Last week I attended the AWS Summit Johannesburg, and it turned out to be one of the more energising conference days I’ve had this year. As someone working in cloud infrastructure and DevOps, I went in expecting the usual mix of product pitches and networking. Instead, I ended up trying to take in a bit of everything on offer, from a hands-on Kiro challenge in the Agentic AI & Generative AI track to a look around the AWS Developer Community Zone. I also spent time exploring Amazon Connect and the AWS Partner ecosystem. In this post I’ll share what stood out across the day, starting with the part I got the most hands-on with: the Kiro challenge. 

A Hands-On Challenge with Kiro 

The most hands-on part of my day was inside the Agentic AI & Generative AI track, where sessions covered building end-to-end platforms and autonomous agents on Amazon Bedrock, using agentic AI to modernise legacy code, and practical applications of enterprise AI assistants. That’s also where I got to try Kiro, AWS’s agentic AI IDE, as part of a team challenge.  

Kiro takes a spec-driven approach to coding: instead of just generating code from a prompt, it breaks a request down into structured specs and tasks first, then works through implementation with you. Out of the four tasks we had, one of the ones I remember was having to use Kiro to detect bugs in an existing piece of code. 

Ninety minutes is not a lot of time so we didn’t get as far into the bug-detection side as I’d have liked, since testing ran out of time before we could put Kiro’s detection capabilities through their paces properly. Even so, watching Kiro turn a rough prompt into a structured spec and then working code, live and under time pressure, was a good demonstration of where agentic coding tools are heading. 

Getting to Grips with Amazon Bedrock 

The other part of the day that stuck with me touched on Amazon Bedrock. For anyone unfamiliar, Bedrock is AWS’s fully managed platform for building and scaling generative AI applications and agents, removing the infrastructure burden of provisioning GPUs or training models from scratch. It gives developers direct access to a choice of foundation models, including Anthropic’s Claude and Amazon’s own Nova and Titan models, all through a single managed API. For someone with a DevOps background, that’s the part that resonates most: it turns generative AI into something you can integrate the same way you’d integrate any other managed AWS service. 

Why Agentic AI Took Centre Stage 

Generative and agentic AI have been the headline topics at every AWS event for a while now, so I went in wondering whether the hype would match the substance. Both the hands-on-challenge and the Bedrock talk that followed focused on practical realities rather than flashy demos. That focus on production readiness over spectacle is what made both worth writing about. 

A Look Around the AWS Developer Community Zone 

I also spent some time walking around the AWS Developer Community Zone. It had interactive sessions with AWS Heroes, people playing AWS Buildercards, a live card game, and user group discussions running on the Community stage. I didn’t sit in on a full session there, but it was a good reminder that a lot of the summit’s value comes from the community side, not just the product tracks. 

Beyond the Sessions: Connect and the Partner Ecosystem 

Part of what made the day worthwhile was how much was happening outside the main talks. I spent time at the Amazon Connect area learning about AWS’s cloud-based contact centre service, and in conversations about the AWS Partner Network, including what’s involved in becoming a partner and how flexible, on-demand cloud consumption models fit into that picture. None of it was as deep a dive as the Kiro hackathon or the Bedrock session, but it gave me a broader sense of how wide the AWS ecosystem is beyond the tooling I use day to day. 

Conclusion 

The AWS Summit Johannesburg was a good reminder of how quickly agentic and generative AI tooling are moving, from Kiro’s spec-driven approach to coding to Bedrock’s production-ready model access. I intentionally tried to sample a bit of everything on offer rather than go deep on just one track. I came away with more directions to explore than I expected, including experimenting with Kiro on a personal project, digging further into Bedrock, and learning more about the partner side of AWS. If you attended the summit too, I’d love to hear what stood out to you.