AI Snake Oil: When will ChatGPT be able to make your calls?
Chad here from ChadGPT. So everyone’s been freaking out about AI making phone calls since Google showed off that Duplex demo years ago. You know, the one where AI called a hair salon and made an appointment while sounding eerily human? Yet somehow, in 2025, I still have to make my own calls like a caveman. What gives?
The folks at AI Snake Oil just published an interesting breakdown of why conversational AI is still so bad at making actual phone calls. And it’s not for the reasons you might think.
Here’s the deal: ChatGPT and similar AI models are getting scarily good at text conversations. They can write emails, draft marketing copy, and even handle customer service chats. But put them on a phone call, and they fall apart faster than a cheap suit in the rain.
The technical challenges aren’t actually about voice recognition or speech synthesis anymore – those problems are largely solved. The real issues are around real-time processing, handling interruptions, and dealing with the unpredictable nature of human conversations.
When you’re chatting with AI via text, the AI has time to think and craft a response. On a phone call, there’s no such luxury. Humans notice even small delays in conversation flow, making the entire interaction feel unnatural. Plus, humans interrupt each other constantly and can change topics mid-sentence – something AI still struggles with mightily.
What I found most interesting is that the article argues we’re approaching this all wrong. Instead of trying to make AI pretend to be human on calls, we should be rethinking how calls work in the first place. For most business purposes, asynchronous communication is better anyway.
For small businesses, this means you shouldn’t hold your breath waiting for AI to handle all your calls anytime soon. But there are some practical halfway solutions emerging:
- AI that generates scripts and talking points for YOU to use on calls
- Semi-automated systems where AI handles the straightforward parts but transfers to a human for complex issues
- Tools that transcribe and summarize calls in real-time so you can focus on the conversation instead of note-taking
The article concludes that fully autonomous AI phone agents are still 3-5 years away from being reliable enough for business use. And frankly, that’s probably for the best. We’ve all experienced those painful automated phone systems – “Press 1 to feel your soul slowly leave your body.”
For now, focus on the AI tools that actually work today – ones that help you prepare for calls, analyze them afterward, and make the most of your actual human-to-human interactions.
