API-first · ESB · EDI · Legacy modernisation

Enterprise system integration and legacy modernisation

ERP, CRM, WMS, MES, banking and manufacturing systems connected under a unified API layer. EDIFACT, OpenPEPPOL, AS2/AS4, Kafka, RabbitMQ. Strangler fig pattern for gradual replacement of legacy systems (AS/400, Oracle Forms, SAP R/3). Audit from €5,300, full integration €40k–€530k.

Services

System integration — six service lines

A system integration programme is rarely one project — it is a strategic partnership, typically 12-36 months with phased delivery. Scroll through the six work packages.

System integration audit

A map of the current IT stack: which systems exist, how they communicate (or don't), where data is duplicated, where the bottlenecks are. Output: integration map + roadmap and priorities.

API-first strategy

A unified API layer over every new and existing system — REST, GraphQL, gRPC. Documented, authenticated, rate-limited and monitored from one place. API gateway: Kong, AWS API Gateway, Cloudflare Workers.

EDI and B2B integration

EDIFACT, ANSI X12, OpenPEPPOL, AS2/AS4 — supplier and customer integrations across automotive, retail, pharma and logistics. Hungarian EDI standard (NAV) is also supported.

ESB and message queue

Enterprise Service Bus (Kafka, RabbitMQ, ActiveMQ, NATS) — asynchronous communication, event-driven architecture, guaranteed delivery, replayability for audit.

Legacy modernisation

Modernising legacy systems (Oracle Forms, AS/400, Lotus Notes, FoxPro, Delphi, monolith Java) — strangler fig pattern, gradual migration, without disrupting business processes.

Cloud & on-prem hybrid

Hybrid integration: critical on-prem systems (SAP, banking core) connected to cloud services (CRM, marketing, BI). Site-to-site VPN, AWS Direct Connect, Azure ExpressRoute.

Integration surfaces

Where we already have hands-on experience

Eight integration surface groups where we are not learning on the job — we have shipped production connections for each. Industry-specific protocols (HL7, SWIFT, OpenBanking) all come from real projects.

Legacy patterns

Six proven patterns for legacy modernisation

These are not academic theories — they come from modernising live AS/400, Oracle Forms, SAP R/3 and monolith Java systems. Cutover is never big-bang.

Strangler fig pattern
A new API layer over the legacy system, with gradual function-by-function replacement. The legacy system runs until the last function is migrated. Big-bang cutover NEVER.
Anti-corruption layer
Vocabulary-level isolation of the legacy data model from the new system. An adapter layer that translates — old mess does not bleed into the cleanliness of the new system.
Event-driven shadow
Every transaction published as an event (Kafka). The new system consumes and computes in parallel — discrepancies trigger alerts and investigation. Cutover only when discrepancies converge to zero.
Database-first decoupling
Logical segmentation of the monolith database: a separate schema per bounded context. Only after that can you split into microservices. Database-level replication, CDC (Debezium).
Adapter for AS/400 / IBM i
RPG code understanding, JDBC adapter or IBM HATS, BAPI / IDoc for legacy SAP R/3. The 1990s code lives on, with a modern API layered on top.
Idempotency for everything
Retry-safe APIs, exactly-once semantics, dedup table — a network failure must not result in a duplicate transaction. Outbox pattern + transactional inbox.
Programme phases

Six phases — from audit to handover

A full programme is typically 12-36 months. Money starts flowing back when the first integrations of phase 4 go live — from there, manual-work reduction is the main ROI driver.

01 — Discovery & audit (4-6 weeks)

Stakeholder interviews, system map, data flow diagrams, risk assessment. Output: integration master plan, budget, ROI forecast.

02 — Strategy & architecture (2-4 weeks)

Target architecture, technology stack choice (REST vs GraphQL vs Kafka), security model, governance. Integration is not a project — it is a programme.

03 — API gateway & ESB foundations (2-4 months)

The integration backbone: API gateway (Kong / AWS / Cloudflare), service mesh, message queue (Kafka / RabbitMQ), monitoring (Datadog / Grafana). Individual integrations sit on top.

04 — Individual integrations (3-12 months)

Prioritised integrations delivered in parallel. A typical integration is 4-12 weeks, depending on how well-documented the source-system API is. Kanban-style flow: documentation → adapter → testing → shadow mode → live.

05 — Data sync & migration (2-4 months)

Historical data migration from legacy systems to the new ones. Data cleansing, deduplication, master data management. Validation, audit and rollback procedures for every step.

06 — Stabilisation & handover (1-3 months)

Performance tuning, monitoring fine-tune, runbooks, on-call setup, documentation. Your team is onboarded, with a mentor role for 3-6 months — after that the system is truly yours.

GYIK

System integration — frequently asked questions

System integration = connecting different business systems (ERP, CRM, WMS, MES, e-commerce, banking, HR) so that data lives in one place and processes flow across them. Typical signs: you re-key the same data into multiple systems, reports take weeks, customer data lives in 5 places with different versions, errors compound. If you see this in your operation — you need system integration.

System integration audit + roadmap

4-6 weeks, fixed price. Mapping of the current IT stack, integration map, prioritised roadmap and budget. The right starting point for a programme.

Start a project