)
)
TL;DR
Every AI blog on the internet ends the same way: you start using AI, work gets faster, roll credits. We’re here to tell the FULL story with what actually happened when our delivery team went AI-native.
Because the plot twist nobody tells you is that AI didn't make the work vanish, it just moves it. More time refining, more time for set ups, more time managing AI outputs.
We caught up with Jess, our Head of Delivery, and Tom, one of our Project Managers, for the inside scoop. Where the hours really go, the tips they swear by, and why neither of them would go back to how it was in March 2026.
Binge the rest of the series:
The version nobody publishes
There’s about a million versions of this blog that pretty much everybody on the internet has already seen.
You start using AI, your work gets faster, you celebrate. The end. But what you don’t know is that those million versions actually skip a whole lot of the story.
All the plot twists, flashbacks, and character development.
Fear not! Your fave oversharing AI-native agency is here with the full story, all the inside scoop, of the good, the bad and the ugly of our going AI AF journey.
We caught up with Jess, our Head of Delivery, and Tom, one of our Project Managers, to find out what actually happened. So get ready to learn where the time really goes when it comes to AI-native projects, why the boring bits got even more important, and what our Factory Rebuild has actually been like for our delivery team.
So this is the honest one. Two of our Project Managers, Jess and Tom, on what changed in the day job when Cube went AI first. Not the pitch. The actual texture. Where the time went, why the boring bits got more important, and what it's genuinely like wrangling this stuff at 5pm on a Tuesday.
)
Where it all began…
A long long time ago (aka April 2026) in the land of 3 Sided Cube, our CEO Duncan threw down the gauntlet. We're going AI native!
News spreads across the land (aka our office, Slack, and Google Meets). Some Cubes feel excitement, others were shooketh.
Over in Project Management town, Jess, our Head of Delivery, went where she always goes, straight into problem-solver mode! Her first worry wasn't the tech, it was the team.
How will people react to this? Are we going to have people on board? Are we going to have people that don't want to do it and who are going to be very adverse to the change? All of which are very realistic questions when you’re the one in charge of leading the changes to our process.
“We have lots of developers of all ranges who welcome change and who don't. So I was mostly, kind of, trying to prepare myself for what kind of reaction they would have. Because we were completely flipping how the process was going to work on a project.”Jess Vaughan, Head of Delivery, 3 Sided Cube
Then the penny dropped for what this meant for her own job.
The possibilities were endless. Less manual reporting, faster communications, answers she could find herself instead of waiting on a developer to build her a magic tool!
“I don't need to rely on the developers to build me a magical tool to do this thing. I can now try and do it myself and fix these problems myself.”Jess Vaughan, Head of Delivery, 3 Sided Cube
Classic eldest-daughter energy, and Jess will happily own that!
Then, Tom's journey was a bit… different. He'll tell you himself that he pulls a face "like a slapped ass" whenever a big change gets announced. He wasn’t exactly nervous, he just wasn’t going to use AI for the sake of it. Tom's still a bit sceptical now, and that hasn't been polished out of him for the sake of a nice blog ending, we promise!
Pep talks about AI didn’t cut it. Instead, what really made the difference was seeing a tool that the Project Management team dreamt of come to life!
“From that point is when I've kind of accelerated what I'm able to do with AI and build tools internally. And that was really cool. And I think I've become a lot more on board because I could kind of see all these efficiencies that we could have.”Tom Trigg, Project Manager, 3 Sided Cube
The scheduler became something he could actually build with. Once he could pull sprint time into a chart instead of validating it by hand every planning meeting, he got more on board. He could seriously see the efficiencies, not hear someone telling him about them.
Same change in the land of 3 Sided Cube. Two fairly different reactions. But actually closer than they look: two people who needed proof, got it in different ways, and still keep their eyes open about the bits that don't work.
The first thing we built for ourselves
Aka, our AI-native origin story
There was a platform the whole team touched every single week. Then the vendor tripled the price overnight.
Our reaction was simple: nu-uh!
So Duncan, while on holiday, up at stupid o'clock, started building a scheduler. Two or three days in and the amount he'd done was frankly a bit mad. But there was a stopping point. Although he was really familiar with the scheduler, Jess was the one who uses it the most. So Duncan handed it over.
