🚀 Live Workshop · $600 / seat · Small Business Track

Put AI to Work in Your Business

A hands-on workshop on Claude Code, Cowork, AI agents and automation — designed for small-business owners and operators, not engineers.

7Practical modules

5–10 hrsSaved per week

27+Curated resources

0Lines of code required

⬡ The big idea

The Claude product ladder

Every tool in this workshop sits on one ladder. Pick the rung that matches the job — most small businesses live on the first two.

	Claude Chat	Claude Cowork	Claude Code
Think of it as	An assistant who answers questions	An employee who completes tasks	A power builder — most customizable
Best for	Quick questions, brainstorming	Multi-step work, files, business productivity	Building custom tools, apps, automations
Who	Everyone	Owners & operators (no terminal)	Builders & technical teams

💡

The promise: “If you can text a friend, you can use Claude Cowork.” By the end of this workshop you will have a repeatable system that takes 5–10 hours of busywork off your plate every week.

Outcome 1

Automate the busywork

Turn file chaos, data entry and repetitive admin into one-line instructions.

Outcome 2

Build a tireless team

Stand up agents that run while you sleep and scale without new hires.

Outcome 3

Sell it as a service

Use the AI Audit Framework to deliver paid AI engagements to clients.

Mark intro complete & continue

Foundations

What is Claude Code?

The agentic engine behind everything else in this workshop — and every surface it runs on.

An agentic AI tool that reads your files, edits them, runs commands and connects to the tools you already use. The same engine runs everywhere, so your settings, instructions and connections travel with you.

Fastest, most scriptable surface. Install once, then run inside any project folder.

curl -fsSL https://claude.ai/install.sh | bash
cd my-business-folder && claude

Install the VS Code or JetBrains extension for inline diffs and edits without leaving your editor.

Standalone app — visual diffs, parallel sessions, scheduled tasks and cloud sessions. The friendliest entry point for non-technical users.

claude.ai/code — zero local setup, ideal for long-running and parallel tasks from any device.

What you can actually do with it

Automate the work you avoid

Reports, reconciliation, renaming, follow-ups, release notes.

Build from plain language

Describe an outcome; it builds the tool or fixes the problem.

Connect your stack via MCP

Drive, Gmail, Slack, Jira, CRM — read & write across tools.

Customize & schedule

CLAUDE.md, skills, hooks; recurring tasks via Routines, Desktop or /loop.

📌

Workshop angle: This is the “what is this thing” source of truth. Module 2 (Cowork) is the non-technical front door to the very same engine.

Mark Module 1 complete

Anchor module · Non-technical

Claude Cowork — 5× productivity, no code

Claude Code's agentic power in an interface you can use if you can text a friend. This is the core of the SMB workshop.

The 6-step quickstart (do this live)

Download Claude Desktop at claude.com/download
Open the Cowork tab, point it at a messy folder, say “organize this”
Install one role-relevant plugin from claude.com/plugins
Build a skill: do a task once, then run /skill-creator
Set a scheduled task: /schedule → daily morning briefing
Share it with someone on your team who needs it

The three power features

Do a task once, then /skill-creator turns it into a repeatable /skill-name command. Fastest way to bottle a process so it runs the same way every time.

Bundles of skills + connectors for sales, finance, legal, HR, marketing. ~30 seconds to enable, free for paid subscribers. The strong SMB angle: instant capability where you can't hire a full-time specialist (e.g. Legal plugin's /review-contract).

Connect Fireflies / Notion / Gmail / Slack, then let multi-step workflows run unattended — e.g. the pre-built /call-summary skill pulls transcripts, extracts action items, builds a tracker and drafts follow-ups.

High-value demos (pick one per audience)

File chaos

2,200 desktop files sorted by category in minutes; screenshots bulk-renamed into logical folders.

Data → deliverable

Raw invoices → renamed, sorted by month, reconciliation spreadsheet with working formulas + tax breakdown.

1 → 100

One press release → Slack summary, Gmail draft, LinkedIn post, IG carousel, one-page FAQ PDF to Drive.

Batch at scale

14 job descriptions + Q1 strategy doc + 47 partner emails + website copy — in 2 hours.

Mark Module 2 complete

Scaling up

The agent-development ladder

When one assistant isn't enough — climb the ladder only as far as the job needs.

Rung 1

Subagents

Specialized helpers for focused jobs inside one session.

Rung 2

Agents in parallel

Several agents working at once in the background.

Rung 3

Agent teams

Coordinated roles handing work to each other.

Rung 4

Agent SDK

Programmatic agents embedded into your own product.

⚖️

When & why (the cost backbone): More agents = more capability and more tokens, latency and complexity. Building Effective AI Agents and Multi-Agent Research System teach the “only add an agent when a single one demonstrably can't do it” rule. Claude Squad is the open-source bring-your-own option.

