Rapid Dynamic Software Development

Looking Forward To The New Year

It was a typical end of year for me – at work that is. I watched some of the Amazon Web Services (AWS) re:Invent conference sessions on YouTube and as per usual it got me thinking about what I should possibly set my mind to in the new year in terms of tech learning.

Artificial Intelligence (AI) has been so pervasive and hyped that I had hitherto kept it at arms length and focused on my usual information technology (IT) learning approach and staples. I was happy to be aware of AI stuff, rather than get "hands-on" with it, at least for the time being.

Whilst I'd been very impressed participating in an internal evaluation of Microsoft Copilot during my at-home web development projects, I felt Copilot was weak in some of the areas I needed it to be strong (weak at refactoring existing code but awesome at creating new code) but I thought maybe next year will see me embracing other AI tools, such as Cursor or Claude Code for example, as I already had a couple of project ideas in mind.

Shattering The Status Quo – Like A Sonic Boom!

The energy and freshness of the AWS re:invent presentations I watched really highlighted what a forward-looking and driven organisation AWS really is. They are seriously "challenging the status quo" to drive out new and better ways of working, as well as innovating new and improved products for solution builders like Capgemini in the harsh world of "projects".

It's clear that AWS are putting generative and agentic AI technologies to work in ways that many of us might never have imagined a couple of years ago and clearly some revolutionary and strategic thinking has gone into this!

AWS eat their own "dog food", i.e. they use many of their own products in anger, so they are born out of genuine business needs, albeit with an initial Amazon focus and bias. (Nothing wrong with that business model.)

In my opinion, thinking out loud, if you continue with merely improved tooling (like adopting Copilot into your code editor for example) but using the "same-old same-old" development approach, then a quantum leap in productivity or capability isn't likely to happen that often or necessarily any time soon, even though no one will complain about "shiny" new tools and the on-going evolution of improvements.

The IT industry has always believed in Continuous Improvement and virtually all customers write it into their contracts as a standard expectation. New tools, technologies and techniques must therefore pay for themselves quickly, regardless of the license costs and steep learning curves!

However, if you dare to imagine, and then go further and attempt a completely different way of working, you may just experience that long awaited quantum leap. Boom! 😊

Hats off to the brave investors and innovators at AWS and also at the Capgemini Research Institute and Applied Innovation Exchange who patiently and relentlessly pursue change and excellence!

The Brave New World

During my learning time I was particularly struck by a presentation about "AWS Kiro" which significantly challenges the traditional development approach.

(Note: At time of writing, AWS has released AWS Kiro in developer preview mode, what most of us would typically describe as a public beta release i.e. it is still evolving based on beta feedback and not yet a final product but it is very impressive and it comes as no surprise that many people have downloaded and started playing with it.)

So what exactly is AWS Kiro and what does it do?

According to the Internet:

AWS Kiro is an AI-powered, agentic IDE designed for building, managing and deploying AWS applications and infrastructure through a chat-driven, specification-based workflow. It enables both developers and non-technical business users to create production-ready solutions using natural language instructions, significantly reducing development cycles.

Can you believe the claims that this product is making? They seem pretty serious.

Having had lots of internal training on Generative AI (GenAI) and Agentic AI, and taken a little time to study AWS Kiro, the product seems like a serious game changer to me!

I'm not saying AWS Kiro is the answer, but products of this genre (i.e. those that put generative and agentic AI technologies to work) will probably become game changers as the concepts involved gain traction, the products themselves mature and the IT industry adapts because the results are too significant to ignore.

AWS Kiro Is A Game Changer

So what makes AWS Kiro a potential game changer?

Faster, Cheaper, Better

Kiro allows you to develop software significantly faster. It can probably do it cheaper and better too!

No Code / Low Code

Non-technical business users can build applications directly based on their own business expertise.

Citizen development tools like Kiro replace traditional coding with natural language specification. Business users describe what they need in plain English and Kiro builds working applications that those users can iterate and refine themselves.

You can either chat (using Vibe) to get started or begin writing specifications and steering directives. There is no need to establish an end-to-end workflow.

The approach succeeds because business users no longer translate their needs through analysts and developers allowing their domain expertise to rapidly become working software.

Natural Language

The ability to express requirements in natural language is a massive benefit and has long since been a "Holy Grail" of software development.

Trying to explain Agile techniques to customers to get them to express their requirements as user stories is always an interesting experience at the start of a project, as is explaining the Product Owner concept and agreeing who that is and having access to that person.

Shifting The Lifting

Kiro harnesses both generative and agentic AI technologies, shifting the lifting in ways that highly impact and accelerate the development process at a time when many organisations are still figuring out how to apply these technologies.

What were once lengthy, manual tasks are offloaded onto highly capable agents that work tirelessly in the background. It's like the high-performance team of 10X developers you always dreamed of but with no egos to manage! 😊

Comparing Kiro To The Traditional Development Approach

Let's compare the traditional agile development approach that we all know and love with using AWS Kiro:

It is apparent that the Kiro agentic IDE approach is radically different to the traditional agile development approach because of the use of natural language specifications driving the creation of user stories, acceptance criteria, technical designs and tasks resulting in much of the work being offloaded onto agents and entirely re-worked when triggered by changes. Consequently, everything is happening at much higher velocity and therefore in significantly more compressed timescales.

