Free Starter Kit · Download, Copy, or Print
The “Start Monday” AI Starter Kit.
Our first year of AI was full of shiny demos nobody used. What actually changed our organization was boring and subtractive: saving 10 to 15 minutes at a time, lowering software bills, and training regular staff to use AI with wisdom. Everything below is ready to use with your team on Monday morning.
Part 01
Five boring AI workflows (ready-to-run SKILL.md files)
Each workflow below is packaged in the open Agent Skills (SKILL.md) format used by Claude, ChatGPT/Codex, Gemini, and Cursor. Download the file, copy it to your clipboard, or grab the full .zip pack above.
How to use a SKILL.md file in 60 seconds
.md file inside a Claude Project. .md file inside a ChatGPT Project, or paste it into Custom GPT instructions. .md file as instructions inside a Gemini Gem. .claude/skills/ or .agents/skills/. - 01
The Meeting-to-Decision Filter
Exec teams, pastors, & department leads
Drop a meeting transcript or post-meeting voice memo into this skill to extract three things only: decisions with owners, open questions, and who needs to be looped in before Friday. Prevents silent decision collisions across departments.
Preview
meeting-to-decision-filter/SKILL.md--- name: meeting-to-decision-filter description: Extracts decisions with owners, open questions, and who needs to be looped in from any meeting transcript or post-meeting voice memo. Use after executive, pastoral, staff, or project meetings to prevent silent decision collisions. license: CC-BY-4.0 metadata: author: Paul Yerrick (paulyerrick.com / COKI Studio) version: "1.0" category: executive-operations source: https://www.paulyerrick.com/ai/ --- # The Meeting-to-Decision Filter Turn a messy meeting transcript or a two-minute post-meeting voice memo into a crisp alignment summary containing **three things only**: decisions made, open questions, and who was not in the room who needs to be looped in. ## When to Use This Skill - Right after an executive team, pastoral leadership, elders, board, or cross-department planning meeting. - After a 1-on-1 or project check-in where multiple people could walk away remembering different outcomes. - When you have a raw transcript (Zoom, Teams, Fireflies, Otter, Apple Notes) or a rough two-minute voice memo recorded right after the meeting. ## When Not to Use This Skill (Red Lines) - **Never paste confidential pastoral care, counseling, HR discipline, payroll, donor giving records, or minors' information** into unapproved or consumer AI tools with model training enabled. - Strip personal names and sensitive identifying details before uploading whenever a meeting touches sensitive personnel or pastoral matters. - Never use this skill to replace direct human follow-up on a difficult relational or leadership conversation. ## Inputs Needed Ask the user to paste or attach: 1. **Meeting transcript or post-meeting voice memo** (required) 2. **Meeting name and date** (optional, infer if present) 3. **Key stakeholders or departments not in the room** (optional, helps flag who needs to be looped in) If the user gives a very brief note without clear owners or deadlines, extract what is explicit and clearly mark missing owners or dates as `[Owner needed]` or `[Date needed]` rather than guessing. ## Step-by-Step Instructions 1. **Read for alignment, not play-by-play.** Ignore small talk, tangents, and circular debate. Do not write a chronological recap. 2. **Identify confirmed decisions.** Only list items where the group actually landed on a direction. Separate firm decisions from tentative ideas. 3. **Attach an owner and target date to every decision.** If no owner was explicitly named in the transcript, flag it as `[Owner needed]` so leadership closes the loop before leaving the week. 4. **Capture unresolved open questions.** List every decision that was deferred, blocked, or left ambiguous, along with what information or person is needed to resolve it. 5. **Run the "Not in the Room" check.** Scan the decisions for downstream impact on other teams (communications, campus operations, worship/production, finance, HR, volunteers, elders, or external partners) and list exactly who needs a heads-up and why. 6. **Check for silent decision collisions.** If two statements in the transcript appear to contradict each other on scope, timeline, or ownership, call it out briefly in a single alert line under Open Questions. 7. **Keep it scannable.