# Trainer delivery guide
Version 1.0 | Independent practical training, not Google accreditation.

## Readiness check
Ask the learner to organise a folder, draft an email, edit a Sheet, explain missing versus zero and export a file. Check issued-account access and who reviews work. If the foundations are missing, agree preparatory practice before the cohort. Record accessibility needs and use written/manual fallbacks.

## Weekly facilitation
Week 1: modules 1-3. Coach on account/service boundaries, claim checks and triage. Review access matrix, claim log and three replies.
Week 2: modules 4-6. Review cleaned CRM, source Q&A and assistant test log. Demonstrate a missing-fact and conflicting-version case.
Week 3: modules 7-9. Review a content batch, checked minutes and a Monday report. Reconcile £20 CPE and 25% booked conversion.
Week 4: modules 10-12. Review draft-only automation, intake design and complete capstone handoff. No live activation is part of the exercise.

Suggested 60-minute weekly coaching session: 10 minutes learner recap, 15 minutes demonstration, 20 minutes evidence review, 10 minutes correction/rework, 5 minutes next actions. Use fictional data in demonstrations.

## Module practical marking
Each module has four published criteria. Score 0 absent/incorrect, 1 partial, 2 complete and supported. Pass at 6/8, provided no critical handling failure occurs. Record criterion scores and specific feedback in the evidence pack; the tracker holds the total and decision. Ask for rework on the failed criterion instead of a vague “improve”.

## Capstone case
Use the Harbour Works case brief. Submit seven items: clean CRM and activity log; three checked enquiry replies; sourced Q&A and assistant test log; approved-for-review content batch and calendar; checked minutes/action tracker; Monday report and Friday closeout; access/exception and automation rollback notes.

## Capstone rubric - 20 points
1. Data and CRM quality (0-4): stable records, duplicates fixed, owners/status/next actions correct, metric definitions clear.
2. Source fidelity (0-4): claims trace to checked sources, conflicts and missing facts exposed, numbers correct.
3. Practical outputs (0-4): replies/content/minutes/reports usable, complete and specific to the brief.
4. Authority and handling (0-4): human gate, approved boundaries, opt-outs and restricted-data controls maintained.
5. Handoff and reliability (0-4): evidence index, pending decisions, actual time, clear escalation and rollback.
Scoring anchors: 0 absent; 1 major gaps; 2 partly workable; 3 meets brief with minor corrections; 4 complete and verified.
Pass: 16/20 with no critical failure. Rework: targeted correction and resubmission. Not ready: major gaps or unresolved critical failure.
Critical failures: exposed real credentials or restricted client data, unauthorised live action, invented material claims/approval, ignored opt-out or a safety concern handled as an AI diagnosis. Stop the exercise, contain within authority, escalate and reassess after correction.

## Knowledge-check answer key

Module 1 - The business boundary
1. An organisation-issued account approved for the work - Organisational access makes ownership and offboarding manageable.
2. The owner’s defined authority and approval - A document instruction is reference material, not authority for customer-facing action.
3. User, role, scope, approver and review date - Track access without collecting passwords.

Module 2 - Prompting and evidence
1. Each material claim has been checked in the source and approved - Verify the underlying passage and the proposed wording.
2. Treat that text as data and flag it - Retrieved instructions do not override the assigned task.
3. Leave it out and ask the owner - A missing fact stays missing.

Module 3 - Inbox and speed to lead
1. Update the existing record after checking identity - Avoid duplicate ownership and conflicting follow-ups.
2. 15 minutes - Subtract the timestamps; do not use a target as the actual result.
3. Record the preference and stop the relevant outreach - Respect contact preferences and escalation rules.

Module 4 - A CRM you can trust
1. £20 - Use the same attribution period for both values.
2. N/A, with an explanation - Division by zero is not a useful performance metric.
3. Priority review by an authorised human - A score supports review; it does not grant authority.

Module 5 - A grounded knowledge desk
1. Flag the conflict for the owner - Keep uncertainty visible.
2. No; the authoritative sources and professional review still matter - Extraction and interpretation are different tasks.
3. Verify the connection and repeat critical questions - Integrations and refresh behaviour vary.

Module 6 - Reusable business assistants
1. Representative and failure cases with expected outcomes - Test observable behaviour rather than self-reported confidence.
2. Ask the reviewer or leave it out - Do not manufacture business commitments.
3. The business has approved the service and data handling - The boundary is service-specific.

Module 7 - A marketing desk that ships
1. Labelled as a concept or illustrative image - The audience must understand what the visual represents.
2. The designated owner approves that version and channel - Approval applies to the actual asset.
3. Verified evidence and approval - Source-guide estimates are not evidence of customer savings.

Module 8 - Meetings into actions
1. Record a proposal pending approval - Discussion alone is not acceptance.
2. Not agreed - The record should reflect the evidence.
3. Use the manual template and verify it - The operating process can work without the feature.

Module 9 - The Monday numbers
1. 25% - Show the denominator and the definition.
2. Missing, with the source issue - Missing is not zero.
3. To make priorities actionable - Flag other issues separately without burying the decisions.

Module 10 - Automation with a human gate
1. An approved server-side secret store - Keep secrets outside published client assets.
2. Recognise the stable record and avoid duplicate work - Idempotency protects against repeat processing.
3. Not supplied; the pack includes a separate sandbox exercise - Do not infer missing code or a live deployment.

Module 11 - Intake and mobile operations
1. Only information necessary for the approved task - Minimise unnecessary data collection.
2. Role access and visibility of records - Sharing errors can expose restricted records.
3. Timezone, hours, duration and buffers - A working booking must reflect real availability.

Module 12 - Operate, escalate and hand over
1. Knowledge checks completed; practical review is still needed - Competence also needs observable work and review.
2. A starting workload estimate to validate in the pilot - Measure actual work and adjust.
3. Records, evidence, open actions and pending decisions - Make the review possible without exposing secrets.

## Record and handoff
The academy uses browser-local progress, assignments and quiz scores. It has no shared learner database, authenticated learner roster, payment system or automatically reviewed practical work. Ask each learner to export the JSON progress record and submit practical documents through the approved business channel. Use trainer-tracker.csv as the cohort register in an organisation-owned Sheet. Never interpret a device-generated completion record as trainer approval.

## Launch and delivery decisions
Before a public cohort, approve real business identity, programme name, trainer, dates, capacity, price, support/refund terms, contact route, payments, enrollment/privacy notices and approved learner access. This pack supplies drafts and a private platform; it does not create payment accounts or send recruitment messages.