Problem and data
Review business objectives, data access, existing software and decision impact. Image, tabular or time-series availability determines feasibility. Applications extend beyond factories.
Development and integration
Plan proof of concept, data preparation, modelling and integration in stages. Clarify rights, privacy, cost and maintenance boundaries before a proposal.
Evaluation and deployment
Define success using real test data and business decisions. Document limitations, release conditions and human review after the pilot. AI is not assumed necessary for every problem.