🦄 Making great presentations more accessible.
This project aims to enhances multilingual accessibility and discoverability while maintaining the integrity of original content. Detailed transcriptions and keyframes preserve the nuances and technical insights that make each session compelling.
Overview
📖 AWS re:Invent 2025 - Migrate, modernize, and move your business into the AI era (INV212)
In this video, Dr. Asa Kalavade introduces AWS Transform, an agentic AI service for migration and modernization launched in May 2024. The session covers three critical workloads: VMware migrations, IBM z/OS mainframe modernization, and .NET Windows applications. BMW Group shares their mainframe transformation journey, achieving 75% time savings in test case creation and migrating seven applications in six months. New announcements include full-stack Windows modernization with SQL Server to Amazon Aurora PostgreSQL migration, AWS Transform Custom for any-to-any code transformations with six out-of-the-box transforms, and composability capabilities enabling partners like Accenture, Cognizant, and AWS ProServe to create personalized workflows. Since launch, customers have processed over one billion lines of COBOL code, saving an estimated 800,000 hours of manual effort.
; This article is entirely auto-generated while preserving the original presentation content as much as possible. Please note that there may be typos or inaccuracies.
Main Part
From Paper Maps to GPS: Reimagining IT Transformation with AWS Transform
You wouldn't drive across the country using just a paper map. But that's how most enterprises approach their IT transformation. No visibility for the road ahead. Legacy systems forcing detours, manual processes compounding delays, innovations stalling in gridlock. $2.41 trillion in US technical debt. 70% of workloads still on premises, 70% of legacy IT software for Fortune 500 companies written over 20 years ago. Until you've spent more on friction than the finish line ultimately returns. What if you could see the entire route before you start? AWS Transform reimagines your entire transformation journey with AI-powered navigation for modernizing legacy systems. What was once impossible, now inevitable. Fortune 500 and global enterprises operating at the speed of innovation. Discover the science, the proof. Your transformation journey starts here. Please welcome to the stage Vice President of AWS Transform at AWS, Dr. Asa Kalavade.
Who remembers carrying paper maps and printing directions to plan travel? Today, we don't think twice about using a GPS-enabled phone that gives you turn-by-turn, real-time directions, reroutes you if needed, and even gives you an alert depending on where you need to be and when. That's our journey for migration and modernization. I'm Asa Kalavade. For the last 10 years, I have obsessed about helping our AWS customers migrate and modernize in the cloud.
I started building hybrid storage services that connected customers' on-premises environments to the cloud, helping them start their migrations. I have built data transfer services that allow customers to modernize their data workflows, whether it's gene sequencing, media editing, supply chain, and so much more. Today I lead AWS Transform. Much like GPS has evolved, our tools for migration and modernization are evolving. We started with custom scripts, point-in-time tools, a lot of manual work, sending information back and forth.
Our vision is for these tools to become just like GPS, automatically navigating you based on your destination. We envision a world where there are task agents that are experts, part of your IT team, sitting alongside your global systems integrators and partners. They understand your business, they plan the migration based on your inputs and your business objectives. They execute using your cues. They're continuously learning from their own trajectories, from your input, and from the environment and the enterprise context itself.
Introducing AWS Transform: The First Agentic AI Service for Migration and Modernization
It is with this vision that we launched AWS Transform in May of this year. AWS Transform is the first agentic AI service for migration and modernization. We launched with three critical workloads, and these were determined based on what our customers told us were the most important yet most risky workloads to modernize. First, we started with migrations. Customers want to move to the cloud, save costs, take advantage of the elasticity, but these migrations are hard. You need to understand on-premises tech and cloud tech. So we started with modernizing and migrating VMware workloads.
Second, mainframe. As we all know, mainframes have been around for the last 50 years. They run highly complex business applications, but customers have an urgency to move these to the cloud. First, they want to get ready for AI, but also the expertise in mainframes is declining. So we started with enabling modernization of IBM z/OS workloads. Third, Windows. Customers want to modernize their Windows applications to get better security and performance, but these are monoliths, and modernizing them requires understanding these complex dependencies. So we started with modernizing .NET.
From framework to core so you can go to more open platforms. Now, it's not just the systems that we're migrating that are complex. Migrations and modernizations are complex because of the number of people involved. It includes stakeholders across organizations and multiple experts. It includes lots of documents being passed around and lots of calls, and it is this complexity that we want to simplify.
Here's how we do it in Transform. Transform is a special purpose-built web application. Within Transform, you can be running multiple jobs at any given time across multiple workloads. Within each job, there could be multiple tasks. We make collaboration really easy by allowing multiple users to work within the system, adhering to your authentication policies. We have a single artifacts store, so you can bring all your documents as part of the migration process in one place.
