Machine learning already drives decisions in many Massachusetts workplaces: demand forecasts, fraud flags, lead scores, predictive maintenance alerts, and no-show predictions. The people acting on those outputs are rarely the people who built them. This instructor-led course teaches workers to use, question, and explain machine learning results on the job. It does not teach model development. Trainees learn how predictions are made, what accuracy, precision, recall, and error really mean for their decisions, how data quality and bias change results, and when a forecast should be overruled. Hands-on labs use AI and analytics features in tools trainees already have: Copilot in Excel, Excel forecasting, Power BI and Tableau AI features, and ChatGPT or Gemini data analysis on fictional datasets. Trainees practice reading a vendor model's report card, checking a forecast against what actually happened, finding a data-quality problem, and writing a plain-language explanation of a prediction for a coworker or customer. Examples are drawn from Massachusetts manufacturing, healthcare, insurance, retail, and logistics, using fictional companies and data. Trainees finish with a prediction review checklist and a capstone decision memo on a model used in their own role.
What you will be able to do
By the end of the course, trainees will be able to:
- Explain in plain language how a supervised model turns past data into a prediction and a confidence score.
- Interpret accuracy, precision, recall, false positives, and false negatives in terms of business cost for a given decision.
- Identify at least three data-quality or bias problems in a fictional dataset that would distort a model's output.
- Check a forecast against actuals and decide, with reasons, when a human should override it.
- Use AI-assisted analysis features in Excel or a chat tool to explore a dataset and confirm one finding by manual calculation.
- Write a decision memo evaluating a model or vendor AI feature used in their role, including risks and monitoring questions.
Who this course is for
Business and operations analysts; Production planners and supervisors; Insurance underwriting and claims staff; Healthcare operations coordinators; Sales operations and marketing analysts; Inventory and supply chain planners
Before you start
Comfortable with spreadsheets (sorting, filtering, simple formulas). FAST-01 or FAST-02 recommended. No programming; no statistics course required.
Hands-on activities
- Confusion-matrix card game: teams set a threshold for fictional Pioneer Valley Mutual fraud flags and tally costs of errors.
- Forecast vs. actuals lab with fictional Bayside Precision Parts demand data in Excel.
- Data-quality hunt: find duplicates, missing values, and a skewed sample in a fictional clinic no-show dataset.
- Vendor report-card review: read a fictional model card and list questions to ask the vendor.
- Explain-it-back: write a 100-word explanation of a prediction for a customer or coworker; peers score clarity.
- Capstone decision memo with cohort review.
How your learning is measured
- Attendance at all five sessions.
- Five lab deliverables scored on a 3-level rubric (Not yet / Meets / Strong).
- Short applied quiz on interpreting metrics in context (80% to meet; retake allowed).
- Capstone decision memo scored against rubric; presented to cohort.
