Matt Presland

Vector 01 · 11 min read · · Staying valuable

How I've (somehow) stayed useful through every disruption

From a 747 over the school gate to a needle tuned to 540. Five turning points, and what they taught me about staying useful when the tools keep changing.

I was six when I worked out what I wanted to do with my life.

I was being picked up from Mimosa Primary in Sydney when a Qantas 747 came in low over the school on approach to the airport. I couldn't stop staring. Something that big had no business floating like that, but there it was, hanging over the school gate like it had all the time in the world. From that afternoon, pretty much everything I did was aimed at getting into a cockpit.

I did get there, eventually, though not by any route I'd have planned. I've been a flight instructor, a charter pilot and a Chief Pilot. I've also been a call centre agent, an accidental web developer and a software engineer. Today I'm Head of Technology at Jacaranda Finance and co-founder of Fundamentl, the technology company behind the payments provider SquarePay.

I've been thinking a lot lately about what connects all that, especially now AI is changing what it means to work in software. It definitely isn't planning, because the plan kept getting taken away. As far as I can tell, every time I stayed useful it was because I paid attention to the problem in front of me rather than the tool in my hand.

That feels more important now than at any point in my career, because the tools are about to be handed to everyone.

The plan goes out the window (the first time)

We didn't have the money for flying lessons, so I worked three jobs through high school: a paper round before school, the Woolworths deli after school, and an ice-skating rink on weekends. Through the Air Force Cadets I went solo on 17 December 2003, 100 years to the day after the Wright brothers' first flight. I still wasn't old enough to drive.

In Year 12 I applied to fly for the Defence Force, along with about 10,000 other people. My careers advisor thought I was mad to sit two weeks of flight screening in Tamworth only days before my final exams, and honestly she had a point. But I was one of fewer than 30 pilot candidates selected that year, I still finished school with an OP1, and at 17 I started at ADFA.

I hardly flew there, because flying didn't start until third year. What I did get was a subject called Beginner Programming for Engineers, and a quiet, surprising thought: hang on, I might actually be able to do this.

Then I hurt my knee. Two surgeries and eighteen months of recovery later, I was medically discharged.

I remember driving out of ADFA with everything I owned in the car, heading back to Queensland with no plan, no job and no idea how I'd ever pay for flight training. It was devastating.

What I eventually took from it: the plan wasn't the thing that mattered. I never stopped wanting to fly. I just had to find another way there.

Three binders of blank paper

The other way started somewhere I'd never have guessed.

I needed work to pay for flying. After a stint on the phones doing pay TV support, an email turned up from the bike shop where I'd bought my motorbike. They were "looking for a tech star" to run their online store. I had no experience, told them so, and turned up late to the interview. For reasons I still don't fully understand, their CFO ended it by asking how much they'd need to pay me and when I could start. Later, when I went back to flight school, he asked me to look after the website for a first aid training company he owned, First Aid International.

One day in that office I saw something I couldn't unsee.

They had three training locations, three booking staff and three binders. Each binder was full of blank pages with a date scribbled at the top, and each page was a course. That page was the only record the course existed. To book someone in, you found the right binder, flipped to the right date and wrote their name down.

Staff had to fight over the binders. Pages fell out or went back in the wrong one. Bookings closed three days early so the page could be posted to the trainer. They'd tried hiring more people, and it hadn't helped. They simply couldn't grow past three locations.

Nobody asked me to fix it. I didn't know Microsoft Access, but I figured I could learn enough to try, and I built a rough booking system on the quiet. We ran it alongside the binders as a trial, and within a week nobody was using the binders.

Then the boss asked, "Can we take bookings online?" Access couldn't do that, so I found PHP and taught myself, badly at first, by rebuilding the whole thing. About fifteen minutes after the online form went live, the boss wandered over and asked if I was testing it. I wasn't. Five real people had booked.

Within a year they'd gone from three training locations to seven. When one of the booking staff later resigned, they didn't need to replace her. (In hindsight, I really should have asked for commission on those bookings.)

I didn't think of it this way at the time, but I hadn't learned to code and then gone looking for something to build. I'd bumped into a problem and learned whatever it needed. Most developers I've met started the other way round. I've worked this way ever since.

The board

I did get back into the air: a commercial licence, then instructing, including training pilots for AirAsia. Eventually I was flying freight, mail and passengers across Bass Strait from Flinders Island. People write books about crossing Bass Strait once. Over the years I made more than 5,000 flights across it, through 70-knot winds, bushfire smoke and, once, an engine failure over the water.

Gordon, who owned the business, planned each day on a whiteboard. There was a column per day and a post-it per flight, and he'd stare at it until a plan came to him. That worked at two or three jobs a day. It broke at ten.

