A year-end summary can become misleading when an assistant merges plans with completed work or turns discussion into a final decision. The solution is to make status and evidence visible before writing a polished narrative.
Use this process for your own documents or material you are authorized to process. Remove unnecessary personal information before using an external tool.
Build a small evidence ledger
For each significant event, record date, source document, location in the document, status, and consequence. Use distinct labels for proposed, approved, completed, and verified.
Ask the assistant to extract candidate entries into that structure. Require a source pointer for each entry and an explicit unknown when a field is absent.
Check the important entries
Open the original material for the milestones that matter most. Confirm that a later update did not reverse an earlier plan. Compare numbers against their original time periods and definitions.
If two sources conflict, preserve the conflict until it can be resolved. A fluent summary should not silently choose the version that makes the story cleaner.
Write from the checked ledger
Generate the narrative only after correcting the ledger. Keep accomplishments, limitations, and next actions distinct in meaning, even when they appear in connected prose.
Review the final summary for invented causality: “after” does not automatically mean “because of.” Remove claims of impact that lack supporting evidence.
NIST’s AI framework provides general risk-management context. The ledger and audit sequence are our suggested editorial method, not a certified NIST workflow.
See AI citation auditing for source checks and the original verification guide for prioritizing consequential statements.
