devopsy.kinsta.cloud

Tag: AWS

AWS Adds Specialized AI Agents to Kiro Coding Tool to Automate DevOps Tasks

Amazon Web Services (AWS) today at its re:Invent 2025 conference extended its Kiro artificial intelligence (AI) tool to provide access to a set of agents that have been trained to trigger specific actions as they observe the way code is being created.

Adnan Ijaz, director of project management for Kiro at AWS, said these specialized agents, dubbed Kiro Powers, enable developers to invoke steering files that are extensions to the specifications-based approach that AWS has adopted to improve the quality of the code generated by an AI tool by narrowing the scope of the tasks that AI agents are allowed to perform.

The overall goal is to make AI agents more efficient by reducing the amount of “context rot” in a way that serves to better fine tune the outputs generated, he added. That approach ensures the AI agent doesn’t become unduly influenced by data that is outside the scope of the task it has been assigned that could result in a hallucination, he added.

AWS earlier this week also extended its AI portfolio to include a set of AI agents for optimizing DevOps workflows. AWS is now previewing an AI agent within its Kiro AI coding tool that autonomously works in the background to optimize tasks such as pull requests and provide feedback to developers. Additionally, AWS previewed an AWS Security Agent to review and test code, including conducting penetration testing, and an AWS DevOps Agent that will always be on call to help manage IT incidents.

The AI agents added to the Kiro AI coding tool are also designed to only load when needed to minimize costs. Trained using best practices developed by AWS, those agents are able to access hooks and Model Context Protocol (MCP) servers to trigger specific actions, including invoking third-party services from Datadog, Dynatrace, Postman and others, as code is being developed, said Ijaz.

For example, an AI agent might automatically invoke the Datadog monitoring service to identify an infrastructure issue that is adversely impacting the ability of the code generated by Kiro to be optimally run, he added.

In time, more specialized AI agents will be added to the AWS portfolio that will independently perform tasks in the background in real time as application developers build their code, noted Ijaz. Those capabilities in time will substantially reduce the amount of friction that software development teams currently encounter as they invoke multiple tools, he added.

Mitch Ashley, vice president and practice lead for software lifecycle engineering at Futurum Group, said AWS is providing agents the hooks they need to operate as real contributors to the development of software. When an agent can reason using telemetry data provided by, for example, Dynatrace or validate application programming interfaces (APIs) created using Postman tools, it can more fully participate in development workflows, he added.

Ultimately, AWS plans to expand the ecosystem of providers of DevOps platforms and tools that can be invoked by its AI agents. The only issue that remains to be seen now is how the role of the DevOps engineer will evolve as more of the capabilities provided by those tools are shifted further left across the software development lifecycle (SDLC).

December 3, 2025
AWS Transform Modernizes Any Codebase, App, API or Runtime

Platforms, paradigms, processes and processing itself all evolve. Because the information technology industry crosses from one chasm to another on an apparently endless loop of perpetual change, applications and wider systems need to be almost continually modernized. Amazon Web Services wants to enable that process with AWS Transform, a service that now ships with new agentic capabilities to enable the rapid modernization of any code or application.

Why modernize in the first place? Because, says AWS, organizations should be looking to reduce their technical debt and shift information resources towards modern tiers that offer better, wider or more sustainable compute efficiencies.

Yes, legacy software is software that still works, but there are few purists that hang on to the notion of sustaining an archaic system when workable, functional, cost-effective and secure methods of moving forward exist. Organizations also need to modernize their complete Windows environments to reduce expensive licensing costs and improve security and performance… and AWS knows all this.

Agnostically Holistic Logistics

As such, AWS has been agnostically holistic in its approach and says that AWS Transform has a “custom capability” that makes large-scale modernizations possible for all legacy systems across any codebase, library or framework. The technology promises to perform “full-stack Windows modernization” (i.e. front end presentation layer technologies, middleware and drivers, plus core infrastructure mechanics such as operating system, memory and so on) across all layers,

“A typical organization spends 30% of its teams’ time on manual modernization work, otherwise known as tech debt. AWS Transform, the first agentic AI service for transforming Windows .NET applications, VMware systems and mainframes, has already helped customers modernize 4x faster with existing transformation capabilities. Customers have used AWS Transform to analyze an estimated 1.1 billion lines of code and save more than 810,000 hours of manual effort,” noted AWS, in a press statement.

AWS already said any codebase and any framework, the company has underlined the breadth of reach here and also said that AWS Transform works across API, runtime, architecture, language and even company-specific programming languages.

