From CEO to Tech Support
Last week I was on a call with 20 business owners in my AI for Business Owners course, walking them through setting up Claude Code in a GitHub Codespace. It was not going well. The interface is unfamiliar to non-technical people. Every screen looks wrong. Everyone had clicked one button incorrectly at some point. Some had stopped trying.
One participant said what everyone was thinking: "You're tech support."
He was right. I’d spent five years build a leadership team at my first company so I could step down from day to day management and here I was walking CEOs through GitHub OAuth screens.
I went all-in on AI because I believed this technology would change how businesses operate. But the AI interface problem had knocked me down to tier 3 support.
Then on February 10th, Anthropic released Cowork for Windows. And the wall came down.
What Actually Changed
Cowork does something that sounds trivially simple: it lets Claude work on files on your local computer. You point it at a folder. Claude reads, edits, and creates files inside it. You don’t need GitHub or a Codespace and never need to open a terminal and see a command line.
Here's the backstory. Anthropic's developers already write nearly 100% of their code using Claude Code. Non-technical employees at Anthropic saw what the developers were doing and started using Claude Code themselves to file expense reports, organize files, build marketing materials, etc. Anthropic realized in December 2025 that the tool people actually needed wasn't a chat interface. It was an AI that could work on your actual files. Four engineers built Cowork in ten days. All the code was written by Claude Code itself.
That product launch wiped $300 billion off software stocks in a week. LegalZoom dropped 20% in a day. Atlassian lost 35%.
Why?
Because Anthropic released plugins for basic legal review, marketing plans, and financial analysis. If Claude can review a contract on your laptop, you don't need LegalZoom. If it can manage a project from your files, you need fewer software seats. Wall Street called it the "SaaSpocalypse."
What This Actually Makes Possible
Cowork makes any documented process executable. That sentence is worth reading twice.
Anthropic introduced something called Skills last October. These are basically executable standard operating procedures. A Skill is a file that tells Claude how to do something: how to process a customer service email, how to generate a weekly report, how to onboard a new client. Load a Skill, point Cowork at your files, and it runs the process. Write the Skill once, execute it forever.
At OwnerRx, we're building this right now for our cohort business. We have three phases: marketing the cohort, conducting it, and closing it out. We're writing the SOP for each phase as a Skill. A new employee with Cowork and those Skill files can launch and run each phase. Start the marketing phase and it immediately sends action items to each person, tracks progress in a Google Sheet or PM tool, and follows up. The knowledge that used to live in one person's head becomes a system anyone can run.
We use an automation tool called n8n to connect Cowork to external systems (email, CRM, calendar) because the built-in connectors are still mostly read-only for security reasons. n8n gives us the write layer. But we're deliberate about it: Cowork drafts customer service emails and saves them in the drafts folder. A human reviews and clicks send. My guess is that within a year, you won’t need n8n except for very specific edge cases.
This is a fully executable knowledge and process system. Any company can build one. We're teaching owners how to do it. In fact, I’m currently refactoring our cohort offering around this tool.
The Risk Is Real but Manageable
One issue that will come up is security. Prompt injection is the main concern. This is a set of malicious instructions embedded in a website or document that tell Claude to destroy data or send it somewhere it shouldn't. Every system you connect expands the surface area.
But the risk is roughly the same as uploading files to a chat interface, which we're all already doing. Be deliberate about what folders you share just as you are deliberate about which files you upload to the chatbot. Don't give it permission to send emails directly. Keep write access narrow.
Anthropic is almost hysterical about security, and that's fine. I'd rather they overcorrect. But we can't let edge cases stop us from moving forward. Edge cases exist in everything we do every day.
What to Watch
The Skills market. Anthropic released 11 plugins in January. If the community builds hundreds by mid-year, this becomes the new app store for business processes.
Your software subscriptions. Before you renew anything under $500/month, ask whether Claude plus a Skill could replace it. The build-vs-buy math has permanently changed.
Your own timeline. Gartner says 40% of enterprise apps will have AI agents by end of 2026. That's ten months. If you think you have five years to figure this out, you don't.
I went from CEO to tech support. Cowork means I don't have to be tech support anymore. Now I can focus on what actually matters: helping owners build AI into how they work, not fight with how AI works.
If you’d like to sign up for our next Cohort focused on building this type of system, register here.