** The full output must fit comfortably in a single Slack message, Teams post, or short email that takes under 90 seconds to read. ## Output Format Use this exact markdown structure: ```markdown ## Meeting-to-Decision Summary **Meeting:** [Meeting Name] | **Date:** [YYYY-MM-DD] ### 1. Decisions Made - **[Decision in plain English]** - Owner: [Name or Owner needed] - Target date: [Date or Date needed] - Context (1 line max): [Why this was decided, only if needed] ### 2. Open Questions (Needs Clarity) - **[Unresolved question or potential collision]** - Needed to resolve: [Person, data, or next meeting] ### 3. Who Needs to Be Looped In (Before Friday) - **[Person or Team]:** [1-sentence summary of what affects them] (Assigned to notify: [Name or Owner needed]) ``` ## Voice and Style Rules - Write in plain, grounded English. No corporate jargon, no filler, no hype. - Never invent decisions, owners, or dates that are not in the source text. - Never use em-dashes or emojis. Use periods, commas, colons, or parentheses instead. - Keep every bullet to one or two lines maximum. ## Quick-Start Example **Example Input (post-meeting voice memo):** > Just wrapped the Q3 calendar meeting with Mark, Sarah, and Devon. We decided to move the Volunteer Appreciation Night from October 18 to October 25 so it does not collide with fall break. Sarah is going to lock the venue this week. We still have not decided whether childcare will be provided on site or reimbursed, and Mark needs to get budget numbers from finance by Thursday. Also we talked about launching the new website landing page two weeks early, which probably affects the comms team and campus pastors who were not in the room. **Example Output:** ```markdown ## Meeting-to-Decision Summary **Meeting:** Q3 Calendar Planning | **Date:** [Date needed] ### 1. Decisions Made - **Move Volunteer Appreciation Night from October 18 to October 25.** - Owner: Sarah (venue confirmation) - Target date: This week - Context: Avoids overlap with fall break. ### 2. Open Questions (Needs Clarity) - **Will childcare be provided on site or reimbursed?** - Needed to resolve: Budget numbers from finance (Mark to gather by Thursday). - **Is the two-week early launch for the new website landing page confirmed or still a proposal?** - Needed to resolve: Leadership sign-off before comms adjusts its schedule. ### 3. Who Needs to Be Looped In (Before Friday) - **Communications Team:** Potential two-week shift on the new website landing page launch and updated October 25 event date. (Assigned to notify: [Owner needed]) - **Campus Pastors:** Volunteer Appreciation Night moved to October 25. (Assigned to notify: [Owner needed]) - **Finance:** Needs to provide childcare budget numbers to Mark by Thursday. (Assigned to notify: Mark) ``` - 02
The Reusable Playbook Builder
Operations, ministry coordinators, & comms teams
Interviews any staff member about a recurring weekly or monthly task and turns it into a clean one-page team playbook (goal, inputs, guardrails, copy-paste prompt, and human review checklist) so nobody starts from scratch.
Preview
reusable-playbook-builder/SKILL.md--- name: reusable-playbook-builder description: Interviews a staff member about a recurring weekly or monthly task and converts it into a standardized, one-page Reusable Playbook document (goal, inputs, guardrails, output format, and quality check). Use whenever a team wants consistent, on-brand AI drafts without reinventing prompts from scratch. license: CC-BY-4.0 metadata: author: Paul Yerrick (paulyerrick.com / COKI Studio) version: "1.0" category: operations-enablement source: https://www.paulyerrick.com/ai/ --- # The Reusable Playbook Builder Turn any recurring team task (volunteer briefings, event launch checklists, quarterly board summaries, sermon follow-up emails, onboarding notes) into a **one-page Reusable Playbook** that anyone on your staff can run in five minutes with consistent quality. ## When to Use This Skill - A staff member repeats the same writing, summarizing, or planning task every week or month. - Different team members are typing improvised prompts from scratch and getting wildly inconsistent quality. - An AI Champion wants to document a proven workflow so a non-technical teammate or new hire can run it safely. ## When Not to Use This Skill (Red Lines) - Do not build playbooks that simulate personal pastoral care, grief responses, counseling, or personal relational messages. - Do not build playbooks that require staff to paste confidential pastoral, HR, counseling, donor, or minors' data into unapproved consumer AI tools. - Every playbook must end with a named human review step. Never design a workflow that publishes directly without human ownership. ## Inputs Needed If the user has not already supplied these details, ask **at most 5 quick, plain-English questions** before generating the playbook: 1. **Task name and frequency:** What recurring task are we turning into a playbook, and how often does it happen? 