We are not replacing you. We are multiplying you. This workshop gives you hands-on experience with Claude Cowork so you can take repetitive knowledge work off your plate and focus on the work that actually requires you -- leading people, developing leaders, and advancing the mission.
Learn the concept, see a demo, try it yourself. Each block builds on the last.
The mission frame. What is happening in AI right now, why it matters for church leadership, and what our Easter runway looks like.
From chatbot to intern to team member. How Cowork differs from Chat, what "agentic AI" means, key vocabulary.
Open the app, select a folder, run your first real task. Troubleshoot together.
Role-specific breakout. Each person identifies 3 tasks from their actual week. Try one live. Discover relevant plugins.
Group discussion: what do we never outsource? Safety essentials. Pilot milestones. Launch #ai-wins. Point to this reference site.
Questions, troubleshooting, and "what should I try first this week?"
AI is not coming. It is here. The question is not whether your role will be affected, but whether you will be the one steering how it helps your team and the people you serve.
For the past two years, most people have experienced AI as a chatbot -- you ask a question, you get an answer. That was Stage 1. We are now entering Stage 2: agentic AI. Instead of just answering questions, AI can now plan multi-step workflows, use tools, read and write your files, browse the web, connect to your apps, and execute real work toward a goal you define.
McKinsey reports that 72% of organizations are now using AI in at least one business function. Gartner expects 40% of enterprise applications to include task-specific AI agents in 2026, up from under 5% in 2025. BCG's research shows the impact formula is 10% algorithm, 20% infrastructure, 70% people and process. The technology is ready. The differentiator is whether the people using it know how to think with it.
Every one of you was hired for your domain expertise -- pastoral care, operations, finance, production, nonprofit leadership. Not for your ability to format spreadsheets, chase down data across three systems, or reformat the same report every month. Those tasks are real, they take real time, and they keep you from the work that only you can do.
The goal is not to become an AI expert. The goal is to reclaim capacity -- so you can spend more time developing leaders, serving people, and thinking strategically about the mission ahead.
Easter is April 5 -- six weeks from now. That is our forcing function. The goal is that by Easter, this pilot group has enough experience with Cowork to have real wins to share, at least one cross-functional workflow automated, and the confidence to demo what you have learned to the broader team. Early adopters build capability, then share wins. That pulls in the next wave.
Before you touch the tool, you need a mental model. This block gives you the vocabulary and framework so everything else clicks.
Claude Cowork is a tab inside the Claude Desktop app that turns Claude from a chatbot into a digital coworker: you describe what you want done in plain English, grant it access to your files, and it executes the work -- documents, spreadsheets, organized folders, research summaries -- without you writing a single line of code.
Think of your AI journey as a progression:
You ask questions, you get answers. Most people are here. This is Claude Chat, ChatGPT, Gemini.
You give it tasks, it executes with enthusiasm. Needs supervision. This is where Cowork enters. You must get clear on intent, output, guardrails, and process.
You have built context files, prompt libraries, and workflows. It knows your org's language and standards. It consistently produces good work.
Today we start at Stage 2. The goal over the next six weeks is to mature toward Stage 3. The leadership development angle is real: as you spend less time on repetitive operational work, you gain bandwidth for the things only humans can do -- pastoral care, leadership development, vision casting, and being present with people.
| Feature | Claude Chat | Claude Code | Claude Cowork |
|---|---|---|---|
| Interface | Web / mobile | Terminal (command line) | Desktop app with visual panels |
| How it works | You ask, it answers | You describe code tasks | You describe any task, it plans and executes |
| Best for | Quick questions | Software development | Documents, files, reports, automations |
| File access | Upload only | Full local system | Sandboxed folder you choose |
| Runs on | Anthropic servers | Your machine directly | Isolated virtual machine on your computer |
| Church example | "Explain Romans 12:2 in Greek" | Build a church app | "Turn this CSV into a formatted giving report" |
| Term | What it means |
|---|---|
| Agentic AI | AI that can plan, execute steps, use tools, and work toward a goal autonomously -- not just answer questions. |
| Sandbox | An isolated environment on your computer. Cowork runs inside this so it cannot touch your system files. |
| MCP Connector | A bridge that lets Cowork talk to external apps (Slack, Google Drive, Planning Center, etc.). |
| Plugin | A bundle of skills and connectors that specializes Cowork for a domain (finance, productivity, etc.). |
| Skill | Best-practice instructions that tell Cowork how to produce a specific output type (a spreadsheet, a report). |
| Session | One continuous interaction. Sessions do not carry memory to future sessions. |
| Steering | Typing corrections mid-task to redirect Cowork without starting over. |
| CLAUDE.md | A context file in your working folder. Cowork reads it at the start of every session to understand your organization. |
Everyone opens the app together. We walk through setup, run a test task, and troubleshoot as a group.
