Write a Research Assistance Brief for a Filipino Assistant

The short answer

The best way to delegate research assistance is to define a result and a boundary before discussing hours or tools. This guide is for a manager who needs decision-ready research rather than a folder of loosely related links. The intended result is the assistant returns an inspectable answer to a defined question, with relevant passages, source limits, contradictions, and unanswered points clearly separated. That result is narrower than "handle it for me," and that is useful. A narrow result tells a candidate what good work looks like, gives a manager something observable to review, and prevents routine preparation from drifting into decisions the business owner still needs to make. It also creates a realistic role discussion: the manager can show actual examples, while the assistant can explain how they would handle an ordinary case, an incomplete request, and an exception.

Begin with the input contract. The minimum working set is the decision to support, audience, geography, date range, defined terms, acceptable source classes, excluded material, output format, deadline, citation fields, and the person who will interpret the findings. Put these items in the system where the work will happen instead of scattering them across messages. Mark which record is authoritative when two tools contain similar information. State who may change the instruction, how that change is dated, and when the assistant should ask rather than infer. A complete input does not need to predict every unusual case. It needs to make the normal path repeatable and make missing information obvious. If the manager cannot assemble this material, the lane is not ready for unsupervised execution; the first assignment should be helping the manager document it.

Write the normal path as a sequence of observable actions: restate the question; break it into evidence needs; search authoritative sources first; record relevant passages and scope; test for contradictions; stop under the agreed rule; summarize without stretching the evidence; escalate consequential interpretation. Each verb should produce evidence another person can inspect. "Handle" and "manage" are not completion criteria. A source link, dated record, approved draft, sent confirmation, validation result, or named escalation can be checked. Define the finish line at the same time. An item is not finished merely because the assistant touched it or moved it to another list. It is finished only when the required action occurred, the record reflects that action, and any approval or delivery required by the workflow is visible. If work is waiting on the owner, label it waiting rather than complete.

Consider a realistic boundary case. A manager asks whether a proposed tool supports a required security control. Marketing pages say the tool is secure, but the available documentation does not address the exact control. The assistant should record what was checked and the gap. They should not convert general security language into a yes. The system owner or security reviewer must decide the next step. This example belongs in the operating guide because rules become clearer when they are tested against a choice that could plausibly occur. Add one routine example, one incomplete request, and one stop case to the role brief. Ask a candidate to talk through what they would check, what they would record, and who they would contact. The goal is not to reward a particular phrase. It is to see whether the person preserves facts, notices the edge of their authority, and can hand an exception to the owner without losing the useful work already done.

The stop list for this lane should explicitly include requests for legal, medical, tax, employment, or security conclusions; inaccessible evidence essential to the answer; credible contradictory sources; personal data; invented estimates; and any instruction to find support for a predetermined claim. A stop rule is not a sign that the assistant lacks initiative. It protects the customer, worker, company, and assistant when a request changes from administration into judgment or commitment. For every stop category, name the owner, approved channel, minimum evidence, and expected response window. Also give the assistant a safe interim state: preserve the record, avoid sending or deleting anything consequential, and continue only the unaffected work. Never ask an assistant to resolve a sensitive exception by borrowing someone else's login, copying private material into an unapproved channel, or making a promise simply to close the queue.

Access should follow the job, not convenience. Create an individual account where the system permits it, require multifactor authentication, and grant only the actions needed for the first scope. Avoid shared passwords and broad administrator roles. Decide where exports and working files may be stored, how long they are needed, and how access will be reviewed or removed after a role change. The Philippines National Privacy Commission publishes the Data Privacy Act and related materials, while NIST and CISA provide practical security guidance for remote access and multifactor authentication. These sources inform a sensible control baseline, but the system owner must apply the organization's own legal, security, retention, and incident-response requirements.

Review quality from the evidence of the work. Check whether every material factual statement maps to a source whose publisher, date, population, and definition fit the wording. Verify that quotations are minimal and accurate and that uncertainty remains visible in the summary. Early review should be frequent and focused on a small sample that includes both ordinary and difficult items. Record the specific rule that a correction relates to. "Be more careful" does not help a person reproduce the expected result. If reviewers disagree, the manager should clarify the instruction or add an example rather than making the assistant guess which preference will apply next time. When the same failure repeats across several people or items, inspect the input, access, and review design before treating it as an individual performance problem. The operating system around the role may be creating the error.

Use measures that support a management decision. Measure material claims with retrievable support, reviewer questions caused by missing context, contradictions disclosed, and briefs returned for scope problems. Link count and search time do not establish research quality. Define the numerator, denominator, source, window, exclusions, and owner for each measure before the first report. Keep workload, accepted output, quality, and blocked time separate. A rising queue can mean demand increased, review slowed, inputs worsened, or capacity changed; one count cannot identify the cause. Pair the small scorecard with a short exception list and a decision section. The assistant may assemble and reconcile the evidence, but a manager remains responsible for interpreting it, changing policy, coaching people, or making a commitment to a customer or worker.

A practical launch is deliberately small. Give the assistant one narrow, low-consequence question and an example evidence table. Review the query interpretation before searching and the first few sources before full synthesis. Improve the brief when uncertainty comes from the instruction rather than the researcher. During this period, hold a short daily review around real records: what arrived, what was finished, what was returned, what is blocked, and which rule needs repair. Do not expand scope merely because the assistant has spare time. Expand when the current lane has stable inputs, a usable stop rule, proportionate access, and evidence that the manager can review without reconstructing the work from memory. Add one new task class or permission at a time. Record the date, reason, owner, and rollback condition so a failed expansion can be reversed cleanly.

Managers often ask, "How detailed should an assistant research brief be?" The answer should come from the written authority map, not from job title or seniority. List the decisions the assistant may make, the decisions they may prepare for approval, and the decisions they must immediately escalate. Attach examples to each category and state whether communication may be sent or only drafted. Revisit the map when systems, policies, customer promises, or the role itself changes. A reliable assistant can exercise strong judgment inside a defined lane; the company still needs accountable owners for money, security, employment, privacy, policy, and unusual customer outcomes. Clear boundaries make initiative safer because the next valid action is visible.

Before recruiting, turn this workflow into a one-page role brief. Include the recurring volume, schedule and time-zone needs, systems, required experience, communication cadence, success evidence, access limits, and the first-week sample. Link the matching service description at /services/research-assistance and use /contact-us to request a role plan when the lane is specific enough to discuss. OutsourcedAssistants.com describes Philippines-only talent sourcing and does not claim that a checklist guarantees a hiring or business result. The useful promise of this preparation is more modest: candidates and managers can examine the same work, boundaries, and review method before access is widened. That makes the staffing conversation concrete and gives the first two weeks a fair, inspectable starting point.

Build the work lane

• Write the recurring task and its finish rule
• Share an approved example and the source system
• Set response times, approval limits, and escalation rules
• Review a small sample before widening the role

Check the role before you expand it

• Can a new person complete the task from the examples?
• Who reviews the first week and records corrections?
• Which decisions and systems stay with your internal owner?
• What evidence shows the role is ready for another queue?

Common questions