)
)
TL;DR
We went AI native (get the full story here) and some people whose jobs were never “building”, started shipping tech. We sat down with David, one of our Product Designers, and Joe, one of our Senior Product Owners, to hear what it's actually like now design and product teams are hands-on with the code.
One went in scared. One went in sceptical. Both did a full 180. They've built real projects way faster than the old way ever allowed! But AI still gets things wrong in sneaky ways. It'll pass the first test and be broken underneath, cheerlead a flawed idea, or quietly ignore your design system. Every single time, it took a human reading the actual code to catch it.
AI makes building faster, but human expertise is still key for products and projects that really work.
Binge the rest of the series:
Who says you can’t teach old dogs new tricks?
When Duncan kicked off our factory rebuild back in April, it opened up a world of possibilities and changed the established roles within Cube.
So, introducing David and Joe to give us the lowdown from their experience! David is one of our Product Designers who's been at Cube for nearly five years, and Joe is one of our Senior Product Owners who's clocked four and a half years here.
Between them they've lived through every pivot Cube's ever made and everything in-between. It’s exactly why their reaction to our Factory Rebuild is worth listening to. This was a crash course into a whole new process.
One went in nervous. The other didn't buy it at all.
Two reactions, two very different starting points
David felt nervous at first, and he's honest about why.
"I was a bit scared because it's something that you see in the news, you don't know how it works. I have no coding experience. And for me it was like okay, now I'm asked to ship features, to get involved in the code."David Torres, Product Designer, 3 Sided Cube
The fear didn't last, because the first time he actually used AI and our VPS (Virtual Private Server), something flipped. He felt how much more fun the process was becoming! The delighters, the transitions, the little animations that make an experience feel alive? Those are the first things scoped out of a build when time and budget get tight. Now he could mock them up himself and watch them work earlier on in the process. Almost overnight, scary became exciting.
Then there’s Joe's starting point. He wasn’t scared, he was sceptical. And he wasn’t shy about it either.
"I think I was probably one of the more vocally sceptical people at Cube about AI and how we're gonna implement it."Joe Tattersall, Senior Product Owner, 3 Sided Cube
His whole job as a Product Owner is to guard scope and quality. So when AI is talked about as a way to get more scope in, or hit smaller budgets faster, his instinct was to push back hard. He didn't want to sleepwalk into shipping rubbish just because a machine built it quickly. It took him longer than most to come round, but he eventually did!
"You've got to have a go, you've got to see. You can't just be a kind of ‘AI Philistine’ forever. You've got to get stuck in and it has been a real game changer. I’ve gone that full 180 on being a skeptic to probably being one of the more invested."Joe Tattersall, Senior Product Owner, 3 Sided Cube
Joe and David’s roles work in perfect harmony. Joe's scope-guarding instinct is exactly what stops AI-speed turning into AI-slop, and then David's design discipline is exactly what makes the AI build well in the first place.
)
The proof is in the projects
Talk is cheap. So what actually convinced them was kicking AI into action.
First up: a mobile app under pressure
For Joe, it was one of our mobile app projects. New account creation, a load of maps work, real budget, and timeline pressure. The perfect requirements for an experiment. How much of the scope we'd already planned the traditional way could we actually get built if AI did the coding?
Turns out, it was loads! But there was one condition… good source in, good code out.
Joe talked about an experience one of our other developers had on the project. The initial code was built much faster using AI, but the devs knew it was because they had really good Figma files, really good design files, and a really good design set up. All of that allowed them to prompt the AI effectively
Next: The climate news app that changed how David works
That lesson about putting a good source in pointed straight back at David, who started rethinking how he structures a file in the first place.
"But now we're giving these files to AI to help us with the implementation. And it has impacted the way I structured the file, for example, how we name things."David Torres, Product Designer, 3 Sided Cube
On a climate news app, he started fresh. Named the design system the way the code would name it, so the AI could easily understand what he meant instead of guessing. Fewer iterations. Faster to a working prototype.
"Design and development is now closer than ever in the sense that we need to constantly be collaborating."David Torres, Product Designer, 3 Sided Cube
And this is where the two worlds collide in the best way. When you can see the thing working early, the conversations change. Design, product, development, quality assurance, everyone gets a look while there's still time to change things.
