Last week I published a LinkedIn post which showed a GPT I created. The GPT receives requests to make phone calls and completes any task asked of it (e.g. call a restaurant and make a booking).

The process was pretty straightforward, using ChatGPT GPT capability, and providing it with access to a third party generative AI telephony tool, Bland.AI. By doing this, the GPT is able to take a request provided to it by a user “e.g. call restaurant <insert number> and make a booking for 8pm” and pass it to Bland.AI. You can see it in action in the LinkedIn post (I’d open up access but each call costs $$$).

Ok, the learnings…

1. It’s amazing what can be done with no-code tools, the GPT pretty much coded itself with some plain English prompts from me.
2. Speed to deploy is no longer a barrier, this demo took less than 10 minutes to build.
3. Massive phishing issues are about to emerge, with an additional 5 minutes (of recording my voice) I was able to train Bland.AI to make phone calls in my voice.
4. Generative AI can deliver real value today, it’s not something worth over strategising about, as Nike might say “Just do it!”