A logistics platform, rebuilt in six months.
From final mile to the whole chain.
A UK carrier asked us for a developer, years before AI could write code. This year, their CTO asked us to lead the rebuild of their entire platform.
They needed a developer. We were that developer, for years.
The carrier delivers parcels across the UK. When we met them, they already had a mature system. What they needed was a hand with full-stack work: new features, fixes, maintenance.
That system only covered the final mile. Everything before and around it, from the middle mile to pick-up points and lockers, was handled elsewhere.
Trust came from moments like this one: an incident stopped operations at one of their warehouses, and we built an app in seven days that got the warehouse working again.
Meanwhile, we changed. They noticed.
Our freelance work grew into a company. Around the same time, AI became useful for writing software, and we used it from the first day. Our clients saw the speed and the quality go up. This one included.
Their CTO started trusting our lead on how far AI could push a small team. So together, we decided to rebuild the whole system from scratch.
Not a new final mile. The whole chain.
Middle mile, final mile, pick-up and drop-off points, lockers, several packaging options. One platform, with an operations console and a driver app, built for 100 times the volume the current system handles.
Three people built it. AI wrote most of the code.
We planned it with their CTO: he owned the business side, we owned the engineering and the AI setup. Then he joined the build, with one of their developers.
We brought what we had learned building this way on our other projects. AI writes almost all of the code. Every change goes through automated tests, several layers of AI review, and a person before it ships.
What we brought from our own work.
| What we tried | Verdict | Why |
|---|---|---|
| Tests, then layers of AI review, then a person | Kept | Speed never skips the checks. |
Six months for what would have taken years. Live in December.
The first end-to-end tests have started, and they look promising. We’ll add real numbers here after launch.