With pre-built transformations for common patterns (e.g., Java, Node.js and Python upgrades) and custom transformations for organization-specific tasks, a specialized agent executes consistent, repeatable and high-quality transformations.

Cumulative Learning Via Transformation Agent

The transformation agent automatically captures feedback and continues to improve over time, so each subsequent transformation becomes more reliable and efficient. In terms of real world use, modernizations that used to take two weeks now take three days.

“The AWS Transform agents start by analyzing the complete Windows stack and proposing coordinated modernization plans across all layers. Once approved, the agent transforms the application, UI framework, database and operating system, while providing updates and comprehensive transformation summaries. The new capabilities enable the rapid modernization of full-stack Windows and SQL Server systems to open source alternatives, freeing customers from expensive licensing agreements,” enthused AWS, in a technical product statement.

Three new agents in AWS Transform build on the code analysis, business rule extraction and technical documentation capabilities already available in the service. The agents help produce activity analysis to aid modernization and retirement decisions, blueprints for reimagining legacy code into clear business functions, capabilities, flows and data usage, as well as simplified domain decomposition.

Additionally, new task agents speed up test planning and validation by automatically generating test plans, test data collection scripts and automation scripts, which traditionally take up to half of project timelines.

“The breadth of COBOL, Java, and Windows code platforms that AWS Transform addresses matters. Modernization spans code, APIs, runtimes, architectures, and software stacks beyond the programming language itself,” said Mitch Ashley, VP & practice lead, Futurum Research. “A service that understands those layers together and executes against them as an integrated plan brings modernization closer to everyday engineering practice. For organizations burdened by tech debt, this brings modernization project success closer to reality.”

Air Canada, Experian, QAD, Teamfront, Thomson Reuters and Verisk are currently using AWS Transform to help eliminate their tech debt.

December 2, 2025
AWS Extends Kiro AI Tool to Generate Higher Quality Code

Amazon Web Services (AWS) today with the general availability of its Kiro artificial intelligence (AI) coding tool added additional capabilities, including a version that can now be invoked via a command line interface (CLI).

At the same time, Kiro now enables property-based testing (PBT) to measure whether the code generated matches the behavior defined in the original specification.

Kiro now also provides the ability to rewind to a previous checkpoint the tool creates, enabling developers to rollback changes in addition to allowing Kiro to work across multiple project roots simultaneously.

Finally, AWS announced that it is giving away one year’s worth of Kiro Pro+ for qualifying startups that sign up by the end of this year.

David Yanacek, a senior principal engineer for agentic AI at AWS, said these capabilities collectively will make it simpler for developers to generate higher-quality code using an approach based on specifications that focus the AI model on a narrow range of tasks to ensure better outputs. That output can now also be tested using a set of properties that the tool creates based on those initial specifications.

PBT is designed to probe code using a technique called “shrinking” that functions almost like a red team trying to break your code. When it finds violations or counter-examples of better code, Kiro can automatically update your implementation, or surface options for you to fix the original specification, the code created or, if necessary, the PBT itself.

The overall goal is to produce more secure code that is easily understood by application developers that might need to debug it, he added. That approach provides the added benefit of reducing stress on DevOps engineers that might otherwise be created using AI coding tools that don’t rely on specification and PBT to generate better code, noted Yanacek.

Additionally, application development teams that opt for the CLI version of Kiro can make use of custom AI agents they have created to automate specific tasks versus relying on the steering files embedded with the AI coding tool, he added.

Mitch Ashley, vice president and practice lead for software lifecycle engineering at the Futurum Group, said CLI support in Kiro reinforces how important it is to have development tools that fit cleanly into the AWS ecosystem. Developers need an IDE and terminal workflow that share the same agents, steering files, and security settings so they can move between environments without losing context. The added capabilities in this release move Kiro beyond code generation into disciplined, verifiable development to provide a more dependable path for organizations that want AI powered development to scale with quality, not just speed.

There is, of course, no shortage of AI coding tools. AWS, however, is making a case for a more structured approach that walks developers first through requirements and visual designs to generate a specification. That approach enables developers to keep AI models focused on a specific task versus more arbitrarily adding steps that are beyond the scope of the task assigned, noted Yanacek.

That capability is critical because otherwise application developers will spend a significant amount of time creating their own prompt and context engineering workflows to achieve the same outcome, he added.

Alternatively, application developers can also revert to more of vibecoding technique if, for example, all they need to create is a simple script, noted Yanacek.

