Create a data dictionary with human-approved definitions
A field name can look obvious to a person and still be ambiguous to an AI system. Microsoft Digital’s October 15 report highlights the value of explicit business context and describes AI drafting data descriptions for people to review.

This ATC workflow turns that idea into a small documentation task:
- Pick one table. Choose a dataset used in a bounded pilot and state its intended use.
- Name an owner. Assign someone who can explain the business process and resolve uncertain meanings.
- Define fields. Record meaning, units, allowed codes and how blank values should be interpreted.
- Let AI draft. In an approved tool, use permitted metadata and existing definitions to organize a draft. Ask it to mark unknown meanings rather than invent them.
- Check real meaning. Have the owner compare every definition with the actual process. Use safe examples to distinguish similar statuses.
- Version and review. Save the approved dictionary with an owner and review date. Revisit it when fields or workflows change.
For example, “closed” might mean the work ended, while “resolved” might mean the customer confirmed the fix. Your team must supply the real distinction. A useful dictionary records field name, definition, type, units or codes, blank-value meaning, owner and last review date.
This is an illustrative ATC process, not a certification or measured performance result. Start with existing approved tools; no new purchase is required to document definitions.
Source context: Microsoft Inside Track: AI-ready data, October 15, 2026; reviewed October 18.
American Tech Consultants | americantechconsultants.com | (954) 998-2994 | support@americantechconsultants.com