She started tweaking it herself, in her words, going ‘slightly crazy’ on it for a couple of weeks. Late nights. Weekends. The instant gratification of asking for a thing and seeing the thing almost instantly, instead of waiting for the reveal, was a HUGE incentive for Jess to keep going.
“There were too many late nights and I may have started working on it at weekends. It was just, it was so addictive to start with”Jess Vaughan, Head of Delivery, 3 Sided Cube
(Side note: we don’t encourage this, no matter how cool it is seeing the tools! See our Head of People’s take for how we’re now managing our approach to AI so we all get a good nights sleep 💚)
Now the scheduler could do things it never could before: Deadline checker at the click of a button, colour coded projects, manual reformatting of spreadsheets, super speedy reporting, deadlines organised in the right date order and with the right emojis to categorise a release. A scheduler that fits exactly how we work at Cube.
It might seem like ‘small’ stuff, but Jess reckons it's saved her at least half a day of admin a week!
)
The plot thickens
Oh, it sure does! These are our character defining moments, the tips, tricks, and things we’ve been through so you don’t have to:
Skip the setup, ruin the ending
The boring admin matters MORE now. You'd think AI would let you skip the unglamorous scoping? Nope.
We’ve already talked about the importance of sharing a project’s history, and it’s pretty much the same when you’re working on a project in the present.
Every bit of discovery, every immersion session, every ticket listed out with the whole team in a room, that's the instruction set you're handing the AI. Cut corners and everything will go downhill from there. Generic and riddled with assumptions!
“If you trust [AI] too much to go too far, it will make loads of assumptions for you. Whereas actually, you being in that meeting, making sure you ask those things and getting the clarification from the client is even more important at the start.”Jess Vaughan, Head of Delivery, 3 Sided Cube
Skip the set up and there’s a whole long list of problems: nothing to manage, missing scope, duplicating scope, spawns new tickets to cover the gap, and suddenly there's a chaotic project board you're trying to rein back in.
It’s the same trap with documents and generating tickets. Let AI run and you get a beautiful 45-page scope doc that nobody reads because it's 45 pages. Ask for ‘relevant accessibility criteria’ in a ticket and it dumps the entire WCAG double-A spec when it only ever needed colour contrast and text resizing.
A well-thought out set up means you avoid the hiccups later!
AI loves a blank page. It hates a backstory
Give AI a blank page and it is in heaven. New project, no baggage, defined requirements, working with you instead of playing catch up. It’s like an author writing the first draft, full of ideas and creativity.
What AI hates doesn’t love is anything with history behind it.
On a project with years of context, AI is actually worse to work with. It’s not AI’s fault, you just forget how much information lives in a developer's head (or any other member of the team) that was never written down anywhere. Years of ‘we did it this way for a reason’ decisions, the odd ‘following you gut’, all of them load-bearing.
AI can't read what was never written so it fills the gaps by assuming. It thinks it's being clever, quietly ‘fixing’ something that was that way on purpose, and off it goes.
“It thinks it's doing a great job. And actually you're like, no, please stop ruining it. but with new projects, like you are defining all of those requirements at that time.”Jess Vaughan, Head of Delivery, 3 Sided Cube
Now, PMs become the context provider. The job stops being ‘manage the build’ and also includes feeding AI the history, structuring the data properly, catching it before it goes off piste.
AI only ever knows what you tell it!
)
We finished a project in a week, then came a plot twist
For another project, Tom and the team went all in on AI. Within a week, the app was testable and had pretty much all the functionality it needed. No one expected that level of speed!
Then the actual shape of the work showed up with more bugs than you'd normally expect at that stage.
And more bugs means more time with the team to fix said bugs.
“It's been a kind of double-edged sword in the sense that you get a project or a product quicker, it's probably just slightly less better quality than it would have been had it been built from the ground up.”Tom Trigg, Project Manager, 3 Sided Cube
You get the product quicker, but you also need time to get it polished enough to hand over and still be proud of. So, the dev time didn't disappear, it was redistributed into Project Management, Design, and Quality Assurance.
