is using it, including Google use VS Code behind the scenes to cook up their IDs. To me, the ID is what browser is to the masses. The ID is to the developers. So if you don't have IDE, which is integrated development or what we call it as developers. So if you don't have that, you are just relying on JetBeans or Eclipse, or some VS Code or Microsoft's IDE, like Microsoft had their own local native IDE. Google just got a couple of years back and they're maturing it. Finally, I think Amazon got it. I wanted Amazon to acquire Cursor, but they're already building this similar stuff. But I think the very refreshing thing is that it's a spec-driven development. It's a team-driven development. So that's pro code, pro-application. Bigger applications. Complex applications. What we usually cook up all these enterprises. So I think it is a great thing, but it's a first step. They're just dipping their feet into this. And I mean, in the beginning, actually, I got some color into that. When you launch KiiRo, it tells you that, "Do you want to do white coding or do you want to do spec-driven coding," right? And it gives you that fork in the beginning. I said like, "Can you change your mind later after you do some development?" And yes you can, they told me. And it's a work in progress, but it's a very sort of forward-looking from the industry point of view, not only the AWS point of view, industry point of view. Hey, you have specs, you give me the specs and when I do hit the code completion, or give me some code snippets or go code this, it is looking at your specs. And then, when the code changed, you can tell it to update the specs backward, which is very hard thing for developers to do. Developers don't want to write specs. I mean, they hate that. Developers don't like to macro the database as much. They don't like identity part. So all these things are getting taken care of.

>> All what they call the heavy lifting.

>> Yeah, heavy lifting.

>> What about this Amazon Nova customization? They made a big deal out of that and they brought out Rohit Prasad, who's the senior vice president of Amazon. He's an AGI.

>> His thought process is pretty good, actually. I like him.

>> Yeah, but I'm not sure why they use the AGI term, but any rate, they're not going after Open AI. But I kind of like, I wish they said with enterprise AGI.

>> Coming back to that-

>> They don't want to-...

>> they're forced to use AGI. They're forced to use white coding, even though they're not those people.

>> Right. And so, he actually laid out in his talk a number of items that they had developed throughout the Nova lifecycle. So Nova Titan was the internal code name and they launched Nova at, they launched at a re:Invent I believe. But at any rate, he laid out his roadmap, don't know what are your thoughts on that capability? They claim that they were the most highly customized capability for agents.

>> Okay, let me tell you what that is. I got some color into that. Talking to the product-

>> Yeah, go ahead....

>> managers who are developing this stuff and what's in their mind, and they have blind spots. When you are working at a company like Amazon, you are not keeping an eye on IBM or you can't really study all the comparative, you can't do that. That's our job actually. That's what we do, right? So I mean, IBM, sorry, if you're using their instructor lab for training their models, right? Native model, right? Granites. I mean, that's light model. And these guys are telling me that our model's heavier. So it's harder to infuse your own data into it. Number one is that, number two is that your data stays separate. So Nova sort of ensures you that, that's a good thing. It's a bigger model and you are infusing the data to bigger model, fine-tuning, bigger model is harder. So that's another plus thing. S3 pricing goes with it. If you are infusing your data, S3 is cheaper than traditional vector databases. So that goes with it. So they're trying to solve that problem of infusing your own data into the models, either using any sort of modes, like fine-tuning or RAG, or distilling the model. So they're tackling the problem from all angles. But having said that, it's going to be a table stake. It is table stakes going forward from Google. Google will do the same thing. Microsoft will do the same thing, but they're not ahead in the industry in that sense, but they have some unique advantages, which IBM doesn't have because of scale. But IBM started earlier. So they put that sort of-

>> And IBM's strategy is to really be kind of horizontal integration.

>> . Yeah.

>> Yeah. We're up against the clock, but a couple of the other announcements that you can pick whatever you want. Amazon Nova Act enables agents to take action and web browsers with complex UI interaction capabilities. Amazon S3 Vector is a new capability to store and query vector embeddings in S3, with built-in support that's optimized for scale and low TCO. I'm sure John has been talking about that a little bit. Amazon Q for developers is being enhanced. This is with GitLab integration, enhanced MCP support. Okay, so you'll see that. AWS Transform, they've been talking about that for a while. So the code transformation. And the other piece is the capability to interact with, they're leading or at least participating in the standards of MCP and A2A. That's part of, I guess strands for agents 1.0, right? Support for A2A protocol data, persistence, multi-agent workflows. So a lot of typical AWS announcements. I'll give you the final thoughts on all this.

>> I think final thought is that we just watched a movie last week with the family, the Formula One movie, Brad Pitt comes in as an old guy and a young-

>> F1?

>> Yeah, and a young guy. So the combination of young guy and old guy. So there's a dialogue said a few times repeatedly. Slow is steady and steady is fast. So the young guy was going fast and not winning the race. But then you need to know the maneuverability, right? You need to know the basics of the stuff, right? The fundamentals, the primitives. If you don't know that, you can be fast, but you can't win, right? So I mean, that's what they are. Amazon is slow is fast, fast is... Slow is steady, steady is fast. So they won't sort shake the ocean or boil the ocean if you will. But they are moving in the right direction for the people who build stuff, systems at scale on one end and the people who are building that B2B applications from scratch. So they're very startup friendly on the one end and very enterprise friendly and ISP friendly on the other. In between that gap there, like local developer, inside developer where we talked about earlier, I think that's where they need to focus a little more on. But I don't blame them. They can't focus everywhere. And one only word of caution to AWS folks is that hey, yes, white coding is being talked about by all these new entrants and these tourist investors, and all that stuff. And tourist tech companies, including Facebook is trying to get into this. Don't get confused with your messaging. Focus on the developers. You can talk about low code, but don't start using all the new terms and all that stuff.

>> Well, they're using them.

>> Be careful.

>> All right, Sarbjeet, thanks so much. Appreciate you coming back on.

>> Thank you very much.

>> All right, keep it right there, everybody. We'll be back right through the short break. This is Dave Vellante for John Furrier and the entire CUBE team. We're here at NYC Summit 2025, the AWS Summit. Be right back right after this short break.