GPT Live 1: building on the setup we already have.

Building on what we already have.

We already have an AI customer service setup running with ElevenLabs. Behind the voice, there are tools, rules and systems we want to keep using.

The idea is simple: use that work to test GPT Live 1, OpenAI’s voice model. We’ll manage the connection on our own VPS: a server that links the OpenAI API to the systems we already use.

That’s what interests me about this test. I want to see how much we can reuse and what needs adapting to try this new technology in our operation.

What we want to test.

The plan is to set up that connection and test it with our existing customer service setup. The ElevenLabs configuration will remain available throughout this stage.

The first conversation left me excited. We still need to see how it performs on the phone, how it handles interruptions and whether it can carry out the tasks we need. That’s what we’ll be testing next.

How we’ll compare the solutions.

In the comparison, we want to understand how the Portuguese sounds, how quickly it responds and what we can do with each solution.

We want to understand what changes, what needs more work and whether this solution makes sense for our clients.