The AI assistant
What it is for
Art2link ESB ships with an AI Assistant in the portal. Its job is the part of integration work nobody enjoys: reading message samples, figuring out fields, and helping you build the maps, custom functions, and pipeline components that handle them.
It arrives already trained on the transaction formats integration teams actually meet: EDI, HIPAA, HL7, FHIR, and most other popular transaction types. Hand it a sample message and it recognizes the fields on its own. Custom fields, if you have any, are the only thing you explain.
Where your data goes: nowhere
The first question every regulated customer asks: "we handle patient data, where do these prompts go?"
Here is the answer, and it is the whole design: the AI Assistant runs locally in your tenant. Its infrastructure is deployed into your own subscription, next to the rest of the platform. Your prompts, your sample messages, and its answers never leave your tenant. Nothing crosses to Cerebrum City or to any outside AI service.
The boundary runs both ways. The assistant has no access to the actual transactions crossing the bus, no access to your tracking information, and no access to the internet. Its only input is what you paste into the chat: a sample here, a question there. Production message traffic is simply not visible to it.
What it costs
No base cost. The AI Assistant is billed on consumption only: use it, pay for what you used; leave it idle, pay nothing.
One module to go
Publish-select-deliver, the building blocks, the adapters, the habits, and an AI assistant that keeps your data at home. One thing is left: seeing what actually ran. Module 6 covers Tracking, the searchable historical record of every integration run.
Check yourself
Two quick questions on this module, in the same style as the exam. Not graded, not recorded, just practice with instant explanations.