Another obvious difference is in development team make-up, with Kiro having a stronger leaning towards business people rather than IT people. (In a world where there is a recognised shortage of IT skills, this is a significant benefit but does not fully negate the need.)

Rapid Dynamic Software Development

What Makes The Kiro Approach Rapid?

A key strength of Kiro is "shifting the lifting" so that agents do the bulk of the work.

Once initial specifications and steering directives are set in place, the work is then performed by agents. You can start with very little and incrementally add to your requirements based on the results you are seeing. (This approach is traditionally referred to as rapid prototyping.)

User stories are generated from the requirements as are acceptance criteria, tasks and architectural designs. Whatever is generated can be "tweaked" and even reverse-engineered automatically back into specifications.

Iterative refinement lets users test with real data immediately instead of waiting for complete implementations.

Integrated tools eliminate the need for expert developers to assemble complex tool chains.

Whenever you have a smaller team, communication improves, especially when that team has good business domain knowledge from the outset.

What Makes The Kiro Approach Dynamic?

Kiro is dynamic because different development activities can trigger agentic tasks and agents are continually improving their knowledge and skills. As tools, technologies and techniques evolve, so too do agents.

Model Context Protocol (MCP) allows agents to access assets in systems to gather information that will enable them to achieve their goals. This may be the latest data or documentation.

Agents are able to work together and new agents are becoming available all the time.

My Ephinany And The Radical Writing On The Wall

The more I thought about Kiro, Agentic AI and what people were saying about how it will change the way we develop software in the future, I wasn't sure if I was having an epiphany or seeing the writing on the wall!

The epiphany said:

"This completely different way of working solves so many problems and takes so much time out of who, what and how we do things today! Why would you carry on working the same way when you could work this new and improved way?"

The writing on the wall said:

"You need to learn to do things in an entirely different way to achieve the same successful outcomes faster, cheaper and better or exit the competition as a loser. Adapt or quit!"

So I watched more videos and read a few more articles.

I eventually concluded that it takes a brave man (RIP) to continue using a sword when your enemy is using a machine gun!

Think about it ...

If I asked you to get me a copy of a very thick book, would you even contemplate writing it out by hand? No way! Yet that's exactly what people did in the past until someone invented the printing press and later the photocopier etc. (These days you'd probably just order it on Amazon and get it the next day. 😊)

Similarly, AWS Kiro aims to solve the same problems, differently.

It's like comparing how NASA designed and built rockets with how SpaceX do it today. It's not that NASA weren't awesome and still are – things have simply moved on.

New people, fresh ideas, lots of learning from the past leading to new and different approaches and serious advances. The problem itself may not have changed but the alternative approaches to solving it have yielded progress!

I've been designing and building software solutions for more than four decades. I think I know what permanent change in the IT industry looks like. I also thought I'd learned how to consistently move with the times. But this new AI-powered, rapid, dynamic software development approach is radically different!

Shocking But Inevitable

Two recent revelations had shocked me, yet somehow didn't really surprise me.

First, hearing that a certain large organisation had replaced several thousand people with a single AI chatbot to achieve huge savings and significant business improvements. Yes, the chatbot did it all. There was no competition.

Second, colleagues at the Capgemini Research Institute looking at Agentic AI tools had gone from a Scrum team of 10 people doing two-week sprints down to a two-person team doing two-day sprints! Better yet, the two-person team had a stronger business bias than technical leaning. Again, how can you compete with that?

How Does This Affect Architects?

In case you hadn't noticed, we ask AI to do stuff and it kind of just does it. Often it makes many decisions on our behalf without even asking. For some things that is okay but certainly not for all things.

Some decisions, typically relating to non-functional requirements, are too important to be made without asking and require various options to be considered, and sometimes even formally evaluated, before a decision can be made. That's exactly where Architects fit in!

Sure Architects can use AI to rapidly identify options and make suggestions but that still doesn't negate the need to properly explain things to stakeholders and allow them to make the important decisions based on their strategy, costs, timescales etc. and having properly understood the trade-offs and potential risks and consequences.

Another important consideration is when, where and how to apply AI in software solutions. Again, this is where Architects should be stepping up!

So architects still have an important role to perform in this ever advancing world of AI.

The Nagging Question

So what will people do if generative and agentic AI start doing what some people once did? At this moment in time, that's a good question. But we'll figure it out. Eventually.

Yet Another Learning Dilemma

So after all this, my new year probably isn't going to bring just a few new technologies – it could be a drop everything, turn around and start learning a whole bunch of new stuff from first principles!

It's all just one more reminder of the continuous learning dilemma we all face working in the IT industry!

But when I look back over my career and think about it, this is not a new problem. It's actually happened many times before!

Like

Thank you for liking this page.

Tim Simpson
16th January, 2026
#AI-Architecture | #LifeAtCapgemini

« Previous Blog Post Blog History