Most agentic AI coverage assumes you write code. Claude Cowork is different. It runs real routines on your computer while you sleep and helps you stay on top of things you would otherwise miss.
Claude Cowork runs scheduled routines on your computer without any engineering required. The first sign it is working is often money already in your account or a task already finished by the time you wake up.
I don't write Cowork routines in a terminal. I describe what I want in plain English and pick a schedule. Claude Cowork runs on my desktop, hooks into the apps I already use, and does the work while I'm asleep or putting kids to bed. That's the whole pitch. The interesting part is what you actually automate once you believe it.
Southwest rebooking. Every night at 2am, a routine checks all my upcoming Southwest flights for the same exact flight and fare class at a lower price. Southwest keeps the difference as a credit. So far: over $200 in flight credits, zero mental overhead.
Auto-refunds. A bunch of retailers will refund you if the price drops after you buy. I know this. I still never do it, because finding the drop, reading the policy, and filling out the form is the kind of admin I will always deprioritize. A routine runs overnight when I have tokens left and has already clawed back money I would've written off.
Wish list research. I keep a note of stuff I want to buy. Sunday mornings, when I'm rarely at my computer anyway, Cowork works through the list: current price, refurbished options, rebates, whether a newer model exists. I enjoy doing this research when I have time. I just don't have time often enough. It's already surfaced sales below what I was prepared to pay on items I would've bought at full price.
Around 8pm, after I put the kids to bed, a routine scans conversations where I wasn't the last to reply, the thread is at least a week old, and we've been talking for more than three months. For each one it suggests a natural way back in.
When I come back upstairs, I have a short list of people I could reach out to and something specific to say. That window after bedtime is when I have the least energy for social maintenance and the most guilt about letting threads die. This one actually gets used.
Before I walk away from my desk for anything longer than a few minutes, I run a routine that reads my open apps, calendar, email, and recent texts, then writes a one-page "here's where you were" note for when I return.
I have ADHD. The expensive part of leaving a task isn't leaving. It's coming back and spending twenty minutes remembering which tab mattered. A summary from something that actually looked at my open windows sounds silly until you've lost an afternoon to re-entry twice in one week.
I basically communicate in one mode. Direct, a little long, assumes you already like me. That works with some people and lands flat with others. I know this in theory. I rarely adjust in the moment.
When I have an email that can wait, I point Cowork at it and ask it to read my history with that person and suggest framing, channel, and tone. Not magic. Just a forced pause before I send the same paragraph to someone who needed three sentences and a question mark.
Almost everything above works because Cowork can drive Chrome: open pages, read them, fill forms, click through checkout flows. That's how rebooking and refund requests happen without me.
It breaks sometimes. New page layout, weird multi-step form, site that loads like it's 2009. I retry. The wins are big enough that an occasional failure doesn't bother me.
Anthropic has a product overview and an intro webinar from launch. Cowork ships on paid Claude plans through the desktop app.
If you want to try it: claude.ai/referral. You get a discount on your first month and I get a credit.