These documents then get archived and versioned so they become a digital footprint of your IT infrastructure as it goes through the transformation process. In addition, we have multiple task agents that can create a job plan and execute the job plan with the human in the loop providing the input as needed, all through chat. Now, to make this really simple experience possible, we focused on three fundamentals.
Under the Hood: Expert Task Agents, Goal-Based Orchestration, and Continuous Learning
First, we have dozens of expert task agents. These expert agents range from understanding your COBOL code, your .NET code, your VMware TCO, and everything in between. These agents have been trained based on our learnings from hundreds of thousands of migrations over the years. They also can understand the business context and your enterprise context. Second, these agents get orchestrated based on goals. So based on your specific goal for your specific workload, a job plan is created that could be very personalized. These agents are then orchestrated through this job plan. Not only does this give you automation, but it allows each migration to be highly personalized.
Third, these agents are learning. They're learning from trajectories of their own runs, they learn across runs, and because of this learning, we can now recommend the next best steps. Let's look one level deeper under the hood. Here's the AWS Transform stack. At the topmost layer, you can interact with multiple surfaces. We showed the web interface earlier, and you can also use an IDE. Under this are the three domain-specific purpose-built workflows: VMware, mainframe, and .NET.
Now, within each of these, obviously there are lots of task agents that get invoked, but all of them share a common set of fundamentals. These include your artifact store, your knowledge bases, the chat agents, and most importantly, the job orchestrator. It's the job orchestrator that understands dependencies, routes tasks, and also captures a lot of the work logs. Because at the end of the day, as you can see, Transform is multi-agent, multitask, multi-user collaborative. So that's very complex to operate. We capture all of the work logs and the operational pieces that allow us to run Transform.
Transform also fits within your enterprise context, so it can connect to your code repositories, it can connect with coding agents, your identity providers, so it's really meant to be an extension of an enterprise's IT environment. Since launch, hundreds of customers have used Transform. They have processed over a billion lines of COBOL code as they're modernizing mainframes, and we estimate to have saved over 800,000 hours of manual effort. And we're just getting started.
VMware Migration Made Simple: From Discovery to Deployment with Five New Task Agents
Now, let's look at each of these three workloads in some more detail. Let's say you have a large data center exit coming up. The first agent you would turn to would be the VMware agent. Now, as many of you know, these VMware migrations tend to be very complex. Typically, you have an IT admin who's collecting all your on-premises assets to even know what you have on-premises before you can move it to the cloud. And then you have a combination of your tech program managers, your product managers, and even CFOs coming up with a business case and planning the migration.
And then your experts, whether networking, security, migration, all of them have to come in and do the actual migration. Now, you can imagine this is what makes those projects so complex, and they take anywhere from months to years to complete. So how does this work in Transform? We have introduced five new task agents within Transform to make this process much simpler. So first, we've expanded the discovery.
Now you can upload all kinds of formats into Transform, and it takes care of all the deduplication and creates a canonical representation of your entire assets. The next step is the actual TCO or business case analysis. We've introduced more ways in which we come up with recommendations for the right target architecture. We have a real-time connection to the AWS price catalog so you can understand the costs, and now you can have a conversation with this. You can say, "What if I moved from this instance to the other instance or I change regions?" so you can do all of this in minutes. This is what used to previously take weeks or months, and that makes it hard to even start a migration.
Let's say we've done all of this and we're ready to move forward. The next step is planning your migration waves. At GA, we introduced the capability to do this through Transform, but then our customers told us they want to do more of these migration planning. So we have now introduced a new ability to do natural language-based migration planning. This is not only looking at your dependencies and data flows, but now you can give it instructions such as migrate my test and prod servers in different ways or do my HR application first and then the tax application. Or even say, put all the servers with low latency requirements into one wave. So now it's looking at the dependencies and making these plans in a very conversational experience.
Now with the planning, you set these agents to work. You have your network and security agents going into the work. One of the features that we introduced in Transform back in May was automated network generation. When you're moving from on-premises to the cloud, you need to generate the network topologies, and this saves a lot of time and effort. Our customers love this, but they told us they have many more complex environments. Can you help us with that as well? So now we've expanded the ability to do isolated VPCs, hub and spoke networks, many different IP strategies, mapping into multiple accounts. This has really sped up the whole network generation process.
Along with security groups, we also brought in AWS Application Migration Service, or MGN as many of you might remember, into Transform, so now you can do your rehost all within the same application. Another feature we just introduced was the ability to create a support ticket. Let's say you finished your migration process and you have the Countdown premium package. You want your TAMs and support engineers to help you with cutover. All you do now is click a button and open a support ticket. Your entire context is passed over, and the support engineer now becomes part of this extended collaboration team.