But then: The very human catch
Of course there’s a catch! This stuff is addictive. The satisfaction of watching an idea turn into a real screen makes it very hard to step away from the machine. Joe and David are both dealing with the same pull.
"You have to remind yourself to create stopping points, and get up from your desk, and go and get some fresh air, because it's so satisfying to be able to see the idea turn into a reality."Joe Tattersall, Senior Product Owner, 3 Sided Cube
"I'm still learning when to stop because sometimes it's too exciting, like I'm gonna keep trying."David Torres, Product Designer, 3 Sided Cube
)
Shift left, or ship slop
The National Bureau of Economic Research (NBER) tracked more than 100,000 developers using AI tools. Code output went up as much as 180%. Cool, right? Until you follow that code towards an actual finished thing and it bleeds out at every step. NBER went on to report that the 180% more code leads to 50% more finished projects and only 30% more things that actually ship!
Which is exactly what Joe means when he talks about "shift left." Ask AI to do everything and you get a whole lot of slop.
"There's a lot of people out there thinking if you just went and said, 'Hey build me this website, here's my idea for it’, you get slop. Which of course you do. It's gonna guess most of it. And it's probably gonna guess most of it wrong, and it's certainly not gonna build it in a way that's particularly secure."Joe Tattersall, Senior Product Owner, 3 Sided Cube
The move isn't to hand the whole process to a machine. It's to use AI as a tool that makes experts faster, and then keep the expert judgement exactly where it's always been. We absolutely still need expert humans in the room because otherwise, you get stuck in the trap of AI marking its own homework.
If you let AI scope the work, design the work, build the work, and check the work, there's no one left to catch it being wrong.
When it looked perfect and wasn't
For all the wins, AI gets things wrong in ways that are pretty sneaky.
The accessibility fix that passed the first test.
David was working on a site where accessibility really wasn't optional. He spotted a keyboard navigation issue, got AI to fix it, tested it, and it looked spot on.
"I raised my PR, and then it was time for a developer to check it. And to my surprise, even though it looked like it was working, the developer flagged that the implementation was not the right one."David Torres, Product Designer, 3 Sided Cube
The ‘fix’ worked on the surface but it was wrong underneath. Only a developer reading the actual code caught that AI wasn’t following best practices.
The analytics job Joe was very proud of
Joe was building analytics tracking that deliberately avoided storing anything personal. No personally identifiable information, no cookies, all handled page by page. He was pleased with himself and so was Claude.
"Claude was ‘Yeah, you've you've smashed this, this is exactly how you should do it’. And I sent my PR, and one of our developers came back not that long later, and was like ‘This is doing the opposite of what you want’."Joe Tattersall, Senior Product Owner, 3 Sided Cube
AI is eager to please so it'll encourage your idea when what you actually needed was someone to tell you it was flawed. Gotta love Claude for being our own personal cheerleader, but there’s a time and a place.
The component that wouldn't behave
Design survives on reusing the same components across a system. And AI, for all its cleverness, keeps trying to reinvent them!
"Even if I tell the AI, please reuse this component that we've built, sometimes it does not do it. So it really needs that human aspect to review that the implementation is correct."David Torres, Product Designer, 3 Sided Cube
)
The all important question
We asked Joe and David if they’d ever go back to the way we used to work, without AI in the mix. The short answer from both of them was ‘no’.
"Now with AI I can see the code. I can even create my own code. And that's something that I couldn't do before. The reward that I feel every time I design something and then I see it come to life without the need of a developer, at least for a prototype, it just feels really cool."David Torres, Product Designer, 3 Sided Cube
Joe went further. Even if he wanted to go back, he doesn't think he could.
"That bit of everyone being much more actively involved, it will lead to better quality. Provided we keep those quality gates in, you'll make better products with AI."Joe Tattersall, Senior Product Owner, 3 Sided Cube
So, that's it! Two old dogs learning a whole bunch of new AI tricks and not looking back.
Thanks to our Factory Rebuild, more people who care are closer to the process than they've ever been, which is a pretty damn awesome place to be 💚
Ready for MORE factory rebuild?
Hear from more of our CUbes about how the factory rebuild has been for them:
Disclaimer: We used AI in the process of writing this blog, but a human was still at the centre! We'd be poor advocates for going AI native if we didn't use AI to help write about it. The story is ours, the quotes are our team's own words, and every line got the human pass.
Published on 21 August 2026, last updated on 21 August 2026