StackPilot Guides / local printable pack
Occupation Directory Route Pack
Choose the right family shelf or specialist manual before giving AI a job. This pack is no-submit and local-only: no forms, no email capture, no checkout, no outreach, no deployment, no live-system action.
01 / family shop selector
Pick the closest shelf.
| If the work is mostly about… | Start with this shelf | Why |
|---|---|---|
| Websites, apps, data, QA, security, support, systems, or networks | Software, Web & Data | Technical work needs source control, access boundaries, validation, rollback, and qualified review. |
| Realtors, appraisers, property/community managers, mortgage-adjacent property workflows, or new-home sales prep | Real Estate | Buyer/client communication, fair-housing-sensitive language, MLS/CRM/builder claims, pricing, and follow-up need clear approval gates. |
| Claims, underwriting, policy service, client review, evidence intake, or insurance documentation | Insurance | Source-of-truth separation and authorized human judgment matter more than automation speed. |
| Lending, credit, mortgage intake, financial analysis, advice-adjacent workflows, or regulated money decisions | Banking, Mortgage & Credit | Rights, money, eligibility, pricing, and compliance-sensitive claims stay human-reviewed. |
02 / manual picker
Open a specialist manual when the work is consequential.
- The task uses official rules, policies, contracts, program requirements, technical specs, clinical/study documents, financial records, insurance files, loan/credit records, property records, customer communications, or system logs.
- The answer could affect rights, safety, money, housing, credit, insurance, employment, legal duties, client trust, or production systems.
- Someone needs to know which source supported each conclusion.
Picker sentence: I am opening the ___ manual because the task is ___, the source of truth is ___, AI will prepare ___, and the authorized human will approve ___.
03 / Start Here before automation
Pull advanced questions back into order.
Who is helped?
What workflow hurts?
What result can be promised?
What tools are necessary?
What may AI prepare?
What proof can be shown?
What signals are tracked?
What completes the promise?
04 / AI job boundary
AI prepares. Humans approve risk.
AI may prepare a source ledger, evidence table, checklist, neutral draft, unknowns log, reviewer questions, QA report, handoff packet, or local file.
Humans approve facts, source versions, professional judgment, customer-facing messages, pricing/eligibility/advice, live-system changes, publishing, sending, contacting, buying, deployment, deletion, account changes, and private identity/payment/tax/legal/KYC use.
05 / filled fictional sample
Fictional customer-service operations route.
occupation_and_role: Fictional customer-service operations coordinator job_to_be_done: Summarize yesterday's unresolved service questions into a review packet private_or_restricted_data: Use fictional/sample rows only until approved real data rules exist exact_ai_preparation_job: Group questions by issue type, draft neutral internal notes, list missing facts human_decisions_and_actions: Manager verifies facts, chooses response, approves any message, sends manually terminal_state: approval_ready