Business Process Improvement with Machine Learning: 6 Week Pilot

Every business process can benefit from machine learning. Data gathered, during the execution of business processes, are used to train models that score probable actions and next steps.

Key decision points throughout business processes greatly impact results. For example:

• Will a customer churn? • Will an over-due debt be paid? • Is the product price optimized for margin? • Will a device fail?

Insight into these questions, through machine learning, allows your organization to act during the business process. Thus, providing data driven, positive outcomes.

This pilot delivers the entire lifecycle of applying machine learning to an existing business process. In the end, you will have completed an AI experiment that allows your organization to improve that business process.

The Business Process Improvement with Machine Learning Pilot schedule: Week 1: • Identify stakeholders and candidate business processes (customer) • Define clear objectives and measurements • Confirm what data is available to support the process

Week 2: • Acquire historical data for model training • Data profiling and wrangling • Determine feasibility of successful outcome for pilot • Identify how model outcome will change the business process • Plan business process testing

Weeks 3 - 4: • Identify features • Train the model • Test the model • Repeat until reliable outcomes (or not reliable)

Weeks 5 - 6: • Set up testing • Run testing • Evaluate the outcome and determine high level ROI

PILOT Close: Proposal to automate scoring & maintenance.

The following Azure technologies will be used to complete the PILOT: • Azure Data Lake • Azure Databricks • Jupyter Notebooks