An AI project team with a manufacturing company needs to ensure data integrity before moving...

In AI data management, PMI-CPMAI highlights data integrity as the property that data remains accurate, consistent, and reliable over its lifecycle. When the team discovers inconsistencies due to manual entry errors, the most direct and effective control is to prevent bad data at the point of capture. This is achieved by implementing real-time data validation rules—for example, enforcing allowed ranges, formats, mandatory fields, cross-field consistency checks, and lookup constraints before a record is accepted.

PMI’s AI data practices emphasize that “controls at data entry” are preferable to downstream correction because they reduce rework, lower the risk of propagating errors into models, and create cleaner training datasets from the outset. Although automating data entry (option B) can also reduce manual errors, it does not, by itself, guarantee integrity if upstream systems or processes are flawed. Regular audits (option C) are useful as a monitoring mechanism, but they are periodic and reactive rather than preventive. Using ML algorithms to detect and correct errors (option D) adds complexity and itself relies on having sufficiently good data.

Thus, in alignment with PMI-style AI governance and quality management, real-time data validation rules are the most effective method named here to ensure data integrity before moving to model development.