Let me pull it together with a specific customer example. We've been working with CSL, a global biotech company, for the last few months. Over the years they have accumulated a number of data centers because of acquisitions, and they had over 5,000 servers and 1,000 plus applications. They had decided to make a move to the cloud because they wanted better scale and capacity. They started doing the migration and it took them a couple of hours to understand each application. Remember, there are thousands of them, and within one month they had still only processed about a couple hundred servers to plan their migration.
So we then worked with them along with AWS ProServe and Transform, and they got a 10x improvement in the time it took to discover applications and a 12x improvement in migration planning. Now they're well on their way to migrate 17 data centers in just 2 years, and they estimate to see a 30 percent savings in total cost. All right, so now we've migrated those VMs, but we are not quite yet ready to exit the data center because we have that one big mainframe still waiting. So let's see how you can use Transform within the same experience to modernize your mainframe.
Mainframe Modernization Reimagined: Six New Agents for COBOL Code Analysis and Testing
Now if you look at a typical mainframe modernization process, here's what usually happens. There are dozens of engineers sitting and trying to understand that COBOL code because remember, many of the original authors are not with the company anymore. You're trying to understand the code, the dependencies, the databases, and all of that. Once you've done that, only then can you start understanding which pieces you want to start modernizing. This is the piece that makes these mainframe projects go on for 3 to 5 years.
Now with Transform we are actually reimagining how to do these projects. So with this you upload all your data, you get all the analysis, but more importantly you can bring in your business context.
Code documentation, any other tutorials that you might want to bring in, and along with that, any context from the systems integrators. With all of this that flows through the system, you can plan your migration. Connected with coding agents and ready, you're ready to go. To make this happen, we introduced six new agents in the mainframe application.
The first step is to upload all your data and see the analysis. This shows a lot of information, but there's also too much information. We've now introduced a natural language-based analysis on the analyzed code. With a few questions, you can ask questions like what are my batch processes, what are my transactional processes, and tell me more about what's in each of these applications.
Next, we've introduced a new activity analysis agent. You can then start understanding the activity of these different components of your mainframe application. For example, if you decided to do batch processes, you can ask what are the most complex batch processes, what takes the longest time, what is running most frequently, and what's not being used at all, which is obviously a natural candidate to not modernize. You can do all of this with natural language because we bring in all the analysis through knowledge bases that are easily accessible through chat.
Data analysis is another agent that we've introduced. You can now go and ask questions about the data lineage, your DB2 tables, VSAM files, and everything that interacts between the application and data structures. As you can see, we're building up more and more information about this application, including documentation and business logic all in one place. Now you're ready to migrate and change and modernize that application, but there's one more. We've introduced test agents, so as you all know, testing is about half the time it takes for any project. We started building agents to understand your system, create test plans, create test data sets, and compare the two. We're excited about what we've done, but to tell you more about this, I'd like to introduce a customer.
BMW's Journey: How AI Became Both Driver and Enabler for Mainframe Transformation
A customer that doesn't need much introduction. Let's see this video. For over a century, BMW has engineered the ultimate driving machine, creating the world-renowned feeling of sheer driving pleasure. With the BMW iX3, the company reimagined that experience for a new era, more intelligent, more dynamic, and more BMW than ever before. Behind every vehicle stands BMW's global production network. The BMW iFactory. This manufacturing concept embodies efficiency, sustainability, and digitalization, paving the way for future-proof vehicle production. Building on a modern and powerful IT landscape, BMW continues to shape the future of mobility, translating its pioneering spirit into the digital realm. By harnessing the power of AI automation, and cloud technologies, BMW is transforming its mainframe into a modern, scalable digital ecosystem, setting new standards for performance, efficiency, and innovation in the automotive industry, redefining the art of the possible together with AWS.
Please welcome to the stage Director of Customer Solutions and Migrations at AWS Shrirat Nasinga, and Vice President of Transformation Programs at BMW Group, Frank Uslab. Frank, welcome. Nice to be here. Thank you. BMW and AWS have been co-innovating on this journey to a modern IT landscape for some time. Frank, I'd love to hear a bit about your role, your mission, and this journey that you've been on.
Yeah, thanks a lot. Digitalization is one of our focus topics and of course a key enabler to produce such great and innovative cars. We have been working together with AWS for more than ten years on innovative topics like the Alexa assistant or our BMW cloud data hub. We are now in the middle of a group-wide cloud transformation, which is the primary mission of my organization. For us, it's not just an IT upgrade, but it's really a strategic decision to accelerate innovation.
To increase flexibility, and to unlock the full potential of data and AI. That's a fascinating vision to unlock that full potential of data and AI across the company. Now BMW has a storied history over 100 years old. How does that impact your IT landscape?