| Requirement | macOS | Windows |
|---|---|---|
| OS | macOS 11+ (Big Sur or later) | Windows 10 or 11 (x64) |
| Processor | Apple Silicon (M1+) required for Cowork | x64 processor |
| RAM | 8 GB min, 16 GB recommended | 8 GB min, 16 GB recommended |
| Disk | 10 GB free | 10 GB free |
| Subscription | Claude Pro, Max, Team, or Enterprise (paid plan required) | |
Type this and hit enter:
If the file appears in your folder, you are ready. If not, check the troubleshooting table below.
| Problem | Cause | Fix |
|---|---|---|
| Cowork tab missing | Outdated app or Intel Mac | Update Claude Desktop. Check your chip. |
| Setup hangs 15+ min | Network or firewall | Check internet. Disable VPN. Restart app. |
| Cannot see files | No folder selected | Click the folder icon and choose your directory. |
Different roles have different pain points. Find your category, then try one of the suggested tasks live. The goal: identify 3 tasks from your actual week that Cowork could handle.
You hold the whole picture. Cowork can compile department updates into a single briefing, generate attendance and giving dashboards, draft meeting agendas with pre-populated data, and bulk-organize shared drive folders.
You build the systems everyone else runs on. Cowork can create SOPs from rough outlines, summarize and compare vendor contracts, build evaluation spreadsheets, and automate monthly reporting.
Your week moves fast: care follow-ups, volunteer coordination, event planning, and campus logistics. Cowork handles the admin so you can focus on people.
If you oversee one or more nonprofit entities, Cowork is purpose-built for multi-entity work: grant applications, impact reports, donor communications, and compliance documentation across organizations.
You manage the creative and technical production across services and events. Cowork can generate run sheets, organize media libraries, draft social content, and maintain equipment inventories.
You keep the financial engine running. Cowork has a dedicated Finance plugin with skills for reconciliation, journal entries, variance analysis, and statement generation.
Understanding what Cowork should and should not do. Plus: what we commit to as a pilot group.
AI is a tool, not a replacement for the things that require human presence, spiritual discernment, and genuine relationship. Before we go further, let's discuss as a group: What are the things in our roles that should always remain human?
Some categories to consider: pastoral care and counseling, prayer and spiritual discernment, personal relationships and discipleship, personnel decisions, crisis response that requires human presence. This is not a list someone else defines for you. It is a conversation your team has together so everyone shares the same convictions about where the boundaries are.
| Never do this | Why |
|---|---|
| Delete original files without backups | Cowork asks permission first, but always verify before approving |
| Send emails or messages without reviewing | One wrong message can damage relationships |
| Enter passwords, API keys, or banking credentials | Cowork will refuse some of these, but it is your job to never share them |
| Process unprotected donor PII or financial account data | Session data is stored locally with no centralized audit trail |
| Grant access to your entire home directory | Scope the folder to what the task actually needs |
| Target | Goal |
|---|---|
| This week | Try 2-3 real tasks from your actual work. Low stakes first. |
| By mid-March | 4 of 6 pilot members using Cowork weekly. 3+ documented wins in #ai-wins. |
| By Easter (Apr 5) | At least one cross-functional workflow automated. Team confident enough to demo to broader staff. |
| End of Q2 | Evaluate ROI. Decision on expanding to 15-25 seats. Shared CLAUDE.md and Planning Center skill built if value is proven. |
Everything below is your ongoing reference. Come back to this site whenever you need to look something up, share it with a new team member, or explore a new capability.
Common questions people have when they start using AI at work.
All three are strong, and the honest answer is "it depends on the task." ChatGPT (by OpenAI) has the largest user base and is the most versatile generalist. It is great for quick creative tasks and has a massive plugin ecosystem. Gemini (by Google) is strongest if you live in Google Workspace -- it integrates natively with Gmail, Docs, and Sheets. Claude (by Anthropic) leads in writing quality, coding accuracy, and long-document analysis. It is also designed with stronger safety guardrails from the ground up.
For the kind of deep, thoughtful work church leadership involves -- policy documents, strategic planning, sensitive communications, financial analysis -- Claude is a strong fit. But there is no reason to limit yourself to one. Many people use multiple tools for different purposes. The important thing is to start using AI well, not to pick the "right" one.
Claude Chat is a conversation. You type a question or paste some text, and Claude responds. It lives in a web browser or mobile app. It cannot see your files, cannot connect to your apps, and cannot take actions on your behalf. Think of it as texting a smart friend for advice.
