Contact Us
Menu
What we do
Back What we do
Back What we do
Industries
Back What we do
Capabilities

Case Study · Banking × Linkroad

Application modernization: AI-driven requirements & code generation.

A leading banking institution in Portugal needed to modernize a legacy application for regulatory information duties without losing embedded business knowledge or disrupting operations.

The challenge: extract requirements from scattered documentation, legacy code, and usage videos; generate modern code reliably; and automate testing — with full traceability throughout.

≥90%

of requirements extracted automatically

80%

of new code AI-generated, zero manual lines

≥90%

test coverage, frontend and backend

182

requirements, 155 business rules, 18 features

Gallery
Requirements pipeline
Screen catalog
Generated application

Challenge

A legacy application. Knowledge scattered across documents, code and videos. Operations that cannot stop.

The application supported regulatory information duties, and the rules that governed it lived in three places at once: outdated documentation, the legacy code itself, and the way people actually used the screens.

Rewriting it by hand meant re-discovering that knowledge manually, requirement by requirement, and testing a new system against an old one nobody had fully specified.

You cannot regenerate an application you have never fully written down.

The case

Requirement-to-code traceability

The objective was never the code. It was an industrialized way to modernize.

Prove that the AMA framework could industrialize modernization — delivering a requirements-first, AI-driven pipeline that generates code requirement-by-requirement and validates it automatically.

Requirements before code

An automated pipeline extracted and validated requirements from documentation, usage videos and legacy code, and converted usage videos into structured screen catalogs — 182 requirements, 155 business rules and 18 features, written down before a single line was generated.

Generation with traceability

Modern frontend and backend were generated automatically, requirement by requirement, with zero manual lines of code — each requirement traceable to the code that implements it.

A requirement-by-requirement cycle turns generation into something you can audit.

Testing followed the same logic: AI agents with Playwright MCP validated requirements against the legacy application and generated test cases reaching ≥90% coverage on both frontend and backend.

Banking sector imagery

About the client

A leading banking institution in Portugal

182

requirements extracted, validated and generated into working code.

A retail and corporate banking institution operating a regulated technology estate at national scale. Client name withheld under confidentiality.

Impact

What the pipeline delivered

01

Requirements recovered automatically

≥90% of requirements extracted automatically from documentation, videos and legacy code.

02

Code generated, not hand-written

80% of new code generated via the AI-driven requirement-by-requirement cycle.

Zero manual lines.

03

Tested end to end

≥90% test coverage on frontend and backend, with full traceability from requirements to code.

Solution · AI-driven modernization pipeline

From legacy artefacts to generated, tested software.

Four stages took the application from scattered knowledge to a modern, validated system: a requirements pipeline, a video-to-screen catalog, requirement-by-requirement generation, and automated test generation.

Extract, catalog, generate, validate.

01

Automated requirements pipeline

Extracts and validates requirements from documents, videos, and legacy code.

02

Video-to-screen catalog

Converts usage videos into structured screen catalogs.

03

Requirement-by-requirement code generation

Modern frontend and backend generated automatically, zero manual lines of code.

04

Automated test generation

AI agents with Playwright MCP validate requirements against the legacy app and generate test cases with ≥90% coverage.

Technology and approach

Platform: AMA (Application Modernization Approach) Framework.

AI: Multi-agent system, Gemini API, MCP + Playwright agents.

Scope: Legacy application — 182 requirements, 155 business rules, 18 features.

Approach: Requirements-first, AI-driven, zero manual code.

A legacy application nobody wants to rewrite by hand? Start with the requirements.

We can walk through how an automated requirements pipeline, requirement-by-requirement code generation and AI-generated tests would work on your own legacy estate.