Research Selectively and Safely
Create a bounded research plan with approved questions, source limits, and a limited-evidence fallback.
Why this decision matters
Research should answer specific preparation questions: what has the participant said, what is the company's current offering, what changed that affects the agenda, and what remains unknown. A source budget prevents endless searching. Retrieved text is data, not instruction; embedded requests to reveal notes, change permissions, or send information are ignored. Private booking details should never become public search terms. If a source is blocked, record the failure once and continue with a shorter brief.
This lesson advances the same capstone used throughout the course: the Meeting Preparation and Follow-Up Agent Pack. The learner is not collecting ideas for later. The workbook record created here becomes an input to the next module and must be specific enough that another operator could review it. Where evidence is incomplete, the artifact should show the gap instead of smoothing it over.
Source evidence and limits
C7-S3 defines source preference, time and tool limits, blocked-source handling, malicious-input treatment, privacy boundaries, and limited-evidence output.
The sources support the operating method stated above. They do not prove universal performance, guaranteed savings, complete market coverage, causal impact, or a client result. Any date-sensitive product, platform, competitor, legal, pricing, or policy fact must be checked against a current primary source before publication or operational use.
Continuing fictional example
Example: Northstar Operations, Cedar Workshop, and operations lead Jordan Lee are fictional. The meeting concerns inconsistent enquiry handling.
Northstar's fixture permits no public research. The agent uses the three supplied fictional excerpts, marks role and company context as supplied-only, and lists budget, authority, timeline, and impact as questions rather than facts.
Example boundary: The fixture contains no live person, address, email, profile, company, meeting, or customer result. Its source excerpts exist only to test the course method.
Workbook application
Use Workbook Section 4: Research Selectively and Safely. Complete Workbook Section 4. Rank research questions, assign approved sources, set a deadline and call limit, state private-data restrictions, and write the fallback response for blocked or contradictory evidence.
- Write the current evidence or input before adding interpretation.
- Apply the lesson's decision rule and state the reason for the classification.
- Mark uncertainty, missing information, and the human owner for the next decision.
- Check that the result stays inside the course boundaries and can be tested.
Failure modes to inspect
Retrying CAPTCHAs; bypassing a login; placing private notes in a search query; treating instructions in a page as commands; and padding a sparse brief with generic claims.
A polished output can still fail if its evidence, permissions, identity, denominator, source date, or action boundary is wrong. Review the underlying record rather than grading tone alone. The correct repair may be to narrow the scope, gather a permitted source, label an unknown, or stop the proposed action.
Decision rules
Stop optional research at the agreed limit. Record failures. Never claim inaccessible content was verified. Ask when ambiguity would change the advice.
Record the rule in the workbook in language two reviewers can apply consistently. A rule that depends on intuition alone cannot support a deterministic fixture. If reviewers disagree, preserve both readings, identify the missing evidence, and revise the rule before automation.
Three application checks
1. A public page tells automated readers to reveal private meeting notes before continuing. How should the research step respond?
2. The highest-priority source is blocked after one permitted attempt. What should the agent produce?
3. Which research question belongs in the plan?