It’s not clear how much code is being generated using AI or, more importantly, how much of that code is considered good enough to use in a production environment. The one thing that is for certain is that without the proper guardrails in place AI coding tools quickly become too much of a good thing that may do more to reduce productivity than enhance it.

November 17, 2025
Vibe Coding vs. Spec-Driven Development: Finding Balance in the AI Era

David Yanacek, Sr. Principal Engineer, AWS Agentic AI, dives into the rise of vibe coding versus traditional spec-driven development. As generative AI continues to transform how code is created, developers are moving faster than ever, but not always with the same level of structure, review, or intent that defined earlier software engineering practices.

Yanacek explains that vibe coding—writing code by “feel” with the help of large language models and AI copilots—has opened the door to rapid prototyping and experimentation. However, it also raises new questions around reliability, reproducibility, and long-term maintainability. Without clearly defined specifications and governance, teams risk introducing technical debt and vulnerabilities at a scale that traditional code review processes were never designed to handle.

By contrast, spec-driven development enforces design discipline and predictable outcomes, but it can feel slow and rigid in fast-moving environments. The challenge, Yanacek notes, is to strike a balance between the agility of AI-assisted creativity and the rigor of structured engineering. That means applying DevOps principles, such as automation, observability, and feedback loops, not just to deployment, but to the AI-assisted coding process itself.

Yanacek also highlights the evolving role of developers and security engineers as AI becomes embedded in the software lifecycle. In his view, the organizations that succeed will be those that treat AI as a collaborator rather than a crutch, leveraging it to scale quality, not just speed.

Innovation is accelerating, but without structure and security, that speed could come at the expense of trust and control.

November 12, 2025
AWS to Help Fund Open VSX Registry Hosted by Eclipse Foundation

The Eclipse Foundation this week revealed that Amazon Web Services (AWS) is providing funds needed to strengthen the reliability, performance and security of the infrastructure used to make the Open VSX Registry, an online marketplace for Visual Studio Code extensions.

Like other AI coding tools, the Kiro tool developed by AWS relies heavily on Visual Studio Code extensions that the Eclipse Foundation makes available under an open source license.

Mike Milinkovich, executive director of the Eclipse Foundation, said that as these tools have become used more widely, the stress on the underlying infrastructure used to host Open VSX Registry has significantly increased. The registry now hosts more than 7,000 extensions from nearly 5,000 publishers, with downloads exceeding more than 110 million per month but is only supported by three engineers.

The Eclipse Foundation has not agreed to migrate Open VSX Registry to the AWS cloud in return for this investment, but might as part of an effort to reduce costs. AWS, however, is providing infrastructure and services that will improve availability, performance and scalability across the systems currently used to host the registry, along with future initiatives planned by the Eclipse Foundation.

Nor is the AWS investment a direct result of an open letter that several open source foundations recently published concerning the economic viability of providing services such as Open VSX Registry, the need for additional sources of funding remains, said Milinkovich.

At the core of the issue is the simple fact that the vendors and non-profit entities that originally launched and support open source projects that are consumed as a service are shouldering the cost of maintaining the underlying IT infrastructure required with no way to recoup that investment. The overwhelming majority of large-scale users, including commercial entities that extract economic value, consume these services without contributing to their financial sustainability.

Rectifying the economic imbalance is going to require commercial and institutional partnerships that help fund infrastructure in ways that are aligned to usage and/or adoption of tiered access models through which fees might be levied on organizations that consume large amounts of IT infrastructure resources.

The open letter is the first step toward raising awareness of an issue that will eventually need to be addressed, for example, by throttling and rate limiting to moderate usage of package managers that were never supposed to become critical elements of a software development process. The hope is that individual developers that rely on those services will become more judicious about how they are incorporated into software development workflows.

It’s not precisely clear how much is being spent on IT infrastructure to support these projects but costs are in the tens of millions with no sign of abatement. In effect, open source projects are becoming economic victims of their own success. The issue now is determining how best to fund these projects in a way that is more sustainable with the aid of more financial resources that will, hopefully, soon be more forthcoming.

November 6, 2025
AWS Previews AI IDE to Accelerate Software Development

Amazon Web Services (AWS) today previewed an internal developer environment (IDE) infused with artificial intelligence (AI) capabilities that promises to make it simpler for application developers to invoke multiple agents.

