Two different voice agents, same SDK
We're a Chilean startup building AI sales agents for small businesses across Latin America. We're part of their startup grants program, so I'll say that upfront. We built two voice agents on their platform, doing very different jobs. The first one replaced our setup form — instead of filling in fields, a small business owner talks for a few minutes and their agent is configured. The second one guides campaign creation: it asks what they're trying to achieve and routes them to the right channel, without them needing to understand the technical difference between one and the other. Client tools are what made both possible. Having the agent show a field, advance a screen, or save an answer from inside the conversation is what separates a real conversation from a form you talk to. And the fact that the same setup worked for two agents with completely different scripts saved us from rebuilding everything the second time around. Two honest notes: the Flash model occasionally produces audio artifacts on uncommon names, and the agent authentication toggle is hard to find — it's under Settings, Advanced, Security, not where the docs suggest. Neither slowed us down. Our users are business owners aged 55 and up who don't want to fight with software. Letting them talk instead of type changes who can use the product.