Enterprise Architecture

Overview

The EA Modeler is a canvas-based tool for building and maintaining your organization’s enterprise architecture. Models are created in ArchiMate 4 (the default notation for enterprise structure), with BPMN 2.0 for process/workflow behaviour and SysML v2 for systems requirements, constraints, interfaces, and verification. They are intended to be implementable — not just decorative diagrams. They connect directly to the products, technology, and operations managed elsewhere in the platform.

Key Concepts

Reference Models and your archetype

Some reference models are universal and some belong to one industry. IT4IT describes IT management for any organisation that runs IT, so every install gets its full criteria set. An industry model such as the BIAN Service Landscape serves banking, and its criteria are imported only on an install whose archetype is a banking one.

The catalogue entry is kept everywhere on purpose, so you can see that the standard exists and what it is for. On an install it does not serve, the model is listed as not this archetype with the reason, instead of showing a lifecycle status beside empty counts. Empty counts there are correct rather than incomplete: the criteria were never imported because the model is not your business.

If you change the archetype your install declares, the next seed imports the criteria for any industry model that now applies.

Workroom Definitions

Within Coordination, Owner and links shows the accountable person and whether that responsibility is explicit, inherited from a room, or recorded for the organization. Conflicting or incomplete ownership remains a setup gap. Recorded parent, child and dependency links lead to the related room while keeping your filters. A dependency alone does not prove the work is waiting. An incomplete read is labeled; no links observed is not proof that no links exist.

Open Architecture > Workrooms (/ea/workrooms) to see the rooms that are actually running and how they sit across the four portfolios. The page leads with a true count of open rooms — completed and archived rooms are excluded — and says plainly how many have no portfolio recorded. Those are reported as unclassified, never quietly filed under a portfolio nobody assigned. Where team plans exist, each is grouped under its owning portfolio with its shape, participants, queues, triggers and process view.

A team is placed in a portfolio from what the platform actually records: an explicit portfolio role on the team wins, and its portfolio slug or name are used only as weaker fallbacks. When none of those decides the question, the team is listed under Not placed in a portfolio with the reason, instead of being shown inside a portfolio nobody assigned it to. Those teams are counted separately, so a portfolio’s total only ever counts teams that genuinely belong to it. Correct a placement by setting the team’s portfolio role or linking it to the right portfolio.

The overview room count is a real total. The plan list and the Coordination list are each a bounded read. More plans exist identifies a truncated plan list; Next rooms continues through matching open rooms in Coordination. Its count describes the displayed page. If no team plans are configured on your install, the page says exactly that rather than implying you forgot to set one up — there is currently no in-product way to create one, so an install with none is normal.

What You Can Do

Exporting a view as a drawing

Open a view and choose Export / Import at the top right of the canvas to download it as a drawing that people can open without DPF:

The drawing uses the view’s saved layout and the same layer colours as the canvas (business yellow, application blue, technology green). Relationship lines and label text use your organization’s brand colours. Duplicate relationships are drawn once, as on the canvas. A single drawing holds up to 500 elements.

The export runs on the platform’s document engine. If the engine is not set up on your install, Export says so and nothing is downloaded.

You can also ask the Enterprise Architect coworker to export a view. It saves the drawing as a document (the .odg, with its PDF and SVG kept alongside) and gives you the link. This is a picture of the view, not a model exchange: to move the model itself into another ArchiMate tool, use the ArchiMate exchange export.

Importing a Visio or Draw diagram

If your organization already keeps architecture diagrams in Visio (.vsd, .vsdx) or LibreOffice Draw (.odg), you can bring what they show into review. Open any view, choose Export / Import, then Import a diagram…:

  1. Choose the file and select Import diagram. Files up to 900 KB are accepted.
  2. The platform’s document engine reads the drawing. Each labelled shape becomes a candidate element, and each connector that joins two shapes becomes a candidate relationship, with the connector’s label if it has one.
  3. Open the import to see a picture of the diagram beside the candidates. Accept or reject each one.

Nothing is added to your model by an import or by accepting a candidate. Accepting records that the candidate is right; adding it to a view is still done on the canvas. Shapes with no text and connectors whose ends do not reach two shapes are not listed; the import says how many connectors were left out. A diagram DPF exported (see above) imports with its layer shown for each shape.

Importing the same file twice shows the first import again instead of making duplicates. You need permission to edit the architecture model to import or review; anyone who can see the views can see the candidates. If the document engine is not set up on your install, the import says so and nothing is saved.

AI Coworker Identity

Architecture views may still contain older /ea/agents links from governance and route projections. Those links are compatibility paths. They now open /platform/identity/agents, the canonical AI Coworker Identity page for principal coverage, portable identity metadata, and authorization inspection. Use the AI Workforce area for everyday coworker discovery and work assignment.

Following a Concern Across Views

On Workrooms, open Initiatives to find a named initiative and see its recorded purpose, scope and open rooms. Search by name or ID, then select its link to narrow Coordination. Partial or unavailable reads are labeled. Initiative membership does not establish a value-stream link, capability link or reporting relationship. The selected initiative and search stay with you through a room and its process, including the Operation return link.

On Workrooms, expand Coordination to open an actual room. Use Find a room to search by title or room ID and filter by operation or status. Next rooms continues the bounded list; changing filters starts a new first page. A room with no team plan behind it says so explicitly. Select a plan’s room link to narrow the list to that operation first. Recorded stage waits name the responsible role when the execution source records it.

Inside the room, select a process step to inspect its reason, next action, owner and evidence. Operation returns to the same operation, filters and page. Intended process and observed records remain separate: a graph alone does not establish BPMN execution or SysML conformance.

Select an element on an architecture canvas and open Architecture context in the inspector. Related viewpoints lists the nearest views that explain the same element or a directly linked concern. Each link names the notation, target view, and link distance so you can see why it is being suggested.

AI routing elements can also offer Open operational evidence. This opens the AI Operations Map in Compare mode, focused on the same routing stage. The architecture remains the designed projection; the Operations Map supplies the observed evidence and calls out gaps rather than treating missing evidence as proof that the design ran.

The Why this decision works this way section deliberately shows only safe metadata. It can include decision inputs such as sensitivity class or provider eligibility, possible outcomes, the design version, implementation status, and the latest evidence time. It never exposes prompt text or protected payload values.

Edit this page on GitHub