2. **Audience and goal:** Who reads the final output, and what does "good" look like in one sentence? 3. **Typical inputs:** What raw notes, dates, transcripts, or reference docs will a staff member paste in? 4. **Voice and red lines:** What tone should it sound like, and are there words, clichés, or private details to avoid? 5. **Sample (optional):** Do you have one past example of this document done well? ## Step-by-Step Instructions 1. **Confirm scope first.** If the user gave a rough description, ask only the missing questions from the list above in a single short message. Do not overwhelm non-technical staff with technical jargon. 2. **Design for subtraction.** Make sure the playbook genuinely saves at least 15 minutes of blank-page or formatting friction and leaves judgment, voice, and final approval with a human. 3. **Write the one-page Reusable Playbook** with six clear sections: Goal, When to Use, Required Inputs, Guardrails and Voice Rules, Step-by-Step Prompt Block (ready to copy-paste), and Human Review Checklist. 4. **Include a clean copy-paste prompt block** inside the playbook so any teammate can paste it into their approved team AI workspace along with their raw notes and get an immediate first draft. 5. **Keep it to one page.** A playbook nobody reads is useless. Aim for 350 to 500 words total. 6. **Offer a test run.** After generating the playbook, invite the user to paste one real set of raw notes so you can test and tighten the playbook together right away. ## Output Format Use this exact markdown template for the finished playbook: ```markdown # Playbook: [Task Name] **Department:** [Team] | **Frequency:** [Weekly / Monthly / Per Event] | **Owner:** [Role or Name] ## 1. Goal [One clear sentence describing what a great output accomplishes.] ## 2. Approved Tool and Privacy Rule - **Use only in:** [Approved team workspace with training turned off] - **Never paste:** [Specific sensitive data to strip before running] ## 3. Inputs Needed Gather these before you start: - [Input 1, e.g., Event brief or meeting notes] - [Input 2, e.g., Dates, locations, and registration link] - [Input 3, e.g., Target audience] ## 4. Guardrails and Voice - **Tone:** [2 to 3 plain descriptors, e.g., warm, clear, direct] - **Avoid:** [Words, hype, formatting habits, or theological/factual assumptions to skip] ## 5. Copy-and-Paste Prompt > You are helping our [Department] team draft [Task Name] for [Audience]. > Goal: [One-sentence goal]. > Voice: [Tone descriptors]. Avoid [list of things to avoid]. Do not invent dates, names, or facts not provided below. If anything required is missing, mark it as [NEEDED]. > Format the output with these exact sections: > - [Section 1] > - [Section 2] > - [Section 3] > > Here are my raw notes: > [PASTE NOTES HERE] ## 6. Human Review Checklist (Before Publishing) - [ ] All dates, links, names, and numbers verified by a human - [ ] Tone sounds like our real team, not generic AI - [ ] Zero private or sensitive details exposed - [ ] Named human owner approved final version ``` ## Voice and Style Rules - Write for a busy non-technical coordinator, not a prompt engineer. - Never use em-dashes or emojis. - Keep instructions concrete, short, and immediately usable on a Tuesday morning. ## Quick-Start Example **Example User Request:** > Help me turn our weekly Kids Ministry Volunteer Briefing email into a Reusable Playbook. Right now I spend 45 minutes every Thursday writing it from scratch using the curriculum notes, room assignments, and one reminder from our weekend service lead. **Expected Behavior:** Ask 1 to 2 brief clarifying questions if tone/red lines are missing (for example, confirming no children's names or family medical notes are ever pasted in), then generate a crisp one-page `Playbook: Weekly Kids Ministry Volunteer Briefing` following the exact 6-section template above and offer to test it with this week's notes. - 03
The Subscription Audit
Executive pastors, COOs, & finance leads
Analyzes a 12-month software expense export to spot overlapping or single-purpose SaaS tools (transcription, basic formatting, simple summaries) that can be consolidated into one approved AI workspace so you save budget first.
