Last week I attended NDC Sydney, a large global tech conference aimed at software developers. The vibe at the conference was excitement and the AI talks were definitely front and centre.
While there were the usual never ending debates around MCPs (MCPs are dead, MCPs are a must-have, MCPs are just a temporary tool until CLIs can catch up), what stuck with me was how few people were confidently using AI and agentic tools in their daily workflows.
A number of the talks I attended during the two day conference were focused on building confidence with the tools and making sure that there was still a focus on learning and reading code, that we weren't loosing our ability to read and understand code because it's been passed off to an AI agent.
I found these ideas of code literacy and competence really interesting. It occurred to me, listening to some of these speakers, that maybe my willingness to pick up these tools, comes from being a self-taught software engineer. I don't have the same level of code literacy as some of my peers and while that hasn't been an issue, I'm able to catch up and learn faster with the use of AI.
Pairing with my agent and using it as a built-in rubber duck has helped my confidence and my understanding of the code that I am writing. While for some people, these tools are just seen as "code monkeys", doing the work so we can sit aside them mindlessly watching the thought process, I am using it as a discussion tool, asking questions, building my knowledge to bridge any gaps in my education and understanding of what is going on.
While I don't trust the code, and constantly check for mistakes or places for improvement, I am able to ask questions and understand why certain choices have been made. Similar to the discussions I would normally have with senior team members.
I definitely understand some of the hesitancy to utilise the tools and they can definitely be daunting to start with but here are my takeaways from NDC Sydney and from my experience:
AI tools are a copilot: they shouldn't be in the drivers seat without someone keeping an eye on them to make sure they're on track
Ask questions: if you're unsure of why something has been selected or why a particular piece of code has been written, ask! If you don't know how it works, then you don't know that it works
Review everything: you own the code you write, not the AI agent. If you merge code that looks right but hasn't been verified, that is on you as the approver/author of the code
Understand the tools: there are many modes available across all AI tools, it's important to understand them in order to utilise them correctly. Just because the AI tool is acting the way you want, doesn't mean that it's the most efficient way to utilise it
Rubber duck it: AI is a pair programmer, use it the same way you would a rubber duck. Talk to it (out loud if you need to), ask it questions, make it check and re-check everything written to make sure that you're happy with it. Go over the same code suggestions with different models to receive the best outcome
The biggest takeaway for me wasn’t about new tools or workflows, but about mindset.
Staying in the driver’s seat means being intentional about how and when we use AI. We should not be defaulting to it, but at the same time, we should not be avoiding it.
It’s not about handing over control but rather about using AI deliberately, where it adds value, while staying responsible for the outcome.