Starting and learning
Machine learning for executives
Businesses use machine learning to predict which customers will respond or leave, flag defects on production lines, forecast demand and help staff find answers in documents. An executive's job is to pick the decision it improves, the data it needs and the evidence that it works better than today's process.
How businesses use machine learning: a guide for executives
Forecasting demand
Forecasting models predict demand for many products and locations at once and give a range rather than a single number. That range feeds stock, staffing and purchasing decisions.
Real exampleDeepAR at AmazonAmazon researchers describe DeepAR, which trains one recurrent neural network across many related time series to produce probabilistic forecasts such as retail demand.
Detecting fraud
Fraud models score each transaction in real time and block, review or allow it. Because fraud is rare, the model is judged on how many good customers it wrongly blocks as well as how much fraud it catches.
Real exampleStripe RadarStripe describes how Radar's fraud models moved from simpler models such as logistic regression to deep neural networks as its payment network grew.
Reducing energy and operating costs
Models trained on years of sensor logs recommend settings that use less energy while staying within operating limits. The saving is measured against the same equipment run the usual way.
Real exampleDeepMind data centre coolingDeepMind reported in 2016 that applying its machine learning to Google data centres reduced the energy used for cooling by up to 40 percent.
Personalising the product
Models estimate what each customer needs next, such as the right content, offer or difficulty level. The value shows up in engagement or retention measured in controlled tests.
Real exampleDuolingo BirdbrainDuolingo describes Birdbrain, a machine learning model that estimates how much each learner knows and how difficult each exercise is, to set lessons at the right level.
Can executives learn machine learning?
Yes, at the level executives need: what machine learning can and cannot do, what data and people a project requires, and how to tell whether it works. You do not need to code, but a short hands-on course and reading real evaluation results make you a sharper sponsor.
- Core conceptsTraining and test data, baselines, error, and the difference between a prediction and a cause.Learn it in Machine Learning
- Reading evaluation resultsPrecision and recall, uncertainty and A/B tests, so you can question a claimed improvement.Learn it in Statistics
- Generative AI and its limitsHow language models produce text, why they state false facts, and how retrieval and evaluation reduce that risk.Learn it in Generative AI
- Data, governance and operationsData ownership, privacy, monitoring after launch and who is accountable when a model is wrong.
Leading machine learning work as an executive
If you already work in the field
Executives already know the decisions, costs and constraints of the business. Add enough technical understanding to ask for a baseline, a clear metric and a pilot on real data, and start with one frequent, measurable decision owned by a single team.
If you come from data science or ML
Technical leaders moving into executive roles need to translate model metrics into cost, revenue and risk, and to plan for data access, change management and regulation. Practise presenting a project as the decision it improves and the evidence that it works.
Roles that do this workChief data officerChief AI officerHead of AIVP of data science
Tools you will use
| Tool | What you use it for |
|---|---|
| Orange Data Mining | Builds and evaluates a predictive model visually, so you can see a baseline and test score without code. |
| KNIME Analytics Platform | Creates repeatable no-code data and modelling workflows that analysts and managers can review together. |
| Teachable Machine | Trains a small image classifier in the browser to see how inspection-style models learn from examples. |
| Google Colab | Opens your team's Python notebooks in the browser so you can follow how an analysis was produced. |
Public data to practise on
- Default of Credit Card Clients: Credit card customers' payment history and limits, labelled with whether they defaulted the next month.
- Online Retail II: Transactions from an online retailer, useful for analysing repeat purchases, customer value and demand.
Your first project
Open the Default of Credit Card Clients dataset in Orange, predict next-month default with a decision tree, and compare it with predicting no default for everyone using precision and recall.
Common questions
What should executives know about machine learning?
That it predicts from patterns in historical data, so it is only as good as that data and most reliable in conditions like it. Ask what decision it improves, what it is compared against, how errors are caught and who maintains it.
How do you measure the value of a machine learning project?
Compare decisions made with the model against today's process on the same cases, ideally in a controlled test. Measure the business outcome, such as cost, revenue or time saved, not only model accuracy.
Why do machine learning projects fail?
Common causes are an unclear decision to improve, data that is unavailable or unrepresentative, no baseline, and no plan to run and monitor the model after the pilot. Most of these are organisational rather than technical.
Should we build or buy machine learning?
Buy or use a platform for common tasks such as transcription, document search or fraud screening. Build when the problem depends on your own data and processes and gives you a lasting advantage.
Courses for these skills
- Practical Machine Learning in PythonBuild the foundations for classification, anomaly detection and model evaluation.51 published lessons · more lessons in production
- Data Tools: NumPy, pandas, Matplotlib and scikit-learnKeep datasets, joins and feature definitions reproducible.23 published lessons · all planned lessons open
- Statistics and Probability for Machine LearningUnderstand base rates, sampling and uncertainty.41 published lessons · all planned lessons open
- Deep Learning with PyTorchStudy CNNs, transfer learning, object detection and segmentation in the Computer Vision syllabus.Syllabus available · lessons open as produced
New to programming or maths? Start with Python, or follow the machine learning path in order. Published lessons are free to read; compare Free and Pro.