Claude Cowork is a workspace. It runs inside the Claude Desktop app on your computer. You grant it access to a folder on your machine, and it can read your files, write new documents, create spreadsheets and presentations, browse the web through Chrome, and connect to apps like Slack, Google Drive, and Planning Center. Think of it as handing a capable team member a project brief and saying "go."
The rule of thumb: if your task is a quick question, use Chat. If your task involves files, multiple steps, or app integrations, use Cowork.
OpenClaw (originally called Manus, rebranded late 2025) is an open-source AI agent that launched in November 2025. It has grown rapidly -- 175,000+ GitHub stars, 300-400K users -- and runs locally on your computer, connecting through messaging apps like WhatsApp, Telegram, and Signal. It can browse the web, manage files, and execute tasks similarly to Cowork.
However, it comes with real concerns for organizational use. Kaspersky identified 512 vulnerabilities (8 critical) in its codebase. Cisco's security team found third-party extensions performing data exfiltration. Because it is open-source and community-driven, there is no single company standing behind its security or compliance.
For personal experimentation, it is interesting. For an organization handling donor data, financial records, and people's personal information, a managed and sandboxed solution like Cowork is a safer bet. Cowork runs in an isolated virtual machine that cannot touch your system files, and Anthropic stands behind its security model.
An AI agent is an AI system that can plan, use tools, make decisions, and work toward a goal with limited supervision. Unlike a chatbot that responds to a single prompt, an agent can break a complex request into steps, decide which tools to use, execute actions (creating files, browsing websites, querying databases), and adapt based on what it finds.
The simplest way to think about it: a chatbot is a calculator (input in, output out). An agent is more like a team member (give it a goal, it figures out the steps). Claude Cowork is Anthropic's consumer-facing AI agent. It follows a loop: perceive (read your files and context), plan (decide what steps to take), act (create documents, browse, connect to apps), and learn (adjust based on your feedback mid-task).
Gartner expects 40% of enterprise applications to include AI agents by the end of 2026. This is the direction the entire industry is moving. Getting comfortable with agentic AI now puts you ahead of the curve.
The research consistently points in one direction: AI replaces tasks, not jobs. The people most at risk are those who refuse to learn how AI can help them. The people who thrive are the ones who use AI to handle the repetitive parts of their work so they can focus on the parts that require human judgment, creativity, and relationships.
In a church context, this is especially clear. No AI can do pastoral care. No AI can discern what a community needs spiritually. No AI can build the kind of trust that comes from showing up for someone in a hard season. What AI can do is free you from the spreadsheet, the report, the follow-up email template, and the file organization -- so you have more time for the work that actually requires you.
Moving from "AI as search engine" to "AI as capable team member" requires a new way of thinking about your work.
When you give a task to a human team member, you naturally communicate four things: what you want (intent), what the deliverable should look like (output), what boundaries to stay within (guardrails), and how you want them to approach it (process). Working with Cowork is the same. The clearer you are on these four dimensions, the better the result.
"I need to understand our giving trends over the last quarter so I can brief the board."
"Create an Excel spreadsheet with a chart and a one-page summary document."
"Use only the data in the attached CSV. Do not pull from external sources. Round to whole dollars."
"Start with the raw data, clean any duplicates, group by campus and fund, then create the chart."
"I am not technical enough." Cowork is designed for non-technical users. You describe tasks in plain English. If you can write an email to a colleague explaining what you need done, you can use Cowork.
"What if it makes mistakes?" It will. Just like a new team member, it will sometimes misunderstand your instructions or produce something that needs correction. That is why you review before sending, start with low-stakes tasks, and use the steering feature to course-correct mid-task.
"I do not have time to learn something new." The investment is about 2-3 hours over the next two weeks to build basic comfort. After that, it starts saving you time. The question is not whether you can afford to learn -- it is whether you can afford not to, as this becomes the standard way knowledge work gets done.
Beyond Cowork, here are practical tools and techniques that can accelerate your AI fluency.
WisprFlow is a voice transcription tool that lets you speak instead of type -- anywhere on your computer. Since most people speak 3-4x faster than they type, this removes a major friction point. It is especially useful for Cowork: instead of carefully typing a long task description, you can speak naturally and WisprFlow transcribes it into the input field. For church leaders who spend a lot of time in conversations rather than at a keyboard, this is a game changer.
