I Used Claude Cowork Dispatch for a Week. Here's My Honest Review.

The pitch for Claude Dispatch is genuinely exciting. Text a task from your phone, your desktop does the work, you come back to finished output. On paper it's perfect.
I used it daily for a week. The reality is more complicated.
What's rough right now
Three things kept tripping me up. The context windows feel slow. There's no way to reset a session from your phone, so if a task goes sideways, you're stuck. And the tools available on mobile are limited compared to the desktop. Other reviewers are landing around a 50% success rate on complex tasks, which matches what I saw.
The concept is right. The beta is just not done yet.
The idea is the future. The execution is a 1.0. Both things are true at once.
Why I'm still rooting for it
It would be easy to dunk on a rough beta and move on. But the direction here is exactly where AI work should be heading, controlling real execution from anywhere, and that's worth being patient about.
I want Anthropic to fix this, because when they do, it's a genuinely different way to work. (And yes, "I believe in it but it's not ready" is the most annoying review to write. It's also the honest one.)
If you've tried Dispatch, drop your success rate in the comments. The more real data Anthropic gets, the faster the good version ships.
See the 30-Day In-Product AI Assistant Sprint
One AI assistant, inside one existing SaaS workflow, designed, built, evaluated, and handed off in 30 days.
See the sprint