Margins were tiny, and every "dead leg" (a flight with no paying job on board) cost real money. So I got into the habit of memorising the board each morning. When Gordon suggested a plan, I'd suggest a cheaper one. It was always cheaper, and I'll admit it was also carefully engineered to give me the flights I liked best. Saving a dollar always won Gordon over.

Often the best fix wasn't clever at all. One phone call asking a customer if they'd mind leaving thirty minutes later could untangle an entire day and save thousands.

Before long I was organising the operation more than Gordon was, and he asked me to be Chief Pilot. When the business changed hands, we grew from 2 aircraft, 2 pilots and 2 bases to 5 aircraft, 6 pilots and 5 bases, and we picked up doctor runs and contracts with the fire service and police. And funnily enough, we never built any custom software. I quoted it more than once, but Google Calendar and a few iPads did the job.

Looking back, the useful thing was never the whiteboard, or any code I might have written to replace it. It was understanding the problem well enough to know what it actually needed, including when the answer was nothing fancy at all.

"Could you give me your problems instead?"

I left Flinders Island in early 2018, hoping to fly bigger aircraft. Then the job market slumped, and COVID grounded the industry. That was the second time the plan went out the window.

So I nervously started applying for developer jobs. I had no commercial experience, I wasn't sure whether to call myself junior or mid-level, and I didn't know Git. An online travel company gave me a go, I think mostly because they were curious about my strange career. Their lead developer reviewed my code patiently and pushed me to do better, and for the first time someone who really knew the job told me my self-taught instincts were sound.

COVID took that job too. A recruiter went out on a limb for me. He turned down a lowball offer on my behalf and then got me in the door at Jacaranda Finance, where I became developer number three. Back then the tech was a loan application form, and it didn't even store a customer record.

In my first few weeks, the CTO was writing me tickets so detailed it would have been quicker for him to write the code. I'm not sure where the nerve came from, but I said to him, "I'm not really a task jockey. I like solving problems. Could you give me your problems instead, and let me see if I can work them out?"

Thankfully, he did. The first one became a loan assessment system with job queues, escalations, and a way to wait on customer documents. I built it with Luke, the UI developer from the travel company, in about three months. Six years later it's still running in its original form, and it now has AI automation built into it. Over those years I've worked on loan scheduling, forecasting, credit risk decisioning and marketing automation, and in July this year I became Head of Technology.

I think about that conversation a lot now. Turning a detailed ticket into code is exactly the kind of work AI is getting good at. Working out what the problem actually is still feels like our job, and I think the person who asks for the problem is the one who gets to direct the AI.

What the GPS can't do

I'm not anti-AI. Far from it. I build AI into credit decisioning for a living, and I'm building VerdictVector so businesses in regulated industries can use AI and still defend every decision.

But I've also seen where it fits and where it doesn't. In 2022, Brent and I co-founded Fundamentl, SquarePay's technology arm, to rebuild his payments business from an empty repository, by hand, before AI coding tools were really a thing. AI would probably have made progress look faster, but I think it would have boxed us in, because we were engineering a platform that had to stretch in ways we couldn't predict yet. We cut over on the Easter long weekend in 2023, with a complete rebuild and a big new feature set. I'll never forget 1am on the first business day back, watching our first big client's payments come through. Not one dropped. The business has grown about five-fold since.

That didn't come from typing quickly. It came from understanding the current state, the future state, and every step in the gap between them.

My hunch is that AI will make the tool side of software close to free: writing code, turning tickets into features, the boilerplate. That's genuinely exciting. But pilots learned a long time ago that automation is wonderful right up until it isn't, and when it isn't, you want someone in the cockpit who still knows how to fly.

I don't think the people who'll be fine are the fastest typists or the cleverest prompters. I think they'll be the ones who never lose sight of the problem.

540

I'll leave you with a little trick from my Bass Strait days.

Aircraft have a navigation instrument called an ADF. Its needle points at a radio beacon, but it'll point at a normal AM radio station too. If you draw a line from Flinders Island through our base at Bridport and keep going, you reach a town called Scottsdale, which has an AM station on 540.

Gordon always had 540 tuned in on that route. Partly it was so we could listen to the local radio. But it also meant that on the worst days, with weather closing in over the water and the fancy kit playing up, the needle still pointed home.

There's an old rule in aviation: use every tool you've got, and set it up before you need it.

The tools in my career have kept changing: aeroplanes, whiteboards, binders, PHP, and now AI. What's got me home each time, as far as I can tell, is keeping the needle pointed at the problem.

That's what I'll be writing about here. I hope it's useful.