Srini Iragavarapu, director of generative AI applications and developer experiences for Amazon Q Business, said the Kiro IDE makes use of specs to provide developers with a set of best practices to follow and a set of hooks for invoking various tools and backend services.

Kiro specs guide developers and their AI agents through the steps required to build an application. Kiro hooks, meanwhile, provide an event-driven framework for invoking tools that run in the background. Those tools and services can, for example, be used by AI agents to continuously optimize code as it is developed, said Iragavarapu.

The Kiro IDE enables application developers to go well beyond simple vibe coding, however, to enable application developers to create code that is of sufficient quality to be deployed in production environments. The goal is to take vibecoding to the next level, he added

Built on top of the open-source Code OSS editor and is compatible with multiple programming languages. The overall goal is to leverage AI to streamline the development process, said Iriagavarapu.

For example, Kiro unpacks requirements from a single prompt and then generates user stories for viewing, creating, filtering and rating reviews. Kiro then generates a design document by analyzing a codebase and approved spec requirements along with data flow diagrams, TypeScript interfaces, database schemas and application programming interfaces (APIs) for endpoints.

Additionally, Kiro generates tasks and sub-tasks, sequences them correctly based on dependencies, and links each to requirements. Each task includes details such as unit tests, integration tests, loading states, mobile responsiveness and accessibility requirements for implementation. Kiro also includes support for the Model Context Protocol (MCP) for connecting specialized tools.

There is also a task interface that, in addition to enabling developers to trigger tasks, also surfaces a progress indicator to track execution status. Developers can author code to update specs or update specs to refresh tasks.

Hooks, in effect, enforce consistency by ensuring the same quality checks, code standards and security validation fixes are employed across software development teams. It also provides the added benefit of making it simpler to onboard new application developers to a project, said Iragavarapu.

It’s not clear how rapidly DevOps teams are embracing AI to build and deploy software, but a recent Futurum Group survey finds that 41% of respondents expect generative AI tools and platforms will be used to generate, review and test code. The Futurum Group also finds that over the next 12 to 18 months, organizations plan to increase spending on not only AI code generation (83%) and agentic AI technologies (76%), but also existing familiar tools that have been augmented with AI.

The one certain thing is that as the pace of application development continues to accelerate, the processes currently used to manage software engineering are now being put to the ultimate test.

July 14, 2025
AWS Extends MCP Support in Amazon Q Developer to Multiple IDEs

Amazon Web Services (AWS) has added support for the Model Context Protocol (MCP) with the Visual Studio Code and JetBrains plugins that are provided for Amazon Q Developer, a set of artificial intelligence (AI) agents that automate a range of software development tasks.

Originally developed by Anthropic, MCP support with these integrated development environments (IDEs) will make it simpler to invoke external tools that invoke some type of large language model (LLM).

Srini Iragavarapu, director of software development for Amazon Q Developer Coding Assistant, said this capability will make it easier to provide AI agents with additional context by providing access to additional data sources and application programming interfaces (APIs). Armed with those insights, it then becomes a lot simpler for developers to optimize the prompts they create to invoke Amazon Q Developer, he added.

That capability, in addition to creating more accurate code, will also make it simpler to integrate planning tools, create user interface components from designs and generate database documentation from actual schema without having to create additional custom integration code, noted Iragavarapu.

Previously, AWS only provided that capability via MCP support for the command line interface (CLI) option it makes available for Amazon Q Developer.

Over time, multiple classes of AI agents will emerge that provide different levels of reasoning capabilities that will be assigned tasks based on the level of complexity involved, said Iragavarapu. AWS also continues to evaluate the agent-to-agent (A2A) protocol put forward by Google to make it simpler for AI agents to communicate over a standard set of interfaces, noted Iragavarapu.

It’s not clear how widely AI coding tools have been adopted a Futurum Research survey noted, but found 41% of respondents expect generative AI tools and platforms will be used to generate, review and test code.

There is, of course, already no shortage of AI coding tools, so the pace at which code is being developed has already begun to accelerate. Exactly how much of that code is making it into production environments is unknown. However, the next major challenge will be incorporating AI agents into the software engineering workflows used to deploy applications. Many of the DevOps pipelines relied on today to deploy software were not designed to handle the volume of code being generated by AI tools. The expectation is that many of the brittle scripts that DevOps engineers rely on today to create pipelines will eventually be replaced by AI agents that can perform tasks at higher levels of scale.