We are quite proud of a very modern IT landscape supporting core processes like engineering, production, and customer experience with AI. We have more than 600 productive AI use cases running in our business. However, like any other big company, we also have some legacy systems. These systems have typically been in use for a longer time with no significant need for change, so you cannot migrate them easily to the cloud. So far we have kept them on premise.
That's an interesting challenge. You have this mix of core processes infused with AI with 600 use cases, which is amazing, and you have your legacy systems. So what changed, Frank, that made transforming the migration one of your strategic focus topics?
When we started to explore agentic AI use cases in our processes, we saw great opportunity for embedding more intelligence in some of these processes, especially in areas where we also have some legacy systems. For us there is huge potential in automated end-to-end workflows and smarter decisions. However, there is no chance of implementing that in a mainframe environment. So far it was also one of the biggest challenges in IT to actually migrate away from a mainframe system, but here AWS is a big game changer for us. AI became for us both a driver and also an enabler.
That's a fascinating shift to think about AI both becoming a driver for that need to modernize so you can build agentic workflows and also enabling and accelerating that transformation. Now, what were the biggest challenges with getting off the mainframe and how did you sequence your approach with us?
Breaking Down Monolithic Architecture: BMW's Migration Factory Approach
I would say it's the typical challenges that come with monolithic architecture. You have a lot of interdependencies on the mainframe, which means it's really difficult to isolate single applications and also to migrate single applications without causing a ripple effect. On the other hand, there is a lack of proper documentation or even test cases. Some of these systems evolved over a long time with no or just minimal or quite outdated documentation.
Where we started was by first analyzing our applications and then defining our migration patterns. Some of the applications we could quite easily migrate and we found patterns to replace them with existing template solutions. For the remaining applications, we first started with analyzing further the dependencies and then started with decoupling the data. That gave us already a good indication of where to start and how to break things into smaller pieces with fewer dependencies.
What's really interesting about that is this dependency mapping and analysis is often the most complex aspect of migrating off the mainframe, and you found a way to make it better. Now, this co-innovation partnership with Transform has been six months in the making. Can you share with our audience what are some of the specific capabilities you focused on with us?
We saw first of all huge potential in understanding and visualizing these kinds of dependencies and creating documentation based on the actual code, which was not there yet. Of course, everything around testing is something that we have huge potential in. It started from test case generation based on extracting business rules to test data creation or even automated testing. That's what we focused on before we also looked at some of the code improvements.
As a next step, with research capabilities to optimize our coding and make it easier to maintain in the future for software engineers who have not been in touch with any legacy applications, we see big progress on that. What I hear from many customers is that getting the code not only moved but in line with their enterprise standards and Java standards, making it more readable, is very critical for success.
So how did you implement these migrations with AWS Transform within your organization and your team? Our ambition is to do these migrations with the same quality, precision, and efficiency as building cars. We built something we called our migration factory, which is a team of experts with the right mindset and skills supported and enabled by the AI capabilities of AWS Transform. They do the migration together with the product teams, which is a huge advantage since we can take the load away from the product teams and they can really concentrate on creating value for our business.
On the other hand, we have our factory team which can learn with every migration and can provide valuable feedback for further improvements of AWS Transform. While we still need an engineer in the loop at the moment for these transformations, we clearly see that the manual effort is going down, and that is really amazing to see.
Now let's talk about results from the transformation. What are some of the outcomes that you are seeing? It is definitely all about speed and effort reduction. What took us ten days in the past for test case creation, even for smaller applications, can now be done within a couple of hours, including validation. We started with the ramp up of our factory just a couple of months ago with some smaller applications, and we now have the confidence after we implemented them that we can work on the bigger ones.
If we reflect on what you said, that is ten days for test case creation to hours, which represents over seventy-five percent time and efficiency savings. So Frank, what else is different now with AWS Transform? I mentioned that we are much faster, but what has changed as well is that we have also reduced the risks. If you look at the opportunities from automated test case generation, that leads to another improvement: a much more increased test coverage.
In one of our cases, even by sixty percent. In the past when test cases have been created manually, some of the special variants have been missed, later causing issues in production. That cannot happen with automated creation based on business rules and extractions based on the actual source code. The fact that we are able to do this and that we are able to get that test coverage also helps us create much more trust on the business side so they believe that we can master these transformations. That is great risk reduction and coverage.
What else does this mean, Frank, for BMW and your teams in practical terms? While ramping up our factory, we have been able already to migrate seven applications, and we have been able to do that without adding additional resources into our product teams so they can really fully concentrate on creating business value. Seven applications in just six months is pretty unheard of in the mainframe world, and it sounds like you are also seeing that the application owners can really now focus on the business topics and not have to worry about spending cycles on the mainframe.