Is it still faster to use AI and get a good result overall? Yes. It's the fine-tuning that eats into the saving.
And there are other perks too! By having that time to get a project polished, it means the team likely has time to work through the low priority tickets that used to stay in the backlog until the next sprint.
The favicon (a pixel image that represents a website) with the annoying white corners Tom had been ignoring for ages because it wasn't worth an hour of a dev's time? Fixed. The design team's tiny padding niggles, fixed without raising a ticket. The hard-coded copy that PMs used to draft for a developer to paste in, changed directly, no dev required.
Give the AI its character notes
Ask AI to look at a small bug without any context and it'll hand you eight paragraphs of technical waffle about what it's attempting. Not great when you’re not a developer and don’t have time to read and research the meaning of every sentence.
“The amount of stuff it gives you back when you put in a prompt is insane. I've had to start being really strict with the instructions.”Tom Trigg, Project Manager, 3 Sided Cube
So instead, something like ‘I'm not a developer, give me three bullets on what you did, or if you can't do it, say so and stop’ give you a much more focused and useful result! Narrow, specific, no performative cleverness.
Watch out for rabbit holes
Jess had the perfect example. She wanted a headshot sticker in our Cube styling, the ones in our team section. She asked AI to make her eyes green so, of course, the sticker ended up with a green streak through her hair and the whole illustration restyled into something realistic and completely off-brand.
“You do have to pay attention because it will start going completely down the wrong rabbit hole and you're like, whoa, whoa, whoa, whoa.”Jess Vaughan, Head of Delivery, 3 Sided Cube
You have to watch it, catch it, stop it, and hand back the clarification, or you can easily lose an afternoon.
)
Don't fall for the cliffhanger
All the context-switching, the click-and-wait, the "just one more change", keeps you on the edge of your seat! But don’t fall for it and lose hours. Jess knows she can fall into this herself. A binge-watcher by nature and so excited by a project once she's in it. Her solution was to set herself rules.
First, one AI project at a time, no exceptions. Then, a wrap-up ritual where she tells the AI she’s done for the day, to put everything in the backlog, prioritise it, stop suggesting things, and she’ll come back tomorrow. Because if you keep checking and waiting for the responses, you'll always find one more thing.
Not every idea makes the final edit
Last but certainly not least is knowing when to abandon your AI idea. Tom's lived the experience with a sprint-comparison dashboard he built. After getting it working, he admitted it wasn't actually doing what he needed, so he parked it.
“It's cool that I've got this proof of concept, but like knowing when to abandon something when it's not working as well for you as you want it to, I think is also important.”Tom Trigg, Project Manager, 3 Sided Cube
That might sound small, but it's the guardrail against the real mess of having twelve clever tools all half-built, all doing nearly the same thing!
)
So, would either of them rewrite the script?
Would they go back to how it was in March, before the AI-native announcement?
“Probably no. I'd do it, it would be boring. [Now] there's so much more excitement and enjoyment that you can get with building something that's personalised to how we work.”Jess Vaughan, Head of Delivery, 3 Sided Cube
Now Jess know’s what’s possible, she doesn’t want to go back to reformatting spreadsheets, manually tracking project deadlines every week, and bottling up every "why won't it just do this" frustration.
Tom’s in a very similar, but not quite the same, boat. The main perk he’s seen is just how much more involved he is with the projects, instead of tracking it from the outside. In the past, Tom had a moment where someone asked if he was proud of a product and he realised he couldn't really say yes, because building it had nothing to do with him. Now, on one project, he ran most of the UAT amends because it was largely copy updates.
“Stuff like that where you're that much more involved is pretty cool and is a nice place to be. It would be a struggle to go back.”Tom Trigg, Project Manager, 3 Sided Cube
The challenge is why they both chose agency in the first place: the variety, the rate of change, the problem to solve. Our AI-native shift just turned all of that up to 1000!
There’s even more to the story
Our Factory Rebuild doesn’t end here, it’s the series that just keeps giving! Check out how the factory rebuild has gone across the land of 3 Sided Cube:
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 14 August 2026, last updated on 14 August 2026