Preview
subscription-audit/SKILL.md--- name: subscription-audit description: Analyzes a 12-month software and SaaS expense export to identify redundant, overlapping, or narrow single-purpose tools that can be consolidated into one approved team AI workspace. Use before adding any new AI budget line item to save money first. license: CC-BY-4.0 metadata: author: Paul Yerrick (paulyerrick.com / COKI Studio) version: "1.0" category: finance-and-operations source: https://www.paulyerrick.com/ai/ --- # The Subscription Audit Audit your organization's last 12 months of recurring software charges to spot narrow, single-purpose tools and overlapping subscriptions so you can **save budget before spending a dollar on new AI software**. ## When to Use This Skill - Before requesting budget for a new team AI workspace (Claude Team/Enterprise, ChatGPT Team/Enterprise, Gemini Workspace). - During annual budgeting or a quarterly operations review by an Executive Pastor, COO, Finance Lead, or Operations Director. - Whenever multiple departments have quietly expensed separate small SaaS tools on company cards. ## When Not to Use This Skill (Red Lines) - **Remove all bank account numbers, credit card numbers, tax IDs, personal employee names, and donor/payroll records** before pasting or uploading your CSV/spreadsheet. - Only run this audit inside an approved team AI workspace with model training disabled. - Never cancel mission-critical security, accounting, ChMS, CRM, or compliance tools without department lead sign-off. Treat this output as a structured review sheet for human leaders. ## Inputs Needed Ask the user to paste or upload: 1. **A 12-month software/SaaS charge list** (CSV, spreadsheet table, or copied lines showing vendor name, monthly or annual cost, and department if known). 2. **Primary core stack** (optional, e.g., Google Workspace or Microsoft 365, Planning Center / Rock RMS / Salesforce, Slack, etc.). 3. **Target approved AI workspace** (optional, e.g., Claude Team, ChatGPT Team, or undecided). If cost periods are mixed (monthly vs. annual), normalize every line item to an estimated annual spend and clearly mark any estimates. ## Step-by-Step Instructions 1. **Normalize the list.** Group repeated monthly charges by vendor and calculate annualized spend per vendor. Flag duplicate subscriptions across departments (for example, three different teams paying for separate transcription or PDF tools). 2. **Classify every tool into one of four buckets:** - **Consolidate Now:** Narrow single-purpose tools whose main job is already handled well by a single approved AI workspace or existing core suite (basic transcription apps, simple summarization plugins, standalone grammar/rewriting utilities, lightweight copy generators, basic PDF chat tools, duplicate note-takers). - **Check for Overlap:** Categories where the organization is paying for two or more competing tools (multiple project trackers, form builders, survey tools, scheduling apps, or stock/design utilities). - **Keep (Core Infrastructure):** Mission-critical systems of record (ChMS, CRM, accounting, payroll, email/identity suite, specialized production/broadcast software). - **Needs Clarification:** Unrecognized vendor names or unclear usage where finance needs to ask the cardholder what it does. 3. **Calculate conservative savings.** Estimate low-end and high-end annual savings from the Consolidate Now and Check for Overlap buckets. Never inflate savings numbers. 4. **Build a 30-day action plan.** Give the leader a simple checklist of who to email, what to verify before canceling (data export, contract renewal date), and how the savings compare to funding one approved team AI workspace. ## Output Format Use this exact markdown structure: ```markdown ## Software Subscription Audit Summary **Total Vendors Reviewed:** [Count] | **Total Annualized Spend:** [$Amount] **Estimated Recoverable Annual Savings:** [$Low to $High] ### 1. Consolidate Now (High-Confidence Candidates) | Vendor | Annualized Cost | Category / Function | Consolidation Path | Risk Level | | --- | --- | --- | --- | --- | | [Vendor] | [$Amount] | [Function] | [Replace with approved AI workspace or existing suite] | Low / Medium | ### 2. Check for Overlap (Multiple Tools Doing the Same Job) - **[Category, e.g., Meeting Transcription]:** Paying for [Tool A ($X/yr)] and [Tool B ($Y/yr)]. Recommend standardizing on one. ### 3. Keep (Core Infrastructure) - [Brief comma-separated list of core systems left untouched] ### 4. Unknown / Needs Clarification - **[Vendor ($Amount/yr)]:** Confirm owner, active seats, and renewal date. ### 5. Next-Step Checklist (Next 14 Days) 1. Confirm active usage and export any saved files/transcripts before canceling any tool in Section 1. 