AI Builder

AI Builder is the no-code machine-learning capability inside the Power Platform. It ships pre-built models for common tasks — invoice processing, receipt reading, ID extraction, sentiment analysis, OCR, business card capture — and supports custom-trained models for prediction, classification, entity extraction, and document processing trained on the customer's own data. Models are called from Power Apps, Power Automate, Dataverse plug-ins, and Copilot Studio agents. AI Builder runs on a credit consumption model bundled with some Power Platform plans; tracking credit usage is the most common operational discipline.

In practice, a maker reaches for a pre-built model first — invoice or receipt processing is the most common starting point in a Dynamics 365 context, feeding extracted line items straight into a Power Automate flow that creates a purchase invoice or expense line. Custom models need a training exercise: enough labelled example documents or historical records to train against, then a validation pass before the model goes live on real data. Model outputs carry a confidence score, and a well-built flow branches on that score — high-confidence extractions post automatically, low-confidence ones route to a human reviewer instead of posting a wrong number silently.

The most common operational surprise is credits: AI Builder capacity is metered, shared across the environment (or tenant, depending on plan), and a single high-volume document-processing flow can exhaust a month's allocation faster than expected if nobody is watching consumption. Treat credit usage as a capacity-planning exercise from the first production flow, not an afterthought discovered when processing suddenly stops.