1. The operating brief
A team needs dependable information about markets, vendors, accounts, products, locations, competitors, contacts, or business requirements, but senior staff should not spend hours opening pages and cleaning spreadsheets. The work needs more than a broad instruction to research a topic.
MY-VA can assign a research virtual assistant to a defined question and output. The assistant follows approved source criteria, captures evidence, distinguishes confirmed facts from unresolved items, and delivers the information in the system or format the client will actually use.
This is a deployment example, not a claim about a named client or a promised number of records. Research volume and confidence depend on the topic, source access, language, geography, and verification standard.
2. Turn the research request into a field specification
The brief defines the business question, permitted sources, excluded sources, geography, date range, required fields, acceptable evidence, and the decision the research will support. Each field needs a clear definition so two researchers would capture the same fact in the same way.
The output may be a spreadsheet, CRM list, vendor comparison, location directory, account profile, content brief, or source library. Required fields are separated from optional context, and unknown values remain unknown instead of being guessed.
- Research question and intended business use.
- Required fields, formats, and allowed values.
- Approved source types and evidence requirements.
- Exclusions, duplicate rules, and stopping conditions.
- Sample records for client approval before full production.
3. Capture the source with every material fact
A useful record shows where the information came from and when it was reviewed. The assistant can prioritize official company pages, public registries, product documentation, government sources, and other sources approved for the project. Search snippets are treated as leads, not final evidence.
When sources disagree, the record identifies the conflict instead of silently selecting the most convenient answer. Restricted, personal, or sensitive information stays outside the workflow unless the client has documented a lawful purpose and approved handling method.
4. Keep the dataset consistent while it grows
The assistant works from a controlled template with naming rules, field validation, source columns, notes, and an exception status. Unique identifiers and duplicate checks prevent the same company, contact, location, or product from appearing as several different records.
Production moves in reviewable batches. Early samples reveal ambiguous fields, inaccessible sources, and categories that need adjustment before the entire list is populated. Approved changes are recorded so earlier rows can be updated consistently.
5. Review accuracy, completeness, and evidence
Quality review checks whether required fields are complete, values match the source, links support the claim, formats are consistent, duplicates are handled correctly, and uncertain information is labeled. High impact fields can receive a second source or a second reviewer when the client requires it.
Corrections are returned with a reason code. That lets the team improve the instructions, retrain the assistant on a recurring error, and separate source limitations from execution errors.
6. Deliver information that is ready for the next owner
The final handoff includes the approved records, evidence links, unresolved questions, excluded items, change notes, and any assumptions the next team must understand. If the research feeds a CRM, campaign, proposal, or procurement review, the destination fields and ownership are confirmed before import.
Useful reporting can include records reviewed, records accepted, unresolved items, duplicate findings, returned records, source access problems, and quality corrections. Sales and decision outcomes are measured separately from research completion and quality.
Another person can trace each important fact to its source, understand what remains uncertain, and use the dataset without rebuilding the researcher’s logic.