Most of us still use AI like a browser tab we visit, type into, and copy from. Cowork changes that by putting Claude on your desktop where it can see your files, plug into your tools (Gmail, Calendar, Drive, your CRM), and run on a schedule. Combine that with skills, the written instructions that teach AI how you actually work, and you stop being the middleman moving information between systems. Your morning briefing writes itself, your commission log updates itself, your stale client list surfaces itself.
Sign in to view this.
Sign inSign in to view this.
Sign inIf you use AI today, the experience probably looks something like this. You open a browser tab. You type a prompt. You wait for the answer. You copy what comes back, paste it into your document, then jump back to the chat to ask another question. Maybe you upload a file, wait for a result, download it, paste it somewhere else. You're still doing all the work. The AI is something you consult occasionally, and you're the project manager moving information between systems.
What if you don't have to do the copy pasting any more and allow the AI to edit and menage your files itself. When the AI moves out of the browser tab and onto your actual computer, sees your files, plugs into your tools, and starts running tasks on its own, it has access to your work and can work the way you work.
The previous workshop was about skills. A skill is a written set of instructions that lives as a file. Your brand guidelines, your tone of voice, your process for booking confirmations, the edge cases you've learned the hard way. You write it once. The AI reads it every time. That's the move that takes AI output from "decent first draft, sometimes" to "consistent, branded, the way I would do it."
If skills are about reliability, this workshop is about putting those skills inside an environment where they can actually run on your work, not just on whatever you happen to paste into a chat window.
Cowork is Claude as a desktop app, not a browser tab. The useful difference is that it can see your files. Documents, spreadsheets, PDFs, templates, anything on your machine.
The way I describe it: using Claude in the browser is like texting someone instructions. Using Cowork is like having that person sit at your desk with your screen open. They can see what you're working on. They can open the files, read them, edit them, save the changes back.
Your skills are just files in a folder on your computer. Cowork discovers them and reads them automatically. You don't upload anything, you don't paste anything in. The instructions are right there in the folder, and the AI behaves accordingly.
It goes beyond your local files. Cowork can plug into the tools you already use. Gmail, Google Drive, your calendar, Slack, your CRM. These are one-click connections. You authorize it, and it's connected.
So now your assistant can check your inbox, look at your calendar, pull a record out of your CRM, and write something back into a Google Doc. All in the same conversation, without you switching tabs.
The piece that actually changes the relationship is scheduled tasks. You can tell Cowork to do something every morning at 7am, or every Monday, or twice a week. It picks up the skills you've written, the files it can see, the tools it's connected to, and just runs.
I run my own business through this system, and I want to share what that looks like, because the abstract version makes it sound smaller than it is.
Every interaction with every person in my community is logged automatically. Every call, every meeting. When I need to draft a message to someone, the AI already has the full history, what we talked about, what they care about, where we left things last time.
It creates the slides for my workshops. It manages my project notes and to-do list. Every morning, before I open my laptop, scheduled tasks have already run. Meetings pulled in from the calendar. New LinkedIn connections checked and triaged and Emails drafted. When I sit down at my desk there's a briefing waiting that tells me what happened overnight and what I should focus on today.
None of that is science fiction. It's all built with the same building blocks: skills, file access, connectors, schedules. I'm sharing it because I think it helps to see what is possible before you start designing your own version.
If you're an advisor trying to picture what this looks like for your business, here are three patterns that map directly onto the work.
A scheduled task runs every morning before you open your laptop. It checks your inbox, looks at your calendar, scans whatever else you've connected. And because you've written a skill, it knows what matters to you specifically. Maybe anything from a VIP client gets flagged. Maybe any email sitting unanswered for more than 48 hours. Maybe any trip departing in the next two weeks gets surfaced.
You define the rules once. Every morning the briefing is tailored to your priorities. Not a generic inbox summary, your briefing, for your business, based on your rules.
This one runs weekly. It looks at your client list, whether that's a spreadsheet, a CRM, or a folder of notes, and finds anyone you haven't been in touch with for a while. The skill defines your follow up philosophy. How long is too long for different kinds of clients. What a good touchpoint looks like. When to just check in versus when to actually propose something.
So instead of a list of names, you get specific suggestions. "You haven't spoken to Sarah in three weeks and she's got a Greece trip in May, maybe send a note about packing." "James's birthday is coming up soon and he mentioned he really like chocolates, should we send him a box from that place in Italy that he liked so much?." The relationship maintenance runs in the background. You review and act on what's worth acting on.
Every destination research you have done for one trip can be used to create social content like a newsletter, a blog or a post to show off your expertise. Most of us are drowning in work to the point that we don't have any time left to create the content that increases our book of business.
Instead, every Friday an AI scans the work that we have done the past week and any previous social posts to automatically draft our content for next week.
In the live workshop I demoed a commission tracker agent which is something that every advisor needs to do but no one likes to do.
The setup is simple. There's a spreadsheet on the desktop with a running log, twenty or so rows, color coded by status. Green for paid, yellow for pending, red for overdue. There's a skill that defines what a commission email looks like, what to extract (supplier, amount, status, expected date), and where to put it in the log. Gmail is connected.
The instruction in Cowork: "Check my inbox for any commission emails from the last week and add them to my commission log.", with a skill that explains how to identify commission related emails, and how to work with my spreadsheet.
It searches the inbox. Reads each email. Pulls out the supplier, the amount, the status. New rows land in the spreadsheet. Marriott paid $1,240, expected the 22nd. Four Seasons has $980 that's 45 days overdue, flagged red. Hilton deposited $620.
Now picture that running every Monday morning before you open your laptop. The log is always current. Nothing falls through the cracks. The overdue ones are surfaced so you know exactly who to chase. The 20 minutes of scrolling through your inbox each week to reconcile is gone.
That's one use case but with the same building blocks you can set up any automations for the workflows you already do.
Skills make the AI reliable. You've taught it how you work, in writing, once. Cowork puts it on your desktop so it can see your real files. Connectors plug it into the tools you already use. Scheduled tasks let it run without you. Each piece builds on the last, and together you get something that works for you, not something you have to work with.
You don't have to build all of this tonight. Most advisors I talk to are still on the browser tab and copy paste version, and that's fine. The gap between the version you're using today and the version that runs your operation in the background is smaller than it looks. It's mostly a question of writing down how you actually work, and then putting that writing somewhere the AI can see it.
