For many years, software development was code-centric.
We wrote requirements, then quickly moved to implementation. Tests helped, but code was the real source of truth.
AI changes this completely.
Today, we are entering the age of ReDevTest. ReDevTest is an AI-era development loop where requirements define intent, development (AI-generated) implements behavior, and acceptance tests verify correctness, with feedback flowing back to the specs instead of the code.
With AI, the most important artifacts are no longer the code. They are:
- requirements
- system use cases
- acceptance tests
Code becomes the result, not the starting point.
Why AI Changes the Focus
AI is very good at generating code. AI is also very good at generating tests and even UI.
What AI cannot do well is guessing intent.
If requirements are unclear, the generated code will be wrong.
If system use cases are vague, behavior will be inconsistent.
If acceptance tests are missing, correctness is undefined.
This shifts the main effort upstream. Clarity before implementation becomes more important than ever.
From TDD to ReDevTest with AI
Classic Test-Driven Development starts with a failing test, then code, then refactoring.
ReDevTest in the age of AI starts earlier.
The flow looks like this:
- Write clear requirements
- Describe behavior as system use cases
- Derive acceptance tests from those use cases
- Let AI generate or update code and tests
- Review the result and refine the specs if needed
The feedback loop still exists. But the first “Red” is no longer a failing unit test. The first “Red” is an unclear or incomplete specification.
System Use Cases as the Center
System use cases describe observable system behavior. They do not describe implementation details.
This makes them ideal for AI-assisted development:
- They are precise
- They are testable
- They are stable even if technology changes
From a good system use case, AI can generate:
- application code
- acceptance tests
- supporting integration and unit tests
When a use case changes, AI updates the code. When the use case is wrong, you fix the use case, not the code.
Acceptance Tests Define Correctness
Acceptance tests answer one simple question:
“When is this use case done?”
They connect requirements with real behavior.
They protect business rules from accidental changes.
They also act as executable documentation.
In an AI-driven workflow, acceptance tests are not optional. They are the contract between humans and machines.
Code Becomes Secondary
This does not mean code is unimportant. It means code is no longer sacred.
Code can be regenerated, update, or replaced
What must stay stable is:
- intent
- behavior
- correctness
Those live in requirements, system use cases, and acceptance tests.
A New Definition of Productivity
In the age of AI, productivity is not measured by lines of code. It is measured by how fast you can go from intent to correct behavior.
Teams that invest in clear specifications move faster. Teams that skip this step spend time fixing AI output instead.
Conclusion
ReDevTest is not about writing more tests. It is about putting specifications back where they belong: at the center. AI makes this unavoidable.
If you want predictable results from AI, you must give it predictable input. That input is requirements, system use cases, and acceptance tests.
