- What happens when the AI does not know the answer?
- It says so instead of making something up. If the knowledge base does not cover the question, it flags the conversation and leaves it in the inbox for a person. We would rather have an honest "I don't know" than a convincing wrong answer.
- Where does the data end up, and who can see it?
- In our database, kept apart per company with Postgres Row-Level Security, so your team only sees your own account. Conversation text also passes through the model provider you chose, under your own key, so that part stays under your control. We do not use your content for anything other than running your conversations.
- Do I have to bring my own AI provider key?
- Yes. You choose the provider and the model, and the key is stored encrypted. Model usage is billed to you by that provider directly, with no markup from us.
- How does text get into the knowledge base?
- You paste it and save. It is split and indexed with no steps in between. Edit or add to it whenever you like, and the agent uses the new version from that point on.
- Can a person step into a conversation the AI is handling?
- Yes. Any agent can take the conversation, change its state and keep writing. The person always overrides the AI.
- Which languages does it work in?
- English and Spanish, across three separate layers: your team's interface, the one the other person sees, and the language the AI replies in. Each layer is set independently.
- Can I put it inside my own site or connect it to messaging apps?
- Not today. Conversations come in through the branded portal created with your account. If that does not fit how you work, better to know before you start.
- What do I need to get started?
- An account, a key from your AI provider, and some text to serve as the knowledge base. That is enough for the agent to start answering. You can begin with little material and build it up as you go.