Wired published a piece on 15 November 2026 with the headline “AI DJs Completely Kill My Groove.” It says YouTube Music and Spotify want synthetic hosts to curate your playlists. The writer’s complaint is that the awkward banter and bogus song trivia ruin the vibe.

I couldn’t load the full article, so I’m sticking to that summary. I won’t guess at the details of how each service runs its DJ.

The excitement: a voice makes everything feel personal

The pitch behind a synthetic DJ is easy to follow. A voice feels like a person. A person feels like service. So put a voice on the playlist and the product feels warmer.

I don’t dismiss the technology. Synthetic voice is technically impressive and interesting. But I always ask the same first question about any new feature. Is there a need for it? What problem does this solve that listeners couldn’t already solve?

Nobody was stuck with a playlist and wishing a host would talk over it. People who want to hear someone talk about music already have radio and podcasts. The summary says the banter is awkward and the trivia is made up. That tells me the feature was built around what the technology can do, not around something listeners asked for.

What changed, and what didn’t

What changed is that a machine can now produce a voice that speaks on demand. That is real.

What didn’t change is why people open a music app. They want the next song, and they want it without friction. A voice that interrupts that, and gets facts wrong while doing it, adds a cost. It does not add value.

I hold a position on voice generally. It helps a lot on input. I use Wispr Flow to talk to Claude, and dictating instructions works well. In that instructional space, voice clearly helps.

It helps much less on output and in two-way conversation. A spoken answer is slower to take in than text you can scan. You can’t skim it, and you can’t copy from it. The playlist case is a small example of that.

What this means for a shop, a clinic or a service business

Take a clinic owner who hears that voice agents are the next step for customer contact. A patient wants to know if the clinic has a Saturday slot. The patient doesn’t want a synthetic receptionist to chat about it. They want the answer, and they want the booking.

The same goes for a shop. A customer asking where their order is wants the status. A friendly voice that takes 40 seconds to say it is worse than a message that says it in one line.

Customer conversations that turn into actions in your systems are text and data. More of those conversations will also be between software and software, and software doesn’t talk in voice. It exchanges data. So voice has limited room to grow in commerce conversations. The DJ story is the same mistake on a consumer product. A feature goes in front of customers because it can, and nobody asked whether they wanted it.

Where this connects to what I build

This is part of why AeroChat works at the text and data layer. It runs customer service and sales workflows in the chat channels customers already use, such as WhatsApp. It connects to your own data systems, so it can perform actions, not just answer questions. When a customer asks about an order, it checks the order and replies in the chat.

What a business should do

Ask the need question before you put a synthetic voice in front of customers. Write down the specific problem a customer has today. Check whether voice solves it better than a short text reply would. If you can’t name the problem, you are building around the technology.

Keep voice where it clearly helps. That is dictating instructions to your tools, and doing it faster than you could type.

For customer-facing work, put your effort into getting the answer right and doing the action. Make sure the customer’s question gets resolved in the channel they chose, with accurate facts from your own systems. Made-up trivia is what you get when a feature has a voice but nothing real behind it.

Source: Wired: AI DJs Completely Kill My Groove