Breaking Into AI: Advice From Young Kiwi AI Engineers

Disclaimer: This article was AI generated from my YouTube video transcript.

Utilising AI is fast becoming the most important skill for business owners, employees, and new graduates alike. The trouble is that a lot of universities and traditional education systems just aren’t keeping up with how fast the space is moving. That gap is exactly why I set up Young Kiwis in AI - a community of more than 150 young Kiwis who network together and share what they’re learning so we can all move faster.

In this session I sat down with two young Kiwi AI engineers working at two of New Zealand’s leading companies - Georgia, an AI engineer at Trade Me, and Brody, a junior AI integration engineer at Xero. Both broke into the space recently, both came in through slightly unusual paths, and both had genuinely useful advice to share. Here are the key themes and takeaways from the conversation.

Their paths into AI were anything but straight

One of the most reassuring things to come out of the chat is that neither of them followed a textbook route into AI - and you don’t need to either.

Georgia had been interested in AI since she was a kid. She started a Bachelor of Science in Computer Science at the University of Canterbury in 2020, right in the middle of COVID, which made the study pathway pretty hectic. At one point a college advisor told her she wasn’t smart enough to work in AI, so she switched to a Bachelor of Commerce majoring in Information Systems with a maths minor. She got a foot in the door as a cloud engineering intern at Trade Me, kept saying yes to AI-related opportunities, and when an internal AI engineering role opened up on a brand new team, she applied and landed it.

Brody learned to code in high school - he built a clunky dating app back before you could lean on AI to write it, which he described as brutal but a brilliant lesson in owning something end to end. He went off to study engineering science (more maths and algorithms than pure software), did a few data internships, and then tried his luck as a stockbroker. He didn’t love it, but while he was there he caught the AI bug - particularly an interest in AI safety and where this technology is heading as it gets more capable. He started vibe coding in his own time, applied for an AI integration role not really expecting to get it, and was pleasantly surprised when his genuine interest and a bit of hands-on building got him over the line.

The lesson in both stories is the same - you don’t need a perfect computer science degree or permission from anyone to work in AI. Curiosity, a bit of hands-on building, and a willingness to keep showing up will take you a long way.

Almost everything technical is learned on the job

When asked what they knew going in versus what they had to learn on the fly, both were honest - the technical AI skills were mostly learned at work, not at university.

Georgia put it plainly - tertiary education taught her a lot of the soft skills and work ethic, but the actual technical side of being an AI engineer she picked up on the job, because the courses simply haven’t caught up with how quickly the tooling changes. Brody had an interest in AI before applying, but had done very little in the way of AI automations or workflows. The small amount of vibe coding he’d done was enough to get him through the practical part of the interview.

His one big takeaway here - never let “not knowing everything about AI” be a blocker. Nobody knows everything in a field this new, and that is precisely the point.

The skills that actually matter

A recurring theme was that raw coding ability is no longer the bottleneck. The tools have gotten so good that the differentiators have shifted.

Thinking in systems, not code. Brody’s line stuck with me - anyone who can draw a good systems diagram or flowchart can basically code now, because you can hand that diagram to a coding tool and it will build it for you. But not everyone who can code can think about how a solution fits with real people. Designing the workflow is the skill that holds its value.
The people side. Both roles are roughly half technical and half human. Brody spends a lot of time explaining technical concepts to non-technical marketing, commercial and product teams, and just as much time getting them to explain their day-to-day so he can find the real pain points. That two-way translation matters as much as the build.
Knowing where the human stays in the loop. Designing the whole workflow and deciding where the human touch points should sit was Brody’s favourite part of the job - and a skill in its own right.
Architecting over coding. Georgia noticed her work shifting from writing code to architecting solutions far earlier in her career than she expected. Tools like Claude have pulled a lot of architecture-level decision making down to intermediate engineers, so being able to design and reason about solutions is increasingly valuable.
Cost awareness. Brody once accidentally spent around $4,000 on a data platform in his first few months because AI tools make it so easy to run something many times without thinking about cost. Building good cost dashboards and reporting into your work is a real skill.

Learning by building beats learning for its own sake

If there was one piece of advice both of them landed on independently, it was this - start by building something you actually care about.

Brody’s number one tip was to pick a project you’re passionate about, fire up Claude Code or Cursor, and just have a go building it from scratch. You learn far more from a project you’re genuinely interested in than from drilling skills in the abstract.

Georgia approached it from a slightly different angle - brainstorm problems you actually face day to day, and they don’t have to be work related at all. Maybe you’re trying to line up your physio plan with your gym goals. A lot of people get stuck in a loop of “how do I use AI in this work project?” when stepping back and solving a personal annoyance makes a much better passion project to learn on.

