Make Money with Claude Cowork: The 2026 Full Course, Decoded

Everyone keeps telling you to learn AI while almost no one shows you how to turn it into money, step by step. Oliver tries to close that gap with an end-to-end build: the right settings, durable context and real projects. The promise is that once you finish you will sit ahead of eighty percent of AI watchers because you actually use the tools. Companies are cutting roles because of AI and at the same time hiring people who can build with it; very few are on the right side of that equation.

The first mistake is writing a fresh prompt from zero every time. His example is simple: open a claw-like interface and ask for a website for your business — the model does not know you. You must re-explain your customers, your brand and your offers, and the output stays generic enough to fit anyone. It is not the model’s fault, it is the context you gave it. The fix is not another single prompt but context given once that compounds with every request.

Why One Prompt Fails: Context Is Everything

The hub is Claude Cowork. After installing the desktop app you see three tabs: chat, cowork and code. Chat answers like the website, cowork edits your folders and runs projects with you, code does the same with a software focus. Oliver keeps all work inside cowork: give the AI a job, it goes off, does it and reports back. He frames the Pro plan as affordable as skipping one coffee and a sandwich a month.

A one-person business starter pack forms the foundation. You follow the short guide, select the full pack, download the zip, then in the app go to customize > skills > add > upload skill and bring the folder in. It adds seven personal skills — essentially prompts you would otherwise paste into Claude or ChatGPT, now packaged as runnable commands. Inside cowork you run the AIOS setup command.

Setup first asks where your business context lives: Notion, Drive or local files, or start from scratch with name, goals and offers. Oliver points it to a three-year Drive archive. Via customize > connectors > browse you add the Drive connector, authorize, and tell the chat you have more in Drive and to check a PDF and gather context. The progress bar shows connectors and files in use; contradictions like two pricing models and gaps like a missing ninety-day focus are flagged, and communication style is set to blunt and direct before safety rules are locked.

Safety has two layers. If you are new, ask-before-acting teaches the rhythm; act-without-asking is faster once you are comfortable. Oliver draws a clear line: publishing, deleting files or doing real work must always ask first. That lets the assistant run autonomously for routine steps while keeping sensitive actions gated. Generation then starts: claw.md as the handbook read before any reasoning, a linked business profile, brand voice, working preferences, a context map from Drive and Notion plus OneDrive, a memory log and a to-do list.

You consolidate the output by creating a new folder called AIOS Cla, opening it and asking the assistant to organize everything inside. That makes the workspace explicit and lets you grant file-change permission only inside that folder instead of the whole computer. From then on you use cowork’s work in a project or folder, select AIOS Claude and allow edits there. The daily loop uses good morning to summarize yesterday and priorities, end of day to move unfinished items to the list, to-do to review and edit, and AIOS help to list every command.

AIOS: A Second Brain That Compounds on Your Machine

Files can get messy, so Oliver adds Obsidian as a nicer second brain. You install Obsidian and instead of creating a new vault you open the AIOS folder as a vault, so everything you built with Claude becomes a connected knowledge graph. Small at first, it grows each day as you work and add context, and Claude gets smarter with it. For the to-do list you go to settings > community plugins > browse, find Kanban, install and enable; the list turns into draggable columns you can move to done, and Claude can add or complete items.

He underlines why this matches costly consulting: a consultant also gathers context before advising. Now Claude knows your businesses, offers, audience and voice and can help with pricing, offers, marketing and acquisition decisions. As a service Oliver prices the AIOS setup at 500 to 750 dollars upfront with no monthly retainer unless ongoing help is requested; it is the foot-in-the-door for a package that goes up to 7,000 dollars for the full Cowork system, and because it is plain text it can move between Cowork, Code or other models.

Next the assistant gets hands and a schedule. Right now it answers but does not act for you; it has no Gmail, calendar or team tools. The fix is connectors, which is the MCP idea — the standard that lets Claude use your tools. You go to customize > connectors > add > browse, open the popular list and add Google Calendar, authorize, see the check mark and repeat for ClickUp, Fathom for meeting notes, Gmail, Google Drive, Interactive Brokers for trading and QuickBooks for invoices. Zapier can add eight thousand more connections when a client uses something not listed directly.