Mark Module 3 complete

Recipes

Workflows & automation

Plain-language recipes that work on any surface, plus the official decision table for where a recurring task should run.

“Run on a schedule” — the decision table

Option	Where it runs	Best for
Routines	Anthropic-managed cloud	Runs even when your computer is off; API/GitHub triggers — strongest SMB fit
Desktop scheduled tasks	Your machine (desktop app)	Needs local files, tools or uncommitted changes
GitHub Actions	Your CI pipeline	Repo events / cron alongside workflow config
`/loop`	Current CLI session	Quick polling while a session is open

🎯

Official tip: Scheduled prompts run autonomously — be explicit about what success looks like and what to do with the result, e.g. “Review PRs labeled needs-review, comment on issues, post a summary to #eng-reviews.”

The module arc

Common Workflows (recipe hub) → Skills & Slash Commands (build your own /command) → Hooks (deterministic guardrails) → Headless claude -p (CI/pipeline) → GitHub Actions (automate the repo) → Routines (always-on cloud).

Mark Module 4 complete

Deployment

Keep it running while you sleep

Two axes decide every deployment, and the WAT framework names what you're actually shipping.

Axis 1 — Where does it run?

Your machine (must stay on) vs. the cloud (Anthropic, Modal, Trigger.dev, VPS).

Axis 2 — How deterministic?

Full agentic loop (autonomous) vs. fixed script (same every time).

Interactive: the WAT framework

What are you actually deploying? Toggle the parts that apply.

W — Workflow (the steps)

A — Agent (autonomous loop)

T — Tools (APIs / file access)

Select one or more components to see what you're shipping…

The three deployment methods

Fastest to set up — natural language (“loop every 10 min to…”). Session-scoped crons. Terminal beats Desktop: survives /clear, 7-day duration vs. 3-day. Quirk: up to 30 min random jitter (interval kept, exact times not). WAT: W + A + T, your machine, session open.

Anthropic-managed cloud. Runs even when your computer is off; also API and GitHub triggers. The strongest fit for small businesses that can't keep a machine on 24/7.

Self-managed cloud infrastructure for full control, custom dependencies and heavier scheduled jobs. The advanced rung — used when Routines isn't flexible enough.

🏢

The capstone: “Run an AI-First Business with Paperclip” — the pattern for running a whole AI org (agents with roles, a board, heartbeats and governance). The end state of this workshop's deployment track.

Mark Module 5 complete

Proof

SMB case studies & ROI

Real businesses, quantified outcomes — your evidence base when selling the workshop.

10h → 20m

Layer Agency

A reporting workflow that took 10 hours collapsed to 20 minutes with Supermetrics for Claude.

Cross-functional

Zapier

Claude workflows spanning sales, support and ops — a strong adoption signal.

AI-native

Headstart

A service firm that scaled throughput as an AI-native shop.

Founder patterns

YC startups

Repeatable founder workflows for building with Claude Code.

🔌

Connector spotlight: The HubSpot CRM connector for Claude wires marketing, sales and support workflows straight into the CRM your clients already run.

Mark Module 6 complete

Monetize

The AI consulting method

Turn everything above into a paid client engagement with the AI Audit Framework — the “Murder Mystery” discovery method.

The 7 audit worksheets

Engagement info — client, auditor, date, industry
Map all tools — every platform, use case, department, user count
Find knowledge sources — what feeds the RAG decision
Document processes — input → output flows, frequency
Identify owners — business, technical & budget owners
Measure results — baseline vs. target KPIs
Rollout plan — pilot → document → scale → optimize

The 4-phase rollout

1 · Pilot	2 · Document	3 · Scale	4 · Optimize
Select pilot group, train core users, run 30–60 days, collect feedback	Analyze results, document wins & issues, refine, build playbooks	Train next cohort, deploy to department, monitor	Continuous improvement against the KPI tracker

📏

Measurement tip from the playbook: Don't measure day-by-day — Fridays and Mondays skew the data. Review weekly or monthly.

Mark Module 7 complete

✓

Assess

Knowledge check

Five questions. Pass at 4/5 to unlock your completion certificate.

0/5

＃

Reference

Curriculum & resources

This site is built from Product Compass's curated 27-item AI Library.

📚 AI Library

27 catalogued items across 7 categories — the full curriculum source.

/paperclip/shared/AI Library/

📥 Intake drop zone

Drop new PDFs, notes or links to grow the library.

/paperclip/shared/Intake AI Library/

💼 Sales enablement

GTM collateral for selling the $600 workshop.

/paperclip/shared/Sales Enablement/

🧰 AI Audit Framework

The client-facing 7-worksheet deliverable (Module 7).

Catalogue #27

Attribution & lawful use: All third-party items are user-provided via lawful personal access. Workshop materials are Product Compass's own derivative work, crediting original authors and sources (Anthropic, Allie K. Miller, Nate Herk / AIS+, and others). For internal workshop use only.