Amazon Web Services (AWS) today added additional capabilities to its Kiro artificial intelligence (AI) coding tool that promise to further reduce the time and effort needed to build software.
Kiro now includes a Parallel Task Execution engine that accelerates application development by eliminating the need to complete a set of tasks sequentially, along with a streamlined Quick Plan workflow capability through which Kiro is able to clarify well-understood tasks before autonomously executing them.
Additionally, a Requirements Analysis engine makes it possible to evaluate the feasibility of a software development project before a line of code is ever written.
Darko Mesaros, a principal developer advocate for AWS, said the latter capability makes use of a deep analysis AI model that has been specifically trained using best practices for creating requirements documents to enable AI reasoning to be used to help ensure the software that ultimately is delivered meets expectations. If approved, Kiro will then automatically update the original requirements document.
While most AI coding tools make it simpler to write code, AWS is now moving beyond those tasks to streamline additional tasks that tend to consume a lot of time that could be better spent thinking through how best to solve the issue an application is meant to address, added Mesaros.
Ultimately, the overall goal is to change the role of an application developer in a way that enables them to spend more time on strategic planning, he noted.
It’s not clear to what degree software engineering teams are adopting AI coding tools, but as more capabilities are added in the form of what AWS calls “specifications,” the more widely tools such as Kiro will be adopted.
The specifications-based approach AWS has developed makes it simpler for an AI agent to perform tasks. Specifications are structured artifacts that formalize a development process to provide a systematic approach for transforming high-level ideas into a detailed implementation plan that an AI agent can implement in parallel.
In the case of AWS, the company has created Kiro Powers, a suite of AI agents that have been trained to automate a task, such as reviewing code, using a set of specifications that guide the AI agent in a way that ensures the task is reliably completed.
Mitch Ashley, vice president and practice lead for software lifecycle engineering at the Futurum Group, said AWS is repositioning Kiro from code generator to a specifications-driven, agent-led software development platform. Engineering procurement now shifts from copilot accuracy to specification quality and agent governance, he added.
Teams treating specifications as throwaway prompts will produce unreliable agent output at scale, noted Ashley.
Hopefully, the number of bugs and vulnerabilities being generated during the development of any application will one day drop to zero. In the meantime, however, the one thing that is clear is that building and deploying software in the age of AI is becoming much less tedious for all concerned.
The challenge now is determining exactly which tasks to assign to an AI agent in a way that can be validated by the human developers that are still ultimately accountable for the quality of an application.