Collect appointment requests
Give callers a clear way to share what they need. Prepare a calendar action for review when your integration is ready.
- Reception intake
- Service questions
- Reviewed calendar drafts
PRIVATE PILOT
Build a customer support or reception agent around your business information. Test the conversation, inspect its sources, and review any message or calendar action before it leaves your workspace.
Start with sample data. Live channels require separate setup and testing.
One workspace for the work around a conversation
Choose one repeatable request. Define what a good response looks like and where your team takes over.
Give callers a clear way to share what they need. Prepare a calendar action for review when your integration is ready.
Capture the question, use case, and preferred next step. Review the recipient and message before anything is sent.
Add your FAQs and policies. Inspect the retrieved source passages and decide when a request needs a person.
Open each step to see an example. This walkthrough runs entirely in the page. It does not contact a model, access a calendar, or send a message.
Answer published service questions, collect appointment preferences, and flag requests that need a person.
Do not confirm availability or a booking without a verified calendar result.
In the workspace: configure the role, instructions, language, and workflow in Agent studio.
A useful test checks the greeting, the next question, and the point where the agent should ask for help.
A passing simulation does not prove live model quality or call reliability.
Review the event, time zone, and consent record. Approval applies to that saved draft; execution is a separate step.
Real execution requires an enabled integration and a confirmed provider result.
If the provider outcome is uncertain, check its receipt before deciding what to do next.
Keep the question, the workflow, and the review in view. Start with the parts you can test locally.
Store text and Markdown knowledge. Preview retrieval before testing a model response.
Build a call-workflow simulation or publish a versioned automation. Inspect each saved run.
Inspect message and calendar drafts, record consent evidence, and approve a specific payload.
Read saved transcripts, leave feedback, and review uncertain requests without silently retrying them.
This is an early product for evaluation. It is not a generally available, managed calling service.
Configure agents, edit call-workflow simulations, search uploaded knowledge, run local automations, review drafts, and record contact preferences.
Provider-backed text, browser voice, selected calling routes, and external actions need credentials, access gates, consent checks, and live acceptance tests.
Check authentication, legal notices, regional rules, recovery, monitoring, and real provider behavior. Native apps also need builds, signing, and device testing.
Document the questions, approved source material, escalation path, and result your team will review.
The private workspace includes agent setup, call-workflow simulations, knowledge retrieval, local automation steps, action drafts, and a contact consent ledger. These local paths do not make calls or send messages.
Live routes are implemented as controlled pilots. They need operator configuration, credentials, consent checks, and end-to-end verification with the chosen provider. This preview does not promise live service availability, delivery, or production reliability.
You can store a number and its carrier details in the workspace. Actual calling requires verified ownership, a supported carrier setup, regional permissions, and live testing. The product does not purchase or port numbers for you.
The text runtime and selected voice pilots use provider adapters. A private text-inference endpoint is configurable. Every provider, model, and deployment still needs its own setup and acceptance tests; support is not universal.
The web workspace adapts to mobile screens. Android and iOS packaging source is included, but device acceptance, native builds, signing, and store release are separate steps. No store release is claimed.
Use a small test set. Inspect the results. Decide what needs to change.