With hands comes scheduling. The assistant can now run checks without being told and brief you every day. In Oliver’s setup it handles inbox triage, pulls a briefing from calendar, ClickUp and Fathom, and delivers a stock portfolio update before you wake up. The brain is Claude, the context is the AIOS files, the hands are the connectors, and the schedule ties it together. Each connector is a few clicks; the key is granting the right scope and verifying the status.

A Site in Five Minutes and an Automated Form Funnel

The next big build is a 10,000-dollar website in five minutes. You can aim it at your own business, a demo client or a real client; the more detail you give, the more tailored the result. In the video a Summit Roofing example is used: a roofing clip and business context are dropped into the AIOS folder and Claude generates the site, shipped on a free hosting tier. It loads fast, looks clean and is ready to share; Oliver says he has used the flow since May without burning credits.

The funnel matters more than the page. In the site dashboard you enable form detection, then under project configuration > notifications > form submission notifications > email notification you route every new submission to your inbox. You hand Claude the hosting docs so it fixes the reply-to address correctly, redeploy, and test with a submission: name, phone, testle1@gmail.com, Austin Texas address, full roof replacement. The business inbox receives the inspection request with name, phone, email and message, and the skill is updated to remember the correct form setup.

Those leads are then handled automatically with Gmail wired up. After confirming the Gmail connector is checked, you prompt to check the inbox for new Summit Roofing leads and draft replies in the brand voice — friendly, local, thank them, offer a free inspection and propose two time slots this week without sending yet. Claude pulls the submission with name, phone and storm message and drafts the reply; you give feedback and tell it to insert the phone number and send, or hand off to someone who clicks send every few minutes. Oliver prices this website plus funnel at 500 to 2,000 dollars upfront and 200 to 1,000 dollars monthly if you keep improving Claude’s replies.

The pattern has a name: Anthropic released 31 small-business skills as one plugin. Oliver argues 26 are fine for productivity but only five actually move money. He filters with three questions: does it touch a paying customer, does it return value on day one, and does it save real time on a recurring task. Most accounting requests are a 5,000-dollar-a-year problem, while these five attack 50,000 to 100,000-dollar-a-month problems. Since 2024 that filter helped his agency generate over 900,000 dollars implementing agents.

Five Real Revenue Skills That Matter for Small Business

Installing the plugin goes via customize > personal plugins > plus > browse plugins > search small business. The first skill is lead triage: it scores hot inbound leads and drafts personalized emails. With HubSpot connected the demo shows 74 leads in 72 hours — five hot, eighteen warm, rest cold. Drafts reference specifics: Mike’s ranch-house quote and plan to move before winter with a call to book a measurement, Sarah’s emergency form and master-bedroom photos with a direct call prompt. The five hot drafts are approved and sent, and the remaining inspection-only leads enter a five-email nurture sequence, with warm leads queued the same way.

The second skill, call list, pairs with triage: instead of emailing, call the hottest leads now. Using the same HubSpot data it outputs five top calls with name, property, last communication, best time to call, a two-sentence opener referencing what was said or requested, and the first sales question to advance toward booking. It branches for yes, maybe and no answers. Running daily or every other day, thirty minutes of warm outreach beats three hours of cold. Owners or salespeople can work the list without digging through the CRM.

The third skill, invoice chase, recovers cash. With Gmail plus QuickBooks or PayPal connected it pulls overdue invoices — in the demo eight invoices totaling 9,600 dollars — and shows customer, amount, days late and platform with a payer score that sets the tone: direct for very late, softer for a short slip. Marcus Diaz with a 14-day 4,200-dollar invoice gets a firm note that the five-year warranty stays inactive until paid, with a PayPal link and a phone offer; Jennifer with a 2-day 850-dollar gutter cleaning and good history gets a friendly nudge. Red, yellow and green signals mark severity and letter versions flip if norms look off, and Gmail drafting opens inside HubSpot records for two to three late invoices at a time.

