The toolkit separates three kinds of capability because they need different data, skills, and expectations. Confusing them is how projects get funded for the wrong reasons.
Descriptive: what happened
Reports, dashboards, and segmentation. Most businesses get the largest early return here, and it needs only clean data and clear questions, not machine learning.
Predictive: what will happen
Forecasting demand, scoring leads, flagging likely churn. Needs historical data with outcomes recorded, and it produces probabilities, not certainties.
Generative: producing content
Drafting text, summarizing documents, answering questions from a knowledge base. Powerful for productivity, with its own accuracy and confidentiality risks that the governance module covers.
Keep the scope of this step small on the first pass; the toolkit is designed to be revisited, and a rough complete answer beats a polished partial one.
Record any assumption you make here in the assumptions log so that when the pilot results arrive you can see which assumptions held.
Action Step
Classify every analytics or AI idea currently floating around your business into one of the three kinds on the toolkit’s capability sheet.
This lesson is educational content for founders and operators. It is general information, not professional advice for your specific company. Adapt every framework to your own situation and consult a qualified professional where a decision carries legal, financial, or tax consequences.