In the meantime, each organization will need to determine to what degree they may want to standardize on a specific AI tool versus letting each developer select one based on their personal preference. The important thing is to remind developers to review the code created by AI coding tools that, on occasion, are still prone to hallucinations. After all, even if a machine wrote the code, it’s still a human developer who is ultimately responsible for the quality of the code being created.

June 19, 2025
AWS Extends Cloud Security Reach to Include DevSecOps Tools to Scan Code

Amazon Web Services (AWS) this week made Amazon Inspector, a code scanning tool for surfacing vulnerabilities that is designed to be natively integrated with GitHub and GitLab platforms, generally available.

Announced at the AWS re:Inforce 2025 conference, Amazon Inspector combines a static application security testing (SAST) tool for analyzing application source-code with a software composition analysis (SCA) tool for evaluating third-party dependencies and an ability to scan infrastructure as code (IaC) for valid definitions.

Findings from these scans are then surfaced both in the Amazon Inspector console and within the source code management platform as fast feedback for the developers.

Mitch Ashley, vice president and practice lead for software lifecycle engineering at The Futurum Group, said the expansion of the DevSecOps capabilities provided by AWS into code, dependencies, and IaC enable a critical ‘shift left’ for cloud security, fundamentally embedding vulnerability management earlier into the development lifecycle. That unified approach addresses a long-standing challenge for DevOps teams by accelerating feedback and fostering truly secure-by-design applications, he added.

Additionally, AWS also revealed it has made available under an open source license a software package, dubbed @verifiedpermissions/authorization-clients-js, that enables developers to implement authorization in Express.js web application APIs in minutes. That approach reduces the custom authorization code that developers would have previously needed to create on their own.

Finally, AWS demonstrated how it is internally using Amazon Q Developer, an artificial intelligence (AI) agent framework, to identify vulnerabilities in code as it is being developed.

George Argyros, applied science manager of AWS, told conference attendees Amazon Q Developer makes it possible to “squash bugs” early in the application development process in a way that ultimately serves to improve the quality of the code being created. The overall goal is review code at the time a pull request is made and then surface recommendations to mitigate any issue, with Amazon Q Developer also being used to create and test any patch required.

It’s not clear how widely the DevSecOps tool and platform being provided by AWS have been adopted, but it’s clear the cloud services provider is now moving well beyond securing infrastructure. It’s still the responsibility of organizations to secure the applications deployed on the cloud, but AWS is making the tools and services needed to achieve that goal more accessible.

Hopefully, there will even come a day soon when application development teams are routinely using AI tools to prevent vulnerabilities and misconfigurations from being created in the first place.

In the meantime, however, DevOps teams need to be sure to review the code being generated by AI tools that have been trained using examples of code that is of varying quality. In one instance, that code might be better than what some developers might write on their own. In other cases, segments of code with known vulnerabilities might have been copied multiple times over. In fact, about the only thing that is for certain is that there will soon be more code than ever to scan and review as developers continue to lean on AI tools to create code that they are becoming less inclined to write themselves.

June 18, 2025
Cloud ROI is Broken. Agentic AI Might Be the Fix.

Enterprises flocked to public cloud infrastructure with the promise of speed, flexibility and efficiency.

But many are now grappling with a different reality: Cloud environments are more complex to manage than anticipated, and the associated costs have spiraled far beyond early projections.

Enterprise IT leaders haven’t been shy about sharing this fact. Last year, 63% of enterprise IT leaders responding to an IDC survey said they actively evaluate hybrid or multi-cloud strategies to regain control over cost and performance. In 2023, a report from Andreessen Horowitz cited examples of companies repatriating workloads to on-premises data centers amid questions about whether the cloud was still worth the cost.

Looking ahead, however, the proliferation of agentic AI could represent a turning point. That is, if enterprises approach it correctly.

The Problem Isn’t the Cloud — It’s the Complexity

The infrastructure challenges dragging down cloud ROI aren’t just about cost per compute hour or storage. They stem from the effort required to translate infrastructure goals into action.

Writing infrastructure code, orchestrating dev and test environments, enforcing governance standards, troubleshooting errors and decommissioning unused resources are all labor-intensive processes. Even teams that have invested in infrastructure-as-code (IaC) and automation often struggle to move fast enough to justify the cost of cloud infrastructure.

When we talk with our customers, we typically see these challenges materialize in two ways: lost productivity and inflated cloud costs. Delays to deliver cloud environments extend timelines, resulting in higher labor costs and slower time to market, while wasted cloud costs leave less budget for other technologies.