The other two skills handle complaints and database reactivation. Complaint handling pulls conversations across PayPal, Gmail and HubSpot and drafts a response in your voice that seeks a fair compromise instead of reflex refunds, protecting relationships. Lead reactivation scans older form submissions and stalled contacts and restarts engagement sequences, often with acknowledgment and a clear next step. Together with triage and call list, the set covers inbound capture, immediate follow-up, overdue cash and re-engagement.

From Cold Email to Close: Machine plus Sales

Dashboards make it visible and billable. Teaching live dashboards that clients pay for monthly turns one-off setup into recurring revenue: each business sees its own pipeline, form volume and follow-up status in one view. The video walks through wiring the dashboard to the same data sources used for triage and invoices, so owners see where money leaks and where the assistant already plugged the gap.

Module six is the cold-email machine. Oliver has run it since August 2024 for his own agency to fill calendars, claims well over 300,000 dollars from it and scaling his AI company into 50,000 and 80,000-dollar months. The stack is simple: Claude as the brain, Gmail as infrastructure, Instantly as the hands for sending and replies, and Clay for lists. Rule one is never use the main domain; buy a close variant, create three to five inboxes per domain and let Instantly handle DNS and warm-up, with 20 to 30 days of warming or faster with pre-warmed inboxes. Deliverability is checked with an MX lookup tool; only major errors matter, warnings like SOA serial can be ignored.

Setup hands Claude the keys. You create an Instantly API key under profile > settings > integrations > API keys, name it Claude OS, grant all scopes and copy it, then in Cowork go to customize > connectors > add custom connector, name it Instantly and paste the connector string with the key and connect. The first skill then applies daily sending, warm-up and account settings in one go. For lists Clay is used: filtering by job attributes, pasting 80 decision-maker titles into the job title filter to get 73 targets in a local area, enriching work emails and wiring Clay to Instantly for verification and upload automation so the machine scales.

Module seven is the demo that wins the meeting. The cold-email machine books calls, but closing takes a different skill. The demo is shaped to the biggest money leak you find: missed calls get a voice agent, slow message replies get WhatsApp or social inbox agents, large databases get reactivation. With tech and list correct, the focus shifts to copy and offer; messages must feel personal and not automated, and the demo must bridge from interest to a booked inspection or call.

Module eight is the audit and close. Oliver calls it the doctor approach: diagnose before prescribing, ask enough questions to restate the prospect’s pain better than they did, then the close becomes a byproduct. The frame starts with two questions — where do you get clients and how do you turn them into money — hunting the biggest leak first. A simple consequence calculator is used: at 30 to 40 dollars per lead and a 12 percent pickup rate the cost per conversation sits near 300 dollars, so a booked appointment at 60 to 70 dollars anchors the value. Small businesses can be closed with a card today; enterprises start with a small paid pilot, sometimes at 50 percent off for the trial volume, no upfront heavy integration, manual mode first to launch in days, money-back within seven days if they dislike it, and never leaving without the next meeting on the calendar.

Payments run through Whop with a custom connector that creates a payment link on the call. A one-dollar commitment link gets the card on file for performance-based weekly billing; for larger setup fees buy-now-pay-later lets the owner pay over time while you collect upfront, reducing chase. The finale replays a live sales call with a major Australian solar company at 1 a.m. Copenhagen time: over 500,000 open leads, a 20-person call center making 180 calls a day, a 12.23 percent one-minute talk rate, aged data concerns, and on-year-old purchased leads a 27 percent response and 1.72 percent booking rate — 248 appointments from 14,000 contacts — followed by a low-volume manual pilot of one to two thousand leads before scaling. The ladder is one audit, one service and one upsell to a 7,000-dollar full Cowork implementation, with monthly services as the cherry on top.