Now, how do you bring teams and developers on the ground with you along this journey? Being part of a team that solves one of the biggest challenges in IT, and I really think mainframe migration is one of these big challenges, it is something which actually motivates our team members a lot. Using AWS Transform to find smarter and much more automated ways for mainframe migration is a huge benefit so the teams can bring in their ideas and explore opportunities in a very much experimentation culture, which helps us also get the best results and creates a great spirit of true innovation.
Scaling with Manufacturing Precision: BMW's Vision for Companywide Transformation
So what is next in terms of scaling this migration factory companywide?
A culture of true innovation. We're currently extending our factory, which is supported by our DevOps team in South Africa with some more colleagues from our DevOps team in India, and we call it a manufacturing approach—setting up a second assembly line. What does it mean? It will help us to parallelize much more than we could do in the past and also to work on some very specific technical challenges and to further innovate to use the full potential of AI. Our ambition is to shorten our transformation timeline by at least 12 months, and I'm pretty sure we're going to make it. This is manufacturing precision meeting AI transformation, and it's a repeatable model enabled by AWS Transform across industries.
Thank you for sharing this incredible journey with us. I remember meeting you a few months ago and you had this grand vision that you shared with me. I am so impressed with how far along you've already come, and thank you for your partnership. Thank you very much for the collaboration.
Full-Stack Windows Modernization: From .NET to SQL Server and UI in One Framework
All right, so let's continue our migration process. We moved those VMs, we modernized the mainframe applications, and now we're in the cloud. We have a lot of Windows applications. Customers told us that they really liked the Windows .NET modernization capabilities that we had. They could get about 80% savings in effort in modernizing those Windows applications. But then they told us they want more. Any Windows application has databases, UI, and a lot of interdependencies. So today, I am very excited to introduce to you the latest expansion to AWS Transform: full-stack Windows modernization.
Within AWS Transform, now you can modernize not only your application code but also the SQL Server code, moving it to Amazon Aurora PostgreSQL. You can also modernize your UI from WebForms to Blazor. You can deploy the code and test it all within one framework. So we are very excited about this. Now, why is this so important? If you look at a Windows modernization project, you see lots of different stakeholders involved.
You have DBAs, software developers, IT admins, and DevOps. Now all of them work within AWS Transform. We can very easily do a full assessment of not just the application but the database and the intricacies between those. We then do a transformation of the schema, and we do this using our years of experience with AWS Database Migration Service. Having done tens of thousands of these migrations over the years, we've brought all of that knowledge into this agentic infrastructure. So you can modernize the schema, you can modernize the stored procedures, and you can do the application modernization at the same time. All the code invocations and the UI all get modernized. Then within the same framework, you can deploy the app to EC2 or ECS. The database can be modernized and migrated to Amazon Aurora PostgreSQL. We also generate test cases so you can validate that what you have transformed is what your source was, and this is all done within AWS Transform.
To show you a few quick examples from the UI here on the screen, you can see in one place you're doing data and code analysis. You can also develop your migration waves based on the analysis, the complexity, and the interdependencies.
And then you're doing your transformation, both schema and stored procedure code transformation again in one area. This is where after you finish the transformation, you can deploy it to the cloud. Now, what this gives you is the ability to do lots of repositories in parallel because you're applying them across the organization. But we've also heard that sometimes developers want to do a project at a time. So we've also made significant enhancements to the IDE experience.
Now within the IDE, you can get a full plan of your transformation process. You can edit the plan and change it based on your requirements. After the transformation is complete, if there is a need to involve the human in the loop to complete the transformation, we create a full markdown file that you can now take into your IDE and finish the transformation. We make it very easy to go back and forth between the web app and the IDE.
Here on the screen you can see the work log. Your application might be happening in the IDE, but you can still see the work log in the web application. Or if you're a center of excellence owner, you can log in there and see the status of all your modernization projects.
We've been working with Bobby Land at Teamfront for the last several months. Bobby leads the technology and business operations for Teamfront, which has a number of holding companies that do field service applications. These companies have come together through a number of acquisitions and have multiple coding patterns. They've been deployed in different places. They have old application code. They sometimes don't even know who wrote the code because it's been changed over the years. Bobby started working with us on Transform, first doing .NET modernization. He thought what would take six months was now being able to be completed in just two weeks. Having seen that, he's now working with us on modernizing some of his SQL servers to Aurora PostgreSQL. We're excited to see how Bobby can now take all of this savings and move that towards innovating and creating new SaaS applications.
AWS Transform Custom: Tackling Tech Debt with Any-to-Any Code Transformation
As we were talking with Bobby, he was saying that it's great to do all these large modernization projects, but he also has a lot of other stuff that he needs to take care of. That stuff means tech debt. Many customers like Bobby have told us this—they have problems where they're maintaining the right versions of their code. They're trying to move from Java 8 to 11 to 17. They have all these API upgrades that they need to do, SDK upgrades, and this part of the tech modernization or tech debt management process usually takes 30 percent of a typical organization's time and resources.
