Legacy modernization with AI

Systematically reduce technical debt.

Altanwendungen laufen stabil und binden trotzdem Budget. Die Dokumentation ist Jahre alt, die Fachlogik steckt im Code, und die Entwickler, die das System kennen, gehen in Rente. Ein Neubau derselben Funktionalität kostet Mega-Budgets und schafft wenig Mehrwert. Wir setzen KI genau dort ein, wo bisher der Aufwand lag: Analyse, Spezifikation, Testerstellung, Code-Transformation. Erfahrene Entwickler entscheiden über Architektur und Qualität. Am Ende steht produktive Software.

Is your business logic embedded in code that only a few people still understand?

Are you missing an up-to-date specification for your legacy applications?

Does rebuilding the same functionality blow every budget?

Use Cases

Custom-developed applications with incomplete documentation

The application runs, but the specification is nowhere to be found. We analyze source code, existing documents, and the running system and use them to create a domain model, use cases, and calculation rules. Your application owners review the results in short review sessions.

Monoliths in .NET, Java, and PHP

Mature monoliths bundle ten business domains into a single codebase. We split them into bounded contexts, define a layered target architecture, and map custom-built solutions such as proprietary state machines back to standards like BPMN.

Technologies reaching end of support, mainframe and host systems

When support and patches are coming to an end, the date matters. We measure the functional scope, plan the target architecture, and define the date for migration and decommissioning of the legacy system.

Portfolios with 100 or more legacy applications

We use AI to inventory the entire portfolio and cluster it by criticality, technical condition, and complexity. This results in a roadmap organized into waves across four action areas: remediate, retain and maintain, replace, and monitor. Each wave delivers modernized applications and blueprints for the next.

A case study: 65 person-days instead of more than 500

A central ERP and financial system, grown over 15 years, a PHP monolith, ten business domains, and around 200 elementary processes. We measured it, reconstructed the business logic, and migrated it to Java Spring Boot. The human effort across all four phases amounted to 65 person-days. Traditionally calculated, a system of this size would require more than 500. The work does not disappear; it shifts from writing to reviewing.

Learn more

Factor 8

less human effort compared to the traditional approach

65 person-days

human effort across all four phases

approx. 1,100

Function Points according to IFPUG in the case study

46.653

lines of code in the cross-check of the measurement

Our approach

Four phases take you from measurement to a production-ready application. After each phase, you decide on the next step, because each phase delivers an independently usable result.

Phase 1 measures the application using an AI-supported Function Point Analysis in accordance with ISO 14143. This provides an objective measure for effort, budget, and prioritization before any development budget is committed.

Phase 2 reconstructs the business logic from three independent sources: source code, existing documentation, and automated tests against the running system. You receive a domain model, use cases, as well as rules and algorithms in pseudocode with references to the existing system.

Phase 3 defines the target architecture based on established standards and aligned with your cloud strategy.

Phase 4 implements it: A squad of more than ten AI agents works in parallel on design, implementation, and testing, guided by experienced developers throughout our Software Development Lifecycle. Quality assurance runs as part of the pipeline. The process concludes with migration and cutover on a fixed date.

Benefits

Measured instead of estimated

Function Point Analysis in accordance with ISO 14143 provides an objective, vendor-neutral measure of scope and complexity. Business, IT, and Procurement can therefore work with the same numbers.

Expert knowledge is documented

The domain model, use cases, and calculation rules are documented in writing and remain available independently of individual people.

Functional parity as the baseline

We modernize what works today on a future-proof technology stack. New requirements go into a separate, prioritized change backlog.

Minimal effort for your Application Owners

AI creates the analysis, while your owners validate it in short review sessions. Weeks of workshops become focused reviews.

Fixed price based on Function Points

Each application has a binding fixed price for its replacement. Payment is made upon reaching defined milestones, up to the cutover.

Binding cutover date

For each application, the date for migration and decommissioning of the legacy system is fixed. This makes technical debt measurably lower.

Market-standard technology stack

We build on technologies for which you will still be able to find developers in the market ten years from now. In the case study: Java Spring Boot, Spring Data JPA, Vaadin, MariaDB, Flyway, Flowable BPMN, and Docker.

Independence on the technology and service-provider side

Independence on the technology and service-provider side Open interfaces and established standards keep operations and maintenance predictable. You retain the freedom to choose who continues to develop the application. If desired, Liongate can operate the modernized application with fixed SLAs.