Maersk MapGenie

Strategi

Every day, Maersk moves goods across oceans, air, land, and ports worldwide. Behind every shipment, millions of transactions flow between systems that all speak different digital languages.

To keep that world connected, customers rely on EDI - the universal standard that enables secure, system-to-system data exchange across the global supply chain.
But maintaining those connections had become a bottleneck.

Maersk processes more than 1,300 EDI map changes each year - changes to logic built up over decades. That logic is often undocumented and only fully understood by a small number of specialists.

Each change required engineers to manually reconstruct how the map worked before making any modification.

On average:
• 23 hours per change
• 31,000 specialist hours annually

This created a structural dependency on scarce expertise. Integration capacity scaled with the availability of a limited talent pool - not with business demand.
Traditional automation could not solve this. Because the challenge was not execution - it was understanding.
Each map is unique, shaped by years of changes and exceptions.

As a result, the key question became:
'How do you scale a system where the core problem is understanding legacy logic?'

The answer was to shift from manual interpretation to machine understanding.

The strategy was to introduce a new capability:
• AI that can read and understand existing integrations
• remove the need for manual reconstruction
• and turn integration work into a scalable process

The ambition was not just efficiency.
It was to fundamentally change how fast Maersk can adapt its digital connections - and how quickly customers can transact across its global logistics network.

Løsning

To solve this, Maersk and IBM developed MapGenie - an AI-powered platform designed to act as a translator for global logistics data.

In simple terms:
-> MapGenie understands existing EDI maps, applies a requested change, and prepares it for deployment.

From complexity to clarity:
A user describes the required change in plain language - just as they would in a Jira ticket.
MapGenie then:
• reads the live production map
• identifies the relevant logic
• and cuts through the complexity around how the map works
It reveals the structure beneath - making complex, undocumented logic understandable.

From understanding to action:
Based on this, MapGenie generates a structured code proposal.

The output is presented as a 'visual diff':
• current state vs proposed change
• focused only on the relevant parts of the map
This allows users to review changes in context - not as raw code, but as clear, comparable logic.

Built for trust and control:
The solution is designed with governance at its core.
• Deterministic schema and rule checks validate every proposal
• Only validated outputs are presented to the user
• Every change requires human approval
• Deployment follows the existing CI/CD pipeline
• The AI has no direct path to production. Human judgment, AI execution.

MapGenie enables a new way of working.
Previously:
• Two roles (analyst + developer)
• Multiple tools
• Several days per change

Now:
• One user
• One interface
• Minutes per change

MapGenie does not replace existing systems - it connects them.
And in doing so, it transforms integration work from a specialist-driven task into a streamlined, end-to-end workflow.