Data readiness
What is Data readiness?
Whether your data is good enough for the AI you want to build: accurate, well-governed, findable and relevant to the use case. Gartner is blunt here. There is no generally AI-ready data, only data ready for a specific job. Gaps at this stage stall more AI projects than the models ever do.
Why it matters
Data readiness decides more AI projects than the choice of model ever will, and because it is the least glamorous part of the work, it gets skipped. A team picks a model, designs a feature, then discovers the data it depends on is spread across systems, inconsistent, missing key fields, or locked behind access rules no one can clear in time. Fixing that is slow and political. It means naming who owns a dataset and who is accountable for its quality, questions many companies have dodged for years. So the real work of getting ready begins long before any model does, in the governance and plumbing that no demo ever puts on screen.
In practice
A retailer wants a model to predict stockouts and finds its sales data lives in three systems that disagree on basic product codes. Before any modelling is worthwhile, someone has to reconcile them and keep them reconciled. The project that looked like a data-science problem turns out to be a data-ownership problem, and that is the part that eats the quarter.
