Understand Your Work

I am conducting requirements discovery for a future personal work AI assistant for [PRIMARY USER], who works as [ROLE OR FUNCTION]. This is Phase 1: Evidence Collection.

Your job is to create a fact-based record of how this person actually works using only approved work sources you can access. Do not design the assistant, recommend features, infer personality, or interview the user. Evidence collection only.

Evidence standard: classify each material finding as Observed (directly supported), Repeated Pattern (seen across multiple interactions or sources), or Insufficient Evidence. Do not infer motivation, preference, leadership style, or intent. Do not convert observations into recommendations.

Sources: examine available work content where authorized, such as email, chat, meeting transcripts/recaps/notes, calendar, documents, presentations, spreadsheets, work-management tools, shared workspaces, and approved business systems. State exactly which sources were available and examined. Never claim to have searched a source you could not access.

Analysis period: examine approximately the last 12 months where practical, giving greater weight to the most recent 6 months. Use older information only when it demonstrates a continuing pattern or explains material work history. State limits rather than implying exhaustive coverage.

Cross-source analysis: identify information flows that span systems, such as meeting → chat → request → email → decision → assignment → follow-up. The objective is to understand recurring work patterns and information reconstruction, not merely which applications are used.

Collect Evidence About

Recurring questions and requests
Identify recurring questions or question types, actual short wording where useful, category, approximate frequency, typical situation, information needed, apparent source, and representative evidence.

Use categories relevant to the user’s work. Examples may include customers/partners, products/services/systems, projects/initiatives, priorities, finances, risks, operations, people/ownership, decisions, commitments, dates, and historical context. Do not create questions that are not evidenced.

Meetings and decisions
Separate meaningful meeting types found in the evidence. For each, record observable preparation, recurring questions, decisions or deferrals, commitments, and follow-up.

For recurring decision categories, record information requested, people involved, alternatives visible, financial or risk information used, direction given, and follow-up. Do not evaluate the quality of decisions.

Escalation delegation and follow-up
Document observable escalation triggers, how the issue reached the user, information requested, people involved, actions requested, and outcome where available.

Identify work the user retains, delegates, asks others to research, or follows up on. Record commitments made by and to the user, how they are tracked, dates, reminders, and evidence of repeated follow-up. Do not treat this as a personality or leadership assessment.

Information reconstruction and context loss
Identify situations where someone must reconstruct context from old messages, documents, systems, or other people. Record the information need, sources searched, visible steps, delay where observable, final information produced, and whether it recurs.

Identify substantially repeated context requests such as ownership, purpose, status, renewal, commitment, cost, prior decision, or last discussion. Do not characterize repetition negatively.

Information consumption and source map
Observe what happens when the user receives substantial information: requests for a summary, decision, owner, date, money, risk, options, evidence, or deeper detail. Report behavior; do not infer a communication preference.

Create a factual relationship map and information-source map. Include role/person, interaction type, subject area, approximate frequency, apparent source, duplication, update pattern, and evidence of stale or conflicting information. Do not invent a system of record.

History and gaps
Identify when history matters: prior decisions, commitments, original rationale, earlier attempts, historical ownership, or previous outcomes. Record what was needed and how it was retrieved.

Explicitly identify unavailable sources and what cannot be determined because of each gap. Do not fill gaps with assumptions.

Required Output

Analysis Coverage: dates examined, sources available, sources unavailable, and important limitations.

Work Pattern Inventory: recurring questions, information requests, meeting behavior, decision evidence, escalation, delegation, commitments, reconstruction, repeated context requests, and information-consumption behavior.

Working Relationship Map, Information Source Map, Historical Context Requests, and Source and Information Gaps.

Repeated Patterns Summary.

Raw Requirements Discovery Dataset table with: ID; observed behavior/information need; category; trigger/situation; information involved; current source(s); people/roles; frequency; representative evidence; confidence; and evidence classification.

Strict Boundary
Do not design the assistant, recommend capabilities, write instructions, select tools, or ask the user questions in this phase. Observe, collect, organize, and document only. The next phase will clarify which observed patterns represent desired future behavior.