Both of them also automate their own learning and work. Georgia built a Claude skill that runs a scheduled task each day - it looks at her existing knowledge and her code history, works out where her Python gaps are, and sends her a Slack message every day with something to practise. Brody made a rule for himself - anytime he does something more than three times, he asks whether he could automate it, and tries to “skillify” as much of his job as possible. His tongue-in-cheek goal is to automate his entire role so he could go on holiday.

What they got wrong - lessons from the trenches

The honesty about mistakes was probably the most valuable part of the whole conversation.

You can’t keep up with everything. Georgia, a self-described perfectionist, learned to give herself grace. You can’t watch every YouTube video, read every Anthropic article and catch every new model. You need time to sit with what you’ve learned and actually absorb it - that space is what lets you come back ready to learn more.
Belief in a tool is not the same as adoption. Brody learned the hard way that you can explain exactly how a tool works and people will nod along - but unless it’s visual, exciting, and obviously relevant to their own work, they won’t actually use it. Something exciting beats a thorough but boring explanation every time.
Numbers matter. Brody would proudly tell his manager he’d built three working tools in a week, and his manager would ask for the numbers - how many people used them, how much value they created. It felt frustrating at first, but he came to appreciate that attributing real value to AI work is how you prove it’s worth doing. He now expects to see a lot more formal measurement of AI value across companies.
Keep a running log of mistakes. Brody keeps a list of hundreds of mistakes and learnings from his time in the role. When the whole field is new to everyone, tracking what works and what doesn’t is genuinely valuable.

How they actually ship - and what businesses can copy

Both teams have a strong bias towards moving fast and proving value quickly, which is a useful model for any business getting into AI.

Xero’s motto, as Brody described it, is “fire bullets first, then cannonballs” - try lots of low-effort experiments, then put serious effort only behind the ones that work. Ideas come in from across the business, and the team gives each one three to five days to prove value or get killed early, so they never sink two months into something that doesn’t pay off. The proven ones get handed off to the full development team for proper attention.

On the adoption side, Brody’s team rolled Claude out across the whole business and built systems to drive usage - including a weekly AI leaderboard showing people where they rank, how they compare to other teams, and what they can do to improve (like using more skills or connectors). The real-world tools they’ve built include a copy checker that scores product listings against retailer rules and agent engine optimisation, an ingredients compliance checker that validates products against dozens of international regulatory bodies, and a knowledge assistant that searches internal documents using semantic search before answering. In one case Claude in Chrome reverse-engineered a multi-hour manual research process in under an hour from a single PowerPoint - a great example of pointing AI at the boring, repetitive work first.

Trade Me moves fast too. Georgia was part of the team that shipped the first ChatGPT app in New Zealand, built in a very tight window after OpenAI announced ChatGPT apps - roughly a week and a half after she’d started on the team. Companies running 10% learning time, mandatory in-office collaboration days, and Agile stand-ups give young engineers room to grow while still shipping.

Advice for grads and businesses

For anyone trying to break in, the panel’s advice boils down to a few simple moves:

Pick one tool and go deep. Don’t try to learn every tool. Do a bit of shallow research, pick one that looks interesting, and learn it really well. As Georgia put it, a lot of these tools are “the same thing, just a different font” - the skills transfer.
Don’t wait for the perfect time. The biggest barrier most people hit is simply starting. Once you start, the confidence and momentum follow.
Keep up with the industry. Brody credited a lot of his interview success to genuinely knowing the space - recent models, how they’re trained, job displacement, where things are heading. He keeps up through daily AI podcasts and well-researched newsletters, which is a habit worth building.

And a clear message for businesses - these grads are out there, hungry, and already building real things. Both companies in the conversation were actively hiring for junior AI roles and running intern programmes. Hiring young people who can think in systems and bring the people side as well as the technical side is one of the smartest bets a business can make right now.

Key Takeaways

You don’t need a perfect CS degree to work in AI - curiosity and hands-on building matter more.
Most of the technical AI skills are learned on the job, because formal education hasn’t caught up.
Never let “not knowing everything” be a blocker - nobody knows everything in a field this new.
Think in systems, not just code - anyone who can draw a good flowchart can effectively build now.
The people side is as important as the technical side - learn to translate both ways.
Learn by building a project you’re passionate about, or by solving a real day-to-day problem.
Automate your own learning and anything you do more than three times.
Watch your costs, measure your impact with real numbers, and keep a running log of your mistakes.
Move fast - prove value or kill an idea in days, not months.
Pick one tool, go deep, and just start - don’t wait for the perfect moment.

If you’re a young Kiwi who’s even a little bit curious about AI, take this as your sign to start. Pick a tool, build something you actually care about, and connect with others doing the same - that’s the whole reason I founded Young Kiwis in AI, now 150+ strong. And if you’re a business owner reading this, hire the grads. They’re already building real, valuable things, and they’ll grow with you faster than you’d expect. Come join the community, reach out, and let’s build the future of AI in Aotearoa together. Cheers.