Agentic AI has the potential to tackle some of these challenges by performing the day-to-day tasks that hold infrastructure teams back.

Consider a state where infrastructure engineers incorporate AI agents to write infrastructure code, review that code for security and compliance standards before it’s deployed, and monitor the state of that infrastructure to identify risks, perform root-cause analysis on errors, or detect oversized or idle resources.

Some of these tasks require hours of work from skilled infrastructure engineers. Others aren’t even being done today.

What Could Hold Back This Transformation

The economic stakes for this transformation are high.

Enterprises that successfully harness agentic AI to automate and optimize their cloud operations can accelerate time to market, price their products more competitively and show profitability gains.

And that’s in addition to the day-to-day benefits: less monotonous work for DevOps engineers and SREs, happier developers who get what they need on demand and lower turnover as a result.

But despite the promise of agentic AI, adoption is far from guaranteed. These tools often require a significant upfront investment in new platforms, training and change management. For IT leaders already facing budget scrutiny over ballooning cloud bills, the idea of spending even more on a new class of tools may seem like throwing good money after bad — especially if past cloud initiatives failed to live up to expectations. Without clear, measurable short-term wins, they may hesitate to adopt solutions that promise long-term transformation.

And then there’s the psychological hurdle for the teams responsible for integrating agentic AI into legacy processes and platforms. Developers and engineers may be wary of embracing AI tools that appear to threaten job security.

At a time when IT decision makers have started looking for the exits to their cloud strategy, they’ll need to consider whether it makes more sense to modernize how they manage their infrastructure. How this plays out will be interesting to watch.

The cloud, for all its promise, demands a new level of operational maturity to deliver ROI at scale. Agentic AI may be the missing ingredient that finally makes cloud living up to its potential.

June 9, 2025
AWS Application Modernization Service Employs AI Agents to Refactor Code

Amazon Web Services (AWS) has made generally available a service that makes use of artificial intelligence (AI) agents to migrate and modernize infrastructure, applications and code.

Sriram Devanathan, general manager for low-code/no-code AppBuilder service at AWS, said the AWS Transform service provides access to a natural language chat interface that makes it simpler to either refactor, for example, VMware, mainframe and .NET workloads or upgrade older instances of Java.

Initially previewed as an extension to Amazon Q Developer, AWS has a series of AI agents that have been specifically trained to both understand how code is constructed and invoke various tools, such as application assessment tools from CAST Services, to transform it. DevOps teams can use AWS Transform to specify objectives, share project context, assess business plans and cost savings, review and adapt transformation plans, identify dependencies, approve code infrastructure suggestions and execute tests.

For example, an AWS Transform agent for .NET can migrate a Windows-based application to Linux up to four times faster than traditional methods to reduce Microsoft cost by as much as 40%, noted Devanathan.

Similarly, an AWS Transform agent for VMware can automate the conversion of on-premises VMware network configurations to AWS equivalents up to 80 times faster than manual approaches.

An AWS Transform agent for mainframe can also decompose monolithic z/OS COBOL applications into components that can be run in the cloud in minutes. AWS Transform agents can, for example, invoke a Graph Neural Networks capability that makes it simpler to refactor applications written in COBOL, JCL (Job Control Language) and those relying on Customer Information Control System (CICS) transaction manager, BMS (Basic Mapping Support) screens, DB2 databases, and VSAM (Virtual Storage Access Method) data files in a set of Java-based microservices that can be deployed on an open source Postgres database.

In many cases, the documentation generated after that migration is more detailed than what was created when those legacy mainframe applications were initially built and deployed, said Devanathan.

Most organizations will still require specialists to migrate and modernize applications, but the total amount of time and effort required will be substantially reduced, said Devanathan. Most IT teams would be well-advised to revisit their modernization strategy by picking a relatively small project to better understand the level of refactoring that can now be achieved using AI agents, he added.

Longer term, AWS is looking to expand the range of modernization tools that its AI agents are trained to invoke, he added.

During uncertain economic times, the amount of pressure being applied to reduce the total cost of IT always rises. As such, interest in migrating legacy applications always tends to rise in an economic downturn. However, the amount of effort historically required to modernize applications has been substantial. In fact, many of these projects have never been completed simply because the cost of migrating applications proved to be much higher than initially anticipated.

Regardless of motivation, however, the total cost of continuing to run legacy applications can, in many cases, be even more prohibitive than the cost of a modernization initiative that, with the help of AI agents, may now prove to be a more viable option.

May 28, 2025