2. Check annual renewal dates on Section 2 overlaps and set calendar reminders 30 days ahead. 3. Reallocate recovered budget toward one approved, privacy-protected team AI workspace. ``` ## Voice and Style Rules - Be practical, conservative with savings estimates, and respectful of department needs. - Never guess what an unfamiliar vendor does without labeling it clearly as `[Needs verification]`. - Never use em-dashes or emojis. ## Quick-Start Example **Example Input:** > Vendor list (monthly unless noted): Otter.ai Pro ($16.99 x 4 seats), Fireflies.ai ($19 x 2 seats), Grammarly Business ($180/yr x 5 seats), Jasper AI ($49/mo), Planning Center ($299/mo), Canva Teams ($300/yr), Loom Business ($15 x 3 seats), Unknown SaaS "DocSum" ($29/mo). **Expected Behavior:** Normalize all items to annual spend, flag Otter + Fireflies + DocSum + Jasper (and review Grammarly) as Consolidate Now / Overlap candidates totaling over $2,800/yr in potential recoverable spend, leave Planning Center and Canva in Keep, flag DocSum under Unknown/Consolidate, and output the clean decision table plus 14-day cancellation/export checklist. - 04
The Creative First-Draft Multiplier
Creative, comms, & marketing staff
Keeps humans in charge of vision, theology, and final craft. Takes one finished sermon or keynote transcript and produces small-group questions, a midweek recap email draft, five social clip hooks, and a staff summary.
Preview
creative-first-draft-multiplier/SKILL.md--- name: creative-first-draft-multiplier description: Turns one finished sermon transcript, executive keynote, or leadership article into grounded first-draft derivatives (small-group discussion questions, midweek recap email, social clip hooks, and a plain-English staff summary) while keeping humans in charge of voice, theology, and final craft. license: CC-BY-4.0 metadata: author: Paul Yerrick (paulyerrick.com / COKI Studio) version: "1.0" category: creative-and-communications source: https://www.paulyerrick.com/ai/ --- # The Creative First-Draft Multiplier Take one finished piece of real human communication (a Sunday message transcript, executive keynote, or leadership memo) and eliminate blank-page friction by generating four grounded first drafts for your team to edit in their own voice. ## When to Use This Skill - After a pastor, executive, or speaker has already delivered or written the core message. - When a communications, ministry, or marketing team needs Monday morning starting drafts for small groups, email, social clips, and internal staff alignment. ## When Not to Use This Skill (Red Lines) - **Never use this skill to ghostwrite a pastor's sermon or a leader's original convictions from scratch.** Real human thinking and embodied voice come first; this skill multiplies finished work. - **Never invent Scripture quotes, statistics, personal stories, or theological claims** that are not present in the source transcript. - **Never publish raw AI output.** A human editor must review every line for accuracy, theological integrity, tone, and fit before anything goes public. ## Inputs Needed Ask the user to paste or attach: 1. **The finished transcript or manuscript** (required) 2. **Context:** Church/ministry or business/executive audience (optional, infer from transcript) 3. **Optional voice notes:** Any specific phrases your team avoids or a call to action to include in the midweek email (optional) ## Step-by-Step Instructions 1. **Stay strictly grounded in the source text.** Every takeaway, quote, Scripture reference, and clip hook must trace directly back to what the speaker actually said. If a Scripture verse or fact is referenced loosely in the transcript, flag it as `[Verify reference]` for the human editor. 2. **Preserve the speaker's actual voice.** Pull real language and memorable lines from the transcript instead of rewriting them into generic marketing clichés. 3. **Generate all four deliverables cleanly:** - **Deliverable A: Small-Group or Team Discussion Guide** (3 to 5 thoughtful, open-ended questions grounded in the talk, moving from observation to personal application). - **Deliverable B: Midweek Recap Email First Draft** (150 to 220 words, warm and human, pointing back to one core idea plus a clear placeholder for a link or next step). - **Deliverable C: Five Candidate Social Clip Hooks** (exact quote or moment from the transcript, suggested 1-line hook, and why it works as a 30 to 60 second clip). - **Deliverable D: Plain-English Staff Summary** (4 to 5 bullet points for staff or volunteers who were serving in other rooms and missed the message). 