Data steward
A person responsible for keeping particular data assets accurate, well-documented, and properly used within their part of the organisation.
The Simple Version
A person responsible for keeping particular data assets accurate, well-documented, and properly used within their part of the organisation.
Detailed Explanation
Data stewards occupy the execution layer of data governance: they define business rules for data elements, investigate and resolve quality issues, maintain business glossaries and data dictionaries, and ensure compliance with data policies in their domain. Unlike data owners (who hold accountability) and data architects (who define structure), data stewards perform ongoing operational work. In AI development contexts, a data steward may be responsible for reviewing training dataset composition, flagging representativeness issues, and maintaining data provenance records.
Key Characteristics
- Operates at the domain or dataset level with specific data assets in scope
- Responsible for quality monitoring, issue resolution, and metadata upkeep
- Works at the intersection of business knowledge and technical data operations
- Distinct from the data owner, who holds ultimate accountability
Why It Matters
Organisations building high-risk AI systems benefit from designating data stewards for each major training dataset; the steward's ongoing work supports the data governance requirements of EU AI Act Annex IV.
Real-World Analogy
Like a head gardener who is accountable for the health of specific flowerbeds, they do not own the garden but are responsible for its day-to-day upkeep, monitoring, and improvement.
Common Misconceptions
- Data stewards and data owners are the same role, owners hold strategic accountability; stewards perform operational tasks on behalf of owners.
- Data stewardship is a part-time administrative duty, effective stewardship is a skilled, time-intensive role requiring business domain expertise and analytical capability.