A couple of weeks ago, Anthropic launched Claude Dispatch. The idea is pretty straightforward: you pair your phone with your laptop, text Claude a task from wherever you are, and it goes off and does it on your desktop while you’re away.

I’m a paying Claude subscriber and love the product, so naturally I had to try it. I set it up on my personal laptop last night and came up with two real tasks.

Task 1: The budget spreadsheet

I had a household budget spreadsheet that I wanted to tidy up. it includes our monthly costs like fixed and budgeted expenses, some calculations, nothing dramatic. I typically dedicate some time every week updating it manually. I figured this was a good test of Dispatch. Fire it off from my phone, go for a walk, come back to an updated spreadsheet.

And Claude did open the spreadsheet. It did some calculations. It moved some things around. But between pairing the phone, keeping the laptop awake, and approving permissions every time Claude wanted to do something, I felt that the whole thing took longer than if I’d just opened the file on my phone and done it myself. We’re talking a two-minute task that I turned into a fifteen-minute setup exercise.

Task 2: The house-buying roadmap

More ambitious this time. My wife and I are starting to think about buying a house, and I wanted to put together a visual roadmap of all the steps involved. Financing, viewings, bidding, lawyers, inspections. You name it. I wanted to see if I could build a slide deck that I could show her in the evening and we could talk through after dinner.

I dispatched the task while I was hanging out with my son. The prompt was for Claude to build a slide deck with a timeline, milestones, and key considerations at each stage mentioned above.

When I got back, the deck existed. But it was ugly. The whole deck had that unmistakable “AI made this without anyone watching” vibe. I mean it was good enough for my use cases but never in a million years something I’d present in a corporate setting. Equally, Claude couldn’t iterate on design from the phone. I needed to go back to the laptop to give feedback on layout, which defeated the whole “do it while I’m away” premise.

The PM in me sees the pattern

Every PM has shipped something like this at some point. A feature that works technically but doesn’t solve a problem people actually have.

The question worth asking is: why?

To me, Dispatch feels like a textbook case of technology-push. You build something because you can, not because users are screaming for it. The phone-to-desktop agent loop is awesome tech by all measures. But the moment you ask “who has this problem today?”, the answer gets fuzzy for me.

Personally these are the things that are working against it:

• The overhead problem: My budget spreadsheet took two minutes by hand. Dispatch turned it into fifteen. The gain isn’t big enough to justify the pain of setup.
• The segmentation problem: Casual users like me don’t need it. Power users already have scripts. The sweet spot is someone who can prompt but can’t code, and that’s not a massive customer profile yet.
• The context problem: I can’t use Dispatch on my work laptop. Corporate restrictions mean no Cowork, and that’s where all my jobs to be done actually live. Enterprise adoption means IT approval, security reviews, data governance. All the things that slow tool adoption in large organisations.

Don’t get me wrong: I do think Dispatch will matter once enterprise support catches up and the agent gets reliable enough that you don’t need to babysit it.

So, the product is not bad it’s just early. Sometimes as a PM you build the right thing before the world is ready for it.