Here is one of the most underrated ways to get better results from any AI tool. Instead of trying to write the perfect prompt from scratch, start with this:
For example: "Create a prompt I can easily copy and paste to AI, followed by: I want to analyze our Q1 giving data by campus and create a board-ready summary with charts." The AI will generate a well-structured, detailed prompt that you can then copy and paste into Cowork (or any other AI tool). You are using AI to write better instructions for AI. Over time, you internalize what makes a good prompt, and your direct instructions get better naturally.
| Tool | What it does | Good for |
|---|---|---|
| Perplexity | AI-powered research with real-time web sources and citations | Quick research, fact-checking, current events |
| NotebookLM | Google's tool for uploading documents and having AI-powered conversations about them | Digesting long documents, study prep, meeting prep |
| Claude Chat | Quick Q&A, brainstorming, writing assistance | Fast answers, drafting emails, explaining concepts |
| ChatGPT | Versatile generalist with image generation and web browsing | Creative tasks, image creation, quick research |
Anthropic ships 11 official open-source plugins, free with any paid plan. These are the same tools Anthropic's own teams use. Install from the Plugins sidebar in Cowork.
Reconciliation, journal entries, variance analysis, financial statements, close management, audit support.
Task management, memory system (remembers preferences across sessions via files), daily briefings.
Data exploration, visualization, interactive dashboards, statistical analysis, SQL queries.
Search across Slack, Google Drive, Notion, and other connected sources in a single query.
Content drafting (blog, social, email), campaign planning, brand voice review, SEO audits.
Contract review, NDA triage, compliance checks (GDPR/CCPA), risk assessment.
Ticket triage, response drafting, escalation packaging, knowledge base articles.
Feature specs, roadmaps, stakeholder comms, competitive analysis, user research synthesis.
Account research, outreach drafting, call prep, pipeline review, competitive intelligence.
Connecting Cowork to your church management system.
With the Claude in Chrome extension, Cowork can browse Planning Center just like you do -- view people lists, check service schedules, pull giving data from the browser interface. No technical setup beyond installing the Chrome extension.
For deeper integration, MCP servers can connect directly to Planning Center's API to query people data, pull lists, and filter by roles programmatically. Requires a Personal Access Token and configuration. Worth exploring after the pilot proves value.
Your way of teaching Cowork about your organization without repeating yourself every session.
A file named CLAUDE.md placed in your working folder gets read by Cowork at the start of every session. Think of it as a briefing document for your AI team member. It can include your organization's name and mission, key terminology, campus names, ministry and department structure, fiscal year dates, preferred formats, and nonprofit entity details.
Who should create this? Ideally, your General Manager, COO, or Executive Director creates a master version. Everyone on the team places a copy in their working folder. This way, every staff member's Cowork sessions start with the same organizational understanding.
Courses, communities, and references to continue building your AI fluency.
| Resource | What it is |
|---|---|
| AI For Church Leaders | Membership community with workshops built specifically for pastors and church staff navigating AI adoption. |
| AI NEXT by Exponential | Practice area guiding ministry teams through AI adoption with policy guidance and use case libraries. |
| AI for Ministry Masterclass | Beginner-to-advanced curriculum from Discipleship Ministries. Free intro tier available. |
| Resource | What it is |
|---|---|
| OpenAI Academy | Free courses on prompt engineering and AI literacy. Rolling out certifications for different fluency levels. |
| Google "Make AI Work for You" | Free in-person and online workshops touring the country. Practical, hands-on format. |
| Brown University GenAI Workshop | 90-minute professional workshop. Strategies for productivity and innovation regardless of role. |
| Coursera / DeepLearning.AI | Structured certificates. Andrew Ng's courses on foundation models, RAG, and prompt engineering. |
| Resource | Link |
|---|---|
| Official Help Center | support.claude.com |
| Plugin Marketplace | claude.com/plugins |
| Open-source Plugins (GitHub) | github.com/anthropics/knowledge-work-plugins |
| Claude Desktop Download | claude.com/download |
| Planning Center API Docs | developer.planning.center/docs |
| PCO MCP Server (community) | PCO-MCP on Glama |
Phase 1 (the current pilot): $150/month for 5 Cowork seats on the Claude Team plan. This gives each pilot member full access to Cowork, all plugins, and the Claude in Chrome extension. If the pilot proves value by end of Q2, Phase 2 expands to 15-25 seats across the broader staff.
This is not about becoming an AI expert. It is about raising the awareness level of the team with the latest capabilities, so you can reclaim capacity for the work that matters most -- developing leaders, serving people, and accomplishing a much bigger mission together.
The biggest mistake you can make is not trying it on real work this week. Start small. Iterate. Share what you learn in #ai-wins. The compounding loop starts now.
Built for VIVE Church staff, February 2026.
Course content informed by Anthropic's official documentation, open-source plugin repository, and Planning Center API docs.