Every startup carries tech debt.
Not because teams don’t care — but because speed, uncertainty, and evolving requirements make it inevitable. Most of it never gets “paid down”, and full rewrites are usually too risky to attempt.
AI doesn’t remove that reality.
But it does change how we can move forward without being trapped by it.
But it does change how we can move forward without being trapped by it.
This pattern isn’t new — the leverage is
Long before AI, people like Martin Fowler described incremental ways of dealing with legacy systems — most notably the Strangler Fig pattern — where you build new functionality at the edges, introduce clear seams, and extract only the behaviour that still provides value.
These ideas have lasted because they match how real businesses operate.
What’s changed is the cost of applying them.
What’s changed is the cost of applying them.
AI lowers the friction that used to stall this approach:
- spinning up clean, greenfield systems is faster
- understanding legacy behaviour is easier
- porting logic doesn’t require dragging old architecture along with it
The principles are the same.
The economics are different.
The economics are different.
Greenfield at the edges
Our approach to tech debt is intentionally pragmatic:
- we don’t try to clean the monolith
- we don’t freeze delivery to refactor
- we don’t pretend the debt will disappear
Instead:
- we build new, greenfield systems
- we introduce a clear layer between new and old
- we extract only what still provides value
- and we leave the rest alone — deliberately
One example of this was rebuilding our order status experience.
We built a clean, greenfield implementation, mapped its core components to equivalent behaviour in the legacy system, and used AI to help locate and port the relevant logic. The old system stayed where it was. The new one stayed clean.
This wasn’t a rewrite — it was selective extraction.
Why AI fits this approach so well
AI works best when boundaries are clear.
AI works best when boundaries are clear.
Once a clean surface exists, it becomes very effective at:
- identifying relevant legacy behaviour
- understanding intent
- helping move what matters across the seam
The human judgement still comes first: what to keep, where the boundary lives, what’s no longer worth carrying forward. AI accelerates the work — it doesn’t replace the decisions.
Old ideas, new leverage
Companies like Shopify have long taken a pragmatic approach to system evolution, showing that large systems can be incrementally reshaped without risky rewrites by focusing on seams, boundaries, and developer productivity. Deconstructing the Monolith: Designing Software that Maximizes Developer Productivity
Companies like Shopify have long taken a pragmatic approach to system evolution, showing that large systems can be incrementally reshaped without risky rewrites by focusing on seams, boundaries, and developer productivity. Deconstructing the Monolith: Designing Software that Maximizes Developer Productivity
AI doesn’t change that mindset.
It simply makes it easier to execute.
It simply makes it easier to execute.
AI hasn’t replaced the hard thinking.
It has made good architectural patterns cheaper to apply.
It has made good architectural patterns cheaper to apply.
Old ideas.
New leverage.
New leverage.