4. **Add a brief Human Verification Note** at the bottom listing any quotes, references, or names the editor should double-check before publishing. ## Output Format Use this exact markdown structure: ```markdown ## Creative First-Draft Multiplier Pack **Source:** [Talk or Message Title] | **Status:** First draft for human editing ### 1. Discussion Questions (Small Group / Team) 1. [Warm-up or observation question grounded in the message] 2. [Core tension or reflection question] 3. [Practical application question for this week] 4. [Optional deeper follow-up question] ### 2. Midweek Recap Email (First Draft) **Subject line options:** - Option 1: [Direct, non-clickbait subject] - Option 2: [Conversational subject] **Body:** [150 to 220 word draft in plain, grounded language. Include [LINK] placeholder where appropriate.] ### 3. Five Candidate Social Clip Moments 1. **Hook:** "[Short 1-line hook]" - **Transcript anchor:** "[Exact opening phrase from transcript so video editor can find it]" - **Why it works:** [1 short line] *(Repeat for 2 through 5)* ### 4. Plain-English Staff Summary (For People Serving in Other Rooms) - **Core idea:** [1 sentence] - **Key points:** [2 to 3 crisp bullets] - **Call to action mentioned on stage:** [Exact next step announced, or "None mentioned"] ### 5. Human Editor Verification Notes - [List any Scripture references, quotes, names, or dates to verify before publishing] ``` ## Voice and Style Rules - Sound like a real human leader, not an ad agency or a hype bot. - Avoid cliché openers ("In today's fast-paced world...", "Let's dive in...", "Game-changer"). - Never use em-dashes or emojis. ## Quick-Start Example **Example Input:** > Paste a 3,500-word transcript of a Sunday sermon titled "Wells, Not Towers" on Daniel in Babylon, along with a note: "Keep the email warm and direct, and include a placeholder link to our Sunday message replay." **Expected Behavior:** Produce the 4-part pack grounded strictly in quotes and themes from the transcript, surface exact transcript phrases so the video editor can jump right to the 5 best clip timestamps, and flag every Scripture reference for human verification. - 05
The Plain-English Staff Handbook Builder
Whole staff (non-technical friendly)
Asks six plain-English questions about your organization and drafts a tailored two-page beginner handbook with your approved tools, three safe role examples, three clear red lines, and your internal AI Champions directory.
Preview
plain-english-staff-handbook/SKILL.md--- name: plain-english-staff-handbook description: Drafts a tailored, two-page Plain-English Staff AI Handbook for a specific church, nonprofit, or company, including approved tools, three safe role-based examples, three unmistakable red lines, and internal AI Champions contacts. Use to replace staff fear and guesswork with clear, practical confidence. license: CC-BY-4.0 metadata: author: Paul Yerrick (paulyerrick.com / COKI Studio) version: "1.0" category: governance-and-training source: https://www.paulyerrick.com/ai/ --- # The Plain-English Staff Handbook Builder Create a **two-page beginner AI handbook** tailored specifically to your organization's real staff roles so non-technical teammates stop feeling intimidated or quietly guessing what is allowed. ## When to Use This Skill - Right after your leadership team, elders, or board adopts a one-page AI governance policy. - When onboarding new staff or launching an internal AI Champions group. - Whenever staff are either avoiding AI out of anxiety or using unapproved consumer tools on personal phones. ## When Not to Use This Skill (Red Lines) - Do not treat this handbook as legal counsel. Have leadership, HR, or legal review data and privacy language before distributing organization-wide. - Never publish a handbook that skips privacy rules, human ownership, or clear escalation paths for care/crisis situations. - Do not include real personal phone numbers or private logins in an unapproved AI tool; use placeholders like `[IT Helpdesk Link]` or `[Champion Name]` and fill them in locally if needed. ## Inputs Needed If the user has not already supplied these details, ask these **6 brief questions** in one message before drafting: 1. **Organization name and type:** (e.g., multi-site church, mid-sized SaaS company, nonprofit) 2. **Approved 1 to 2 AI tools:** (e.g., Claude Team, ChatGPT Enterprise, Google Gemini Workspace) and how staff request access 3. **Top 3 staff departments to tailor examples for:** (e.g., Ministry/Pastoral, Creative/Comms, Operations/Admin) 4. **Your 3 to 5 AI Champions (names or roles):** Who staff can walk up to with a quick question 5. **Your Governance Group or Pause Button owner:** Who approves new tools or pauses workflows 6. **Any specific organizational values or red lines** you want highlighted verbatim ## Step-by-Step Instructions 1. **Keep it strictly to two printable pages** (roughly 650 to 850 words). Nobody reads a 40-page technical manual. 