Key takeaways
- Code each question against the document that should answer it, not the person who did.
- Separate a missing document from an outdated one and from a search problem.
- Repair the highest-frequency gap and recheck with a fresh question log.
Question and scope
The question is whether the documents already in use can answer the questions a new virtual assistant asks during the first month. The review starts from the assistant's real questions, not from a list of documents the team wishes existed. Each question is matched to the document that should contain the answer. A match counts only when the answer is findable and current. Exclude questions that were answered by a document the assistant had already opened that day.
Scope the review to one role and one month, and include questions asked in chat, email, and review meetings. Exclude questions about personal schedule or payroll unless they relate to the work process. Write down the cut-off date for the document set. A question answered by a document created after the question was asked does not count as coverage at the time. Ask the assistant to paste the question exactly as it was asked, not a summary.
| Question type | Document that should answer it | Common gap | Repair |
|---|---|---|---|
| Routine task | Task procedure | No current version | Rewrite the step |
| Exception | Decision guide | Silent on the case | Add an example |
| Boundary | Responsibility map | No named owner | Assign the handoff |
| Tool behavior | System guide | Outdated screenshot | Update the path |
Coding the question log
Keep a question log during the first month with the date, the task, the question, and the answer source. Ask the assistant to log a question when they cannot find the answer in the current instructions. Do not wait for a weekly meeting, because the context is lost by then. A short form with four fields is enough. A question that needed a second person is a stronger signal of a gap than a routine lookup.
Code each question by the document that should answer it, not by the document that happened to answer it. The difference matters because a chat reply from a senior colleague is not a durable source. Mark whether the document existed, was current, and was easy to find. Those three marks separate a missing document from a discoverability problem. If two reviewers choose different documents, record both and discuss the reason.
Method and evidence standard
Use a sample of questions rather than every message in the month. Select the first question from each day and every question that required a second person. That rule captures routine gaps and escalated gaps without turning the review into a full-time job. Record the selection rule so another reviewer can reproduce the sample. Report the outdated share separately from the missing share because the repairs differ.
Have a second person code ten of the questions independently. Compare the document chosen and the coverage mark. Discuss the disagreements and revise the coding guide before continuing. The goal is a shared standard, not a perfect count. Report the disagreements in the limitations section. A document that exists but is not linked from the task page should be coded as hard to find.
Measures and findings
Report the share of sampled questions with a current document that answers them, the share with an outdated document, and the share with no document. Then report the tasks where questions cluster. A coverage rate without a task breakdown can make a narrow gap look like a broad failure. Group the boundary questions by the two roles involved, not by the assistant's task list.
Measure the time from question to usable answer as a secondary signal. A question with a document that takes twenty minutes to find is not covered in a practical sense. Record whether the assistant asked a person after searching. That field shows the difference between a documentation gap and a search problem. A responsibility map can be one page if it names the handoff and the owner.
Where the gaps cluster
Questions often cluster around exceptions: a refund outside the normal window, a client who wants a different format, or a system that refuses a standard action. The main process document may be accurate for the common path and silent on the exception. That silence is predictable and can be documented with examples. Do not count a colleague's chat answer as coverage even when the answer was correct.
Another cluster appears at the boundary between two roles. The assistant may not know whether a request belongs to them or to the client's team. A responsibility map with a named owner for each boundary reduces those questions. The map should be short and kept beside the task instructions. A new document written during the window should be coded with the date it became available.
Interpreting the pattern
High coverage with a high question rate can mean the documents are complete but hard to search. Low coverage with a low question rate can mean the assistant is asking a colleague instead of the document. Neither pattern is visible from a coverage percentage alone, so read the question source and the search behavior together. If the task mix changed, report the coverage rate for the old and new tasks separately.
A rise in coverage after the first week may reflect new documents written in response to the log. That is a good outcome, but it should be reported as a change during the window rather than as baseline coverage. A fall in coverage may reflect a new task or a client change. Check the task mix before assigning a cause. The review should not be used to judge whether the assistant is asking too many questions.
Limitations
The review depends on the assistant's willingness to log questions, and a new assistant may under-report to appear competent. The document set may change during the month, which makes the coverage rate time-sensitive. The sample rule favors questions that required help, so it may overstate the gap. A question about a retired task should be excluded and noted in the limitations.
The review does not test whether a document is correct, only whether it exists and is current. A wrong document can answer a question and still cause a bad outcome. Pair the coverage review with a quality sample of the work produced. State both limits in the summary. Write the repair as a procedure rather than a paragraph so the assistant can follow it.
Practical next step
Repair the highest-frequency gap first. Write the answer as a short procedure with the source, the steps, and the finish standard. Put it where the assistant already looks for instructions. Add a link from the task page rather than creating a new folder that nobody visits. Link the new procedure from the task page and from the question log entry that prompted it.
Assign a document owner and a review date for every new entry. Tell the assistant who to ask when the document is silent. Keep a short list of open gaps and review it weekly. The aim is to reduce repeat questions, not to produce a large library. Remove or archive documents that contradict the new procedure. Tell the assistant which version is current.
A documentation repair loop
After the first repair, log questions for another two weeks. Compare the new questions with the old list to see whether the same gap reappears. Check whether the new document was found without asking a person. If it was not, fix the link or the title before writing more content. A two-week follow-up is enough to see whether the same question reappears.
Review the loop with the document owners and the assistant. Ask which repair changed their work and which one was ignored. Remove documents that no longer match the process. A documentation loop is working when the question log gets shorter and the answers get easier to find. If the follow-up question changes shape but points to the same gap, the repair is incomplete. Name the owner in the document header.
What would change the conclusion
The conclusion would change if most questions already had a current, findable answer and the assistant simply did not search. That is a training or search-design problem, not a documentation gap. It would also change if the gaps clustered in a task that is about to be retired. Ask the document owner to confirm that the repair still matches the process.
A larger sample across several roles would show whether the pattern is local. A review of the documents' accuracy would show whether coverage is meaningful. Until then, report coverage as a description of this role and use the question log to choose one repair. A coverage review is useful when it shortens the question log, not when it produces a new library.
Sources and methodology
A bounded review of questions logged by one virtual assistant role during a fixed first-month window. Each sampled question is matched to the document that should contain the answer and coded for existence, currency, and findability. The result describes local documentation coverage only and does not test the accuracy of the work produced.
- U.S. Office of Personnel Management — Training and Development current.
- U.S. Small Business Administration — Hire and manage employees current.
- NIST Cybersecurity Framework 2.0 current.
- CISA Secure Our World current.
Source count: 4. Last verification date: November 7, 2026.
Questions about the findings
What counts as documentation coverage?
A sampled question counts as covered when a current document contains the answer and the assistant can find it without asking a person.
Should every question become a document?
No. A one-off question may not justify a document. Look for repeated questions and gaps that carry operational risk.
Does coverage prove the document is right?
No. Coverage tests whether an answer exists and is findable. A separate review is needed to test whether the answer is correct.
Put the finding to work
Review the full research library, compare cluster coverage inside Compliance Operations, and pair these findings with our VA candidate screening support.
