Most AI projects stall because of the data, not the model. Use this checklist before you invest.
AI is on every leadership agenda, but most projects stall before they deliver value. The reason is rarely the algorithm. It is almost always the data underneath it.
The checklist
- Access: can you get the data you need without manual exports?
- History: do you have enough history to learn from, typically one to three years?
- Consistency: are key fields such as customer, product and date defined the same way everywhere?
- Quality: do you know how complete and accurate the important fields are?
- Ownership: is someone responsible for fixing data issues when they appear?
- A clear question: is there a specific decision the model will improve, with a way to measure it?
If you answered no to several
That is normal, and it is good news: the work that makes data ready for AI also improves your everyday reporting. Centralising data, agreeing definitions and adding quality checks pay off immediately, long before a model is deployed.
Our advice is simple. Pick one high-value use case, such as demand forecasting or fraud screening, get the data for that use case into shape, and build from there.