Build custom AI models from process mining insights

  • 6 minutes

Power Automate Process Mining is powerful in analyzing historical data from a process to help you determine how to optimize the process. To activate that feature, you ingest historical data that includes a wealth of information that you can also use to generate custom AI models. Then, you can optimize your process by using that model in the process automation.

For example, consider a scenario where you want to predict whether a payment might be late or not. You could use process mining to do a root cause analysis (RCA) that focuses on late payments. In the following image, the RCA is set up to focus on late payments and the metrics that might influence that scenario.

Screenshot of the beginning of the root cause analysis.

By exploring the RCA, you can better determine the influences on late payment.

Screenshot of more details of the root cause analysis.

You can export this data to use as training data from the RCA. Then, you can import this data into Microsoft Dataverse to use as training data with AI Builder. AI Builder predictive models learn from your historical data. It analyzes and identifies patterns and associates them with outcomes. For more information, see Overview of the prediction model.

Screenshot of the newly created Dataverse table.

Then, you can train with the data.

Screenshot of the predictive model being trained.

In the following image, notice that the trained model scored a "B" for performance. It tracked the same influencers that the RCA discovered.

Screenshot of the results of the trained model performance.

Next, you can use the model in a flow. The following flow expedites a claim to help reduce late payments.

Screenshot of the predictive model being used in a Power Automate cloud flow.

The following video demonstrates an end-to-end example. The process starts from process mining by using the root cause analysis and then moves to generating the training data for a custom predictive AI model that a Power Automate flow uses.

By using historical data, and by using process mining to focus on problem areas, you can infuse AI into your processes to make them more proactive instead of reactive.