2. **Start with reassurance, not rules.** Open with a short leadership posture statement naming what AI is for (subtracting administrative friction) and what AI will **never** replace (human presence, pastoral care, relationships, leadership discernment, and creative ownership). 3. **Include the four non-negotiable guardrails** in plain English: - Always disclose - Route care and crisis to real humans - Protect private data - Honor the Governance Group's pause button 4. **Write three concrete, safe everyday examples** pulled directly from the user's actual departments so a non-technical teammate immediately sees how it helps on a normal Tuesday. 5. **Write three unmistakable red lines** (what never goes into AI and what AI never does). 6. **Include a clean Green / Yellow / Red quick-reference box** and a **"Who to Ask" directory** naming the organization's AI Champions. ## Output Format Use this exact markdown structure: ```markdown # [Organization Name] Staff AI Handbook **Version:** 1.0 | **Last Reviewed:** [Month Year] | **Policy Owner:** [Governance Group] ## Why We Use AI (And What It Will Never Replace) [2 to 3 clear sentences framing AI as subtractive help for administrative friction, explicitly naming what remains 100% human.] ## Our Approved Tools - **Approved Workspace(s):** [Tool 1, Tool 2] (team/enterprise privacy enabled, model training turned off) - **How to get a seat:** [Exact step or contact] - **Rule of thumb:** Never use personal or free consumer AI accounts for organizational work. ## Four Non-Negotiable Guardrails 1. **Always disclose:** Never let AI pose as a human staff member, pastor, or leader. 2. **Route care and crisis to real humans:** Prayer requests, counseling, grief, safeguarding, or personal crisis always go immediately to a human. 3. **Protect private data:** Never paste confidential pastoral, HR, donor, payroll, legal, or minors' data into unapproved tools. Strip identifying names whenever summarizing notes. 4. **Human ownership and the pause button:** Every published piece has a human name on it. Our Governance Group has full authority to pause any workflow that drifts from our values. ## Three Safe Ways to Use AI This Week 1. **[Department 1 example]:** [Concrete 2-sentence everyday example] 2. **[Department 2 example]:** [Concrete 2-sentence everyday example] 3. **[Department 3 example]:** [Concrete 2-sentence everyday example] ## Three Unmistakable Red Lines (Never Do This) 1. **Never** paste private personal, pastoral, donor, HR, or minors' records into unapproved or free AI tools. 2. **Never** outsource personal relational care, apologies, counseling, or spiritual authority to a machine. 3. **Never** publish AI-generated facts, quotes, or Scripture references without verifying them yourself. ## Quick Traffic Light - **Green (Go):** Internal meeting summaries (names stripped), first-pass checklists, editing for clarity, brainstorming discussion questions inside approved tools. - **Yellow (Ask First):** New AI subscriptions, public-facing website tools, connecting AI to ChMS/CRM/email, public AI imagery or video. - **Red (Off-Limits):** Private data leaks, simulated human care, unverified public claims. ## Need Help? Ask an AI Champion You do not have to figure this out alone. Walk up to or message one of our internal AI Champions: - **[Department 1]:** [Name / Role] - **[Department 2]:** [Name / Role] - **[Department 3]:** [Name / Role] - **Governance / New Tool Requests:** [Contact or Governance Group] ``` ## Voice and Style Rules - Write at a clear, warm, non-technical reading level. Zero tech buzzwords (no "LLM orchestration," "token windows," or "agentic pipelines"). - Lower the temperature: cautious staff should feel protected, and curious staff should feel clearly guided. - Never use em-dashes or emojis. ## Quick-Start Example **Example Input:** > We are Grace Valley Church (45 staff). Approved tool: Claude Team (request a seat from Ops). Top 3 departments: Kids/Students Ministry, Creative/Comms, and Facilities/Ops. AI Champions: Jenna (Kids), Marcus (Comms), Dave (Ops). Governance Group: Executive Pastor, Lead Pastor, Comms Director, IT Director. **Expected Behavior:** Generate a complete, ready-to-print two-page handbook customized for Grace Valley Church with concrete examples for Kids/Students, Creative/Comms, and Facilities/Ops, zero jargon, and all four guardrails intact.
