Why wouldn't we just have Copilot write the plugin?

You could, and for some things you should. But writing the code was never the expensive part. Owning it is. Here's the honest comparison.

The bill arrives later

Day one
It works
Copilot writes the plugin. It works. This is the cheap part, and the only part anyone budgets for.
Month six
It waits
The rate changes on a Thursday. The person who wrote it has moved teams, so the change waits for a sprint.
The next platform update
It breaks quietly
A deprecated call breaks it. Nothing failed loudly: the process just stops doing what it did.
Year three
Nobody touches it
It is code nobody remembers writing, and the safest thing anyone can do with it is leave it alone.

North52 rules vs custom code

Comparison of North52 rules versus Custom code
Dimension	North52 rules	Custom code
Who changes the logic	Analysts, in minutes	Developers, plus a redeploy
Audit trail	Built-in trace and versioning	Git, if discipline holds
Deterministic outcomes	Guaranteed	Depends on review quality
Regression testing	TestShield included	You write and maintain it
Platform updates	Engine kept compatible for you	Code can silently break
Where it runs	Inside your tenant, always	Varies by architecture
One-off data migration	A rule you will never run again	Write it, run it, delete it

How the rules engine works →How TestShield works →

Where custom code still wins

Extreme performance edge cases, exotic integrations, logic that will genuinely never change, and the project with no budget for a licence. We'll tell you when that's your situation.

The questions that decide it

Who changes it when the rate changes on Thursday?

With code: a developer, a review, a release. With rules: the analyst who owns the process, in minutes, with the change logged.

What happens when the person who wrote it leaves?

Code inherits an owner who did not choose it. A decision table is readable by whoever picks it up, because it was written to be read.

Can you show an auditor why it decided that?

Code can show you the source as it is today. The engine shows which rule ran, which rows matched, and what changed.

Isn't a rules engine just code by another name?

No. A plugin is a program written for one job. The engine is configuration: your data goes in, the rules you configured run against it, and the actions come out. When the rule changes, you change the rule. Nothing gets rebuilt.

So what is the licence paying for?

Pedigree. The engine has been through fifteen years of Microsoft platform updates in 300+ organizations, and not one of them raised an issue we could not get past. A new plugin starts that clock at zero.

Couldn't Copilot just write us a rules engine?

Go for it. What comes back is the plugin problem on this page, multiplied: an engine, a designer, a trace, versioning and a test harness, all of it yours to own through every platform update, with the clock at zero.

What the data says

8×

rise in duplicated code blocks across a 211-million-line codebase analysis, as refactoring fell to historic lows.

GitClear, 2024 codebase analysis

89.3%

of issues introduced by AI coding assistants were maintainability code smells, across 6,299 repositories.

2026 empirical study

Bring us your hardest rule.

Book a demo, send us your hardest rule, or take it for a test drive.

With our team

Personalized demo

30 minutes on your rules, not our slides.

We build your use case

Free proof of concept

Send us the rule you would least like to hand to a developer who leaves.

Request a free PoC

Download and go

Free trial

Full product, your own environment.

Sources

GitClear, 2024 codebase analysis of 211 million lines of code.
2026 empirical study across 6,299 repositories, on issues introduced by AI coding assistants.