Organizations have choices. They could either keep managing this tech debt problem and reduce the investment they can do on modernization, or some of them say, let me go to modernization, but then this tech debt keeps building up. Yes, it does build up, which is why we call it tech debt. We've said customers shouldn't have to make that choice. How do we help them with tech debt modernization? Today, I am excited to tell you about the latest addition in Transform: AWS Transform Custom.
With this, you can now create custom transformations and execute them at scale. These transformations will take care of your software updates, API upgrades, and even modernizing your legacy applications. This any-to-any code transformation is an audacious idea, and here's how we do it. Through a very easy experience, you define what you want your modernization to be. You describe your input and the output application. You can give code snippets, you can give some documentation. You can even point it to your internal wikis because you might have been doing this manually.
Through this, Transform comes with an elegant description of your transform. You can edit it. You can now execute this either interactively through the IDE, through a CLI, or what's more interesting, you can even embed it into your existing applications to run at scale. As Transform keeps running, it's going to compare the validation that it put into the transform to make sure that the transformed code is what you wanted it to be. As it learns through these validations, through your input, its own self-debugging trajectories, it starts improving those transformed definitions and creates knowledge items that might be specific to your organization and your transformation. Through this you can now get your end-to-end tech modernization built up. It's pretty exciting.
What we did though, we also wanted to see how it works for specific out-of-the-box transforms. We have been using the custom transformation capability to define several transforms ourselves. Today we're introducing six out-of-the-box transforms ranging from Java, Python, Node.js upgrades, Lambda upgrades, SDK upgrades, and so on. I'm sure the team is building more while I'm talking over here. I always tell my kids, show not tell. To show you how AWS Transform Custom works, I'm excited to introduce Morgan Lunt, who is one of the product managers on Transform. Morgan, come here and show us a demo.
Show Not Tell: Live Demo of Custom Transformations from OpenGL to Metal
Hey, everyone, I'm really stoked to be here and to show you our brand new AWS Transform Custom. When I say brand new, I mean like two hours ago, I was sitting on my laptop launching this thing. So this is fresh for you all. Let's check out the demo video.
What you're seeing here is a terrain simulator that I wrote in C++ and OpenGL. I thought I'd give you something nicer to look at than just a command window. You'll notice I get around 50 FPS with fairly high CPU usage, so it runs okay. What I want to do is use AWS Transform Custom to convert this from using OpenGL to Metal, which is Apple's graphics API, to bump up those FPS numbers, increase performance a bit, but keep the same look of my terrain scene that reminds me of Seattle.
The first thing I'm going to do is open up the ATX CLI and tell the agent what I'm trying to do. In this case, I want to migrate my application from using OpenGL to Metal. It's going to search our internal registry to look for a transformation that's already built that does this. When it returns the list, you'll see we've got several hundred transformations already. People have found all sorts of uses for custom transformations internally at AWS. It didn't find one for OpenGL to Metal, so it's going to guide me through a simple process to create a new one. This is where the agent tries to extract from me every bit of information it needs to understand what I'm really trying to get at with this transformation and what exactly I want to do, including runtime language preferences and library preferences. It uses this to generate an initial transformation definition, which is an encapsulation of the agent's understanding of what I want to do. I can give it feedback on that, or I can just tell it to go run it. I can always tweak it later.
So it's transforming my code now, and as you can see in the bottom left corner, it's making incremental git commits so I can always go back and check partial work for what the agent is doing if I don't like what it did at the end. Here we go, my code is now transformed. The first thing I'll look at is the validation summary that we have here. The agent checks its own work, critiques its own work, and gives a summary of what it did, what it had difficulty with, what it thinks went well, and the areas you may want to double check. Digging into the code that it changed here, you can see that it kept my application logic. It changed my OpenGL code from using OpenGL to Metal, which is a pretty substantial change, and it even went and rewrote all of my shaders into Metal.
So what do you think? Do you think it worked? Hey, look at that. 120 FPS with some reduced CPU usage, buttery smooth. It's still the beautiful Pacific Northwest scene that I wanted to bring to Las Vegas here. I'm pretty happy with that. Now I know what you're saying. Okay, Morgan, cool. I see how this can help an individual developer, but I'm a program manager or a campaign manager. I've got to do dozens, hundreds, thousands of code-based modernizations. How does this help me? Check this out.
The AWS Transform web application has full knowledge of every transform in our CLI, both the ones that we made at AWS for you to use and ones that individual developers in your corporation created. A campaign manager comes here and says, hey, I want to upgrade all of my Python Lambdas, and ATX says, okay, cool. Here's a command. I've created a campaign. Pass this command around to your individual developers and have them run it. It's going to transform their code and log the progress here. Not only that, it writes the email for you, so it drafts it and everything, and you just say, hey, send it to the developers.
