AI added to the systems you can’t switch off.
We map and test the old system first, then change it in stages with tests that prove behaviour stays the same. No big-bang rewrite.
What you get
Change you can prove.
A map of the old system
What it does, where the rules live and which parts nobody has touched in years, written down.
Characterisation tests
Tests that record what the system does today, so any change that alters behaviour shows up.
Staged rollout
One part at a time, each behind the tests, with a way back at every stage.
An AI layer on top
Search, drafting or review added around the old system, without changing its core first.
Documentation
The rules we find, written down as we go, so the knowledge outlives the project.
A real handover
Code, accounts, secrets and build steps your team can use without us.
How it works
Map, pin down, then change.
01
Map
Code mapping and interviews with the people who run it.
02
Pin down
Characterisation tests around the parts we will touch.
03
Add
An AI layer around the system, using its existing interfaces.
04
Change in stages
Each change behind the tests, released gradually, with rollback ready.
05
Hand over
Tests, documentation and a path to production your team owns.
Questions
What engineering teams ask.
Do we have to rewrite the system?
No. We start by adding AI around it, and change the core only in tested stages, if at all.
How do you know a change didn’t break anything?
Characterisation tests record today’s behaviour first. A change that alters it fails the tests before it ships.
Nobody here understands the old code. Is that a problem?
It’s common. Mapping it, and writing down the rules we find, is the first part of the work.
Who owns it afterwards?
You do. The handover covers code, accounts, secrets and build steps, so a change can reach production without us.
What we don’t promise
Some claims we won’t make.
- A fixed date for replacing a system we haven’t mapped yet.
- That every legacy system should be changed. Sometimes the right answer is a layer on top.
Keep it running. Make it better.
Tell us about the system. We’ll tell you where AI could help first.