Part 02
The AI Champions Blueprint (90-Day Staff Rollout)
Top-down IT mandates and flashy all-staff demos almost always stall. Peer-led enablement sticks. Start with 3 to 5 curious, high-trust staff members who test real workflows first and teach their own teams in plain English.
1. Who to pick (3 to 5 Champions)
Look for curiosity and relational trust. Your best Champion is not the person reading AI newsletters every morning; it is the teammate others already walk up to for practical help.
- Operations or Administration: Schedules, meeting alignment, recurring checklists, vendor audits
- Creative or Communications: First-draft email, sermon/keynote derivatives, voice guardrails
- Ministry or Front-Line Team (1 to 2 people): Volunteer briefings, event logistics, peer translation
- Technology or Executive Partner: Privacy settings, workspace licensing, governance alignment
2. The 30-Day Champion Charter
Give your Champions four clear commitments for their first month:
- Meet 30 minutes every two weeks to compare what actually saved time.
- Test workflows inside real daily work only (zero futuristic side quests).
- Write down 2 to 3 one-page Reusable Playbooks per department using Skill 02.
- Flag any tool or habit that feels off-culture or threatens human care.
3. The 30 / 60 / 90-Day Schedule
| Phase | Focus | Concrete Deliverable |
|---|---|---|
| Days 1 to 30: Guardrails & Alignment | Align senior leadership and remove fear before scaling tools. | Adopt the one-page governance policy (Part 03), set up 1 to 2 approved workspaces with training off, and commission 3 to 5 AI Champions. |
| Days 31 to 60: Champions Build Playbooks | Test small, subtractive workflows inside real departments. | Throw away anything gimmicky. Document 5 to 10 one-page Reusable Playbooks that genuinely save 15 minutes or more per week. |
| Days 61 to 90: Plain-English Enablement | Train full staff in small, role-based groups instead of one lecture. | Distribute your 2-page Plain-English Staff Handbook and run 45-minute hands-on department labs led by your Champions. |
The Monthly Subtraction Test
“Did this give a human back time, margin, or clarity for work that actually requires a human?”
If yes, turn it into a shared one-page playbook. If it only looked impressive in a leadership meeting and nobody touched it two weeks later, kill it cleanly and move on.
Part 03 · Detachable One-Page Handout
One-Page Board-Ready AI Governance Policy
Bring this page to your next executive, elders, or board meeting. Fill in the adoption block at the bottom, publish it to your staff, and review it quarterly.
Always disclose
Never let AI pose as a human pastor, counselor, or leader. Label automated logistics tools clearly, and never outsource personal relational care or grief notes to a machine.
Route care and crisis to real humans
The gospel is incarnational, not informational. Any workflow touching prayer requests, spiritual struggle, counseling, safeguarding, or crisis must immediately escalate to a real human.
Protect private data
Never paste confidential pastoral, counseling, donor, HR, payroll, legal, or minors' data into free consumer tools. Use paid team workspaces with training turned off and strip identifying names.
Name an owner & a pause button
Appoint a cross-functional AI Governance Group (3 to 5 people across executive, pastoral/mission, comms, and operations/IT) with explicit authority to approve tools or pause any workflow that drifts.
Green, Yellow, and Red Zone Matrix
| Zone | Rule | Practical Examples |
|---|---|---|
| Green (Go) | Approved everyday workflows inside sanctioned tools; human reviews every output before publishing. | Meeting transcripts to decisions; first-pass event briefs and volunteer checklists; editing for clarity; summarizing internal docs (names stripped); auditing redundant software. |
| Yellow (Ask First) | Requires review by a department lead or the AI Governance Group before launch. | Any public-facing website assistant; new AI software subscriptions; connecting AI to ChMS, CRM, or email systems; public AI-generated images or video. |
| Red (Off-Limits) | Strictly prohibited across all staff and volunteer roles. | Pasting confidential pastoral, HR, counseling, donor, or minors' data into unapproved tools; pretending AI output is personal care; autonomous crisis chatbots; unverified Scripture or factual claims. |
Leadership Adoption Block (Fill in for Your Team)
Human Ownership Rule: Every piece of published work carries a human name. Staff remain 100% responsible for accuracy, tone, Scripture/source verification, and integrity of anything shipped.