Back here in the cozy terminal, I'm an individual developer. I got that email, and I'm told, hey Morgan, you've got to go update all your Pythons. I'm like, okay, fine. So I take this command, I put in the name of my repo, I put in the local file path on my disk where my repo is stored, and I hit go, and it's transforming. As it's transforming, it's extracting knowledge items and learning about how my company's code looks and improving the quality of this transformation. But Morgan, I can hear you say my developers have to do this one at a time, and this is going to take forever. What the heck? If you're not, we built this thing with a totally deterministic machine-drivable syntax, so you can literally just write a little script to pull your code down for them wherever it's stored, transform it, and send it where it needs to go. Because it's a CLI, it's got really minimal dependencies. You can put it in a pipeline, you can put it in a container, put it in an AWS Batch job, put it on a laptop in your basement. I don't particularly care. You can run it however your code modernization infrastructure is set up today.
Back here in the AWS Transform web app, you can see the progress of your campaign. You can see every repo that's been transformed, its validation status, lines of code transformed, files transformed, and even an estimate of how much developer time you've saved. Because this is AWS Transform web UI, you've got the full agentic chat interface. So if you have any questions about the data, you can chat over the data and ask your questions. Thank you for your time. I'm super proud of what we've built here, and I cannot wait to see what you transform with it.
Thank you, Morgan. Now you know why I say show not tell. Customers have been using Transform's custom capabilities for a range of use cases. So we have Air Canada, one of the largest airline providers in Canada. They use the out of the box Lambda upgrades to help modernize tens of thousands of Lambdas, and they're seeing 90% efficiency. Another example on the other extreme is QAD, a software provider for manufacturing companies. They're in the process of modernizing their ERP application, and they wanted to make sure their customers' custom work was easy to modernize along with the ERP application. They worked with us to define this highly precise, specific transformation, and they were able to get more than 7,000 hours of saving through this, and there are so many more examples in between.
Composability: Building the World's Largest Transformation Network with Partner Expertise
We ourselves in AWS have been using the custom capability for many use cases. We're migrating our Lambdas, we are doing bash script upgrades, we're even using this to move from x86 to Graviton. There are so many more ideas here. I'm sure your brains are thinking about everything you could be doing with this. So we are excited and we are proud of where we've come with our developers and customers. As we were talking with our partners, our consulting partners, they said they like Transform, but they challenged us to think bigger. They said, "What would you do if there were no limits on where you could go? Where could you take Transform? What could you do with it?"
So today I'm excited to tell you about Transform's composability capability. Now with composability, what you can do is take all of the Transform agents, but you can compose your own personalized workflows. These could be based on your own agents, your IP knowledge bases that have been collected over years of doing these projects, or you can even bring an enterprise context. So now partners and consulting companies, AWS ProServe, they can create personalized versions of Transform for themselves. So to tell you a little bit more about what we mean by custom composability, let's see this video. Migration partners have spent years building powerful tools and proprietary expertise that may live outside modern workflows. Now, AWS Transform introduces composable transformations, where AI amplifies partner expertise to build and scale differentiated solutions in three simple steps with synergistic effect. Compose, publish, use. We're building the world's largest transformation network. 140,000 partners in over 200 countries and territories, creating differentiated solutions and making their specialized expertise universally accessible to customers. This is the art of the possible, where trusted partner knowledge meets AWS native intelligence delivered at scale.
We have been thinking about composability for the last few months, and these thought leaders have helped us define it for you. Would you please introduce yourselves to the audience? Hi, I'm Chris Wegmann. I'm the CTO for the Accenture AWS Business Group. I'm Ravi Kokker. I run the cloud business and financial services in Cognizant. Hi, my name is Darrell Hammett. I'm a global director for AWS ProServe with the AI platforms and also delivery. Thank you, the three of you. So we've been working on composability ideas for the last few months. Tell us how you see composability fit in and what you've been able to do with it so far. Maybe start with you, Chris.
Yeah, we were very excited when you reached out and started talking about composable agents. We started working with a financial services company who is transforming their mainframe platform to a more modern architecture. One of the big things we liked about the composable agents was the ability during the discovery process to actually pull in business knowledge. We were able to take our business knowledge of how a loan process works, take their business knowledge, build that into an agent so as discovery happened we could compose that with another agent to make that discovery even more impactful for the developers. It's awesome. Ravi, how about you? Well, we have come along with a very short span of time, so thanks for the opportunity and early involvement with the capability, and especially our ability to
ingest external feature sets, in financial services we have deep grounding on how we understand the deposits business. For FDIC insured deposits that you keep in the bank, we were able to create something called a very cohesive forward and reverse engineering together with composability. What it gives essentially is our ability to understand a legacy bank ecosystem and then seamlessly apply a lot of intelligence to it so that the deposits are then insured in a different set of banks overnight. That creates a very different confidence for the bank and the consumers. That's fascinating technology.
