aiExpert overview

aiExpert is a new capability of the Model Expert product that brings the power of large language models (LLMs) into the day-to-day work of enterprise and systems modelers. Whereas traditional validation tools focus on syntax and tool-specific checks, aiExpert takes a broader approach: it uses established, public-domain standards and modelling guidelines to provide direct, contextual feedback to modelers as they build and refine their models.

At its core, aiExpert takes selected data from a model – such as elements, connectors, properties, and diagram structures – and submits it to a trusted LLM. The LLM is guided by a curated set of rules that reference widely available best practices and standards. For example, in BPMN it applies conventions from the OMG specification and published style guides; in ArchiMate it draws on The Open Group’s specification; and in SysML it uses guidance from the OMG SysML specification and INCOSE modeling recommendations. This ensures that the feedback is not arbitrary, but grounded in authoritative sources.

Importantly, aiExpert is not limited to one language. While it currently supports BPMN, ArchiMate 3, and SysML v2, the underlying approach is language-agnostic. All that is required is that some best-practice guidelines are available – in the public domain, and which the LLM has previously absorbed – for the target modeling language. This makes aiExpert adaptable to emerging notations, proprietary frameworks, or even organization-specific modeling conventions. By capturing those guidelines in a form the AI can apply, teams can instantly extend the value of aiExpert across their modeling landscape.

For modelers, the benefit is simple and direct: clear, specific, and actionable feedback. aiExpert can explain why something may be problematic, point to the standard it is derived from, and often suggest an improved alternative. This makes it particularly effective as a teaching tool for less experienced modelers, who can learn “good practice” as they go. More experienced users, meanwhile, gain a fast, consistent quality check that reinforces modeling discipline across teams and projects.

You can save your own customized queries and rules in your EA repository, so they can be used by all your users, or allow users to develop their own queries just for use by a specific team.

Because aiExpert is built into the familiar Model Expert environment, users do not need to learn new workflows or tools. The AI feedback is presented in the same style as other Model Expert checks, but with the added depth and flexibility that comes from a language model. Teams can decide whether to run checks interactively while modelling, or as part of a broader review cycle before publishing or sharing models.

In practice, this means:

• A BPMN process model might be flagged for having multiple start events without a gateway, with an explanation citing the BPMN 2.0 standard.
• An ArchiMate diagram might be checked for viewpoint consistency, reminding the user to adopt a strategy layer where applicable.
• A SysML block definition diagram could be checked for missing multiplicities, with references to the SysML specification.

Over time, aiExpert can be extended to support new languages or to incorporate organisation-specific rules. In this way, it combines the flexibility of AI with the reliability of standards, helping teams to raise the quality of their models without adding complexity.