That's awesome, Darryl. How about you? Yes, I'm so excited. We've been doing a lot of work and leveraging what Chris and also Robbie shared. We used a lot of our knowledge bases and as you all know, delivery is changing and transforming every day. Companies are looking for a seamless process and something to solve the problems of time and materials. A lot of the time shared aides are talking about exploring and figuring out the code and things of that nature. We use our years of experience and we leverage transform to actually build a delivery agent. This delivery agent is shaving 30 to 40 percent of the time that we would have taken before with time and materials to actually be more efficient for customers.
It really plays a big part as we think about the future of fixed pricing and fixed pricing with milestones and fixed pricing with outcomes. That takes a lot of the time out up front and really streamlines the process so you can focus on modernization and really solving problems for your customers. Thank you for all of that. Now, AI is changing how we are doing transformations. How do you see transform AI and your businesses evolve over the next few years?
Yeah, as you mentioned, the ability to mix and match agents is something that's super exciting to me. No single transformation is the same. Every time we go to a different client or work with a different client or do internal modernization, the process is different every time. We have different types of code. It could be .NET, could be mainframe. The fact that we can bring those together, mix and match them to address the specific customer needs, that excites me. I don't have to go and rebuild it from scratch every time.
It's awesome. Ravi, I agree with Chris. What's fascinating is that with the technology we are truly able to reimagine critical business processes. We do the best mainframe modernization when we actually don't talk about mainframe modernization. The technology allows us to step back and look at the deposits process, look at a life insurance process, look at an underwriting process, and really reimagine that. Then we get substantiated with composability and a mix and match approach, and that creates superior customer experience and value.
I couldn't agree more with you, Robbie. One of the things that I think Gartner came out with a report earlier this year that 50 percent of POCs will not be funded this year because the organizations, CTOs, or CEOs did not see adoption or see any productivity gains. One of the things, and building on Robbie's point, is that if you use these agents to really streamline the process, you can actually get a lot of return on your time and the type of value that you're spending with clients. So one of the things I think that agents are enabling us to do is get to the modernization story a lot faster and not take so much time working through a lot of the discovery that you would typically have to do.
So again, enabling us to move faster, be more efficient, leveraging agents and not just one agent, multiple agents as we actually work with partners. Chris and I work together across many different opportunities leveraging our partners and leveraging their agents and putting those agents together. Our customers seeing a seamless process is really the future for our customers. So thank you very much. We could keep going, but I really appreciate the partnership over the last few months. You challenged us to think bigger, but you've been thinking bigger with us, and this joint opportunity is going to take us places we've not even imagined so far. So I really appreciate it. Thank you.
The Future of IT Transformation: Putting Tech Debt Modernization on Autopilot
That was pretty exciting. So let's pull all of this together. This is the evolved AWS Transform stack. Over the last hour or so, we showed you how we've expanded the purpose-built agents to add full stack Windows modernization. We have made transform extensible. You can do that through the custom agent where you can define your own transforms. You can connect to a CLI, you can connect to coding agents.
We've made it composable, so you can personalize the experience and, as Chris was saying, for every specific customer you can create a very personalized transformation experience. Through all of this, we see Transform as an extension of your IT workbench—the IT transformation workbench, whether you're a customer, a partner, or trying to do it yourself.
Let's put all of this together. Transform gives you a bunch of capabilities that help you modernize your entire system. As we saw earlier, you can use the VMware agent to do the migration. You can then use the Windows agent in conjunction to modernize that Windows application. There might be some custom off-the-shelf components there. You can use custom to define what that transformation should look like, and there you go.
We've talked a lot about mainframe modernization, but there is such a long tail there that even if you build all the agents, there still might be niche things that we don't know how to transform. For example, we are now working with the airline industry to create transformations for TPF, which is very unique to specific verticals. So you can now mix and match these things. You can imagine using composability not only to create your own workflows but also to create a lot of these custom agents through the AWS custom capability and bring them into this flow as well.
This becomes a place where you can mix and match, and our vision is really to help become this one place that fits into your enterprise ecosystem. Just as our GPS has evolved from paper maps to self-driving cars, we see Transform evolving. We've evolved to become smarter and more automated, and with this, we want you to put your tech debt modernization on autopilot so you can invest in innovation and the future. Thank you for your time and enjoy the rest of the show.
; This article is entirely auto-generated using Amazon Bedrock.
Top comments (0)