Start here
How Korenact works
Follow a request from the first message to a confirmed action or staff handoff.
The path a request takes
Every turn is handled in the context of your organization, the selected agent, and that agent’s rules. Korenact can stop to ask a question, verify identity, wait for approval, or hand over. It does not need to complete every stage to answer a simple approved question.
- ReceiveA configured channel identifies the workspace and agent.
- UnderstandIdentify the request and collect missing details.
- CheckApply identity, permissions, data access, and approval rules.
- Act or hand overUse a supported connection, ask for clarification, or involve staff.
- RecordKeep the transcript, decision trace, action status, and references.
Read the decision trace
| Stage | What to look for |
|---|---|
| Conversation | The incoming turn and conversation context. |
| Identity | Whether a person is unknown, matched to a record, or verified. Matching alone does not establish verification. |
| Intent | What the person appears to want, and the confidence in that interpretation. |
| Details | Required information still needed to carry out a request. |
| Organization rules | Whether the action and required data access are allowed, and whether approval or verification is required. |
| Action / System update | The attempted operation and provider response. Read the action status and reference. |
| Person | The handover reason and intended staff destination. |
The conversation detail page shows the stages actually reached. Older conversations may have a shorter trace reconstructed from their recorded actions. A stop at a rule or missing detail is a useful explanation of why a system update did not happen.
Approved facts and optional AI
Rules, exact commands, approved knowledge, and typed decisions are used before optional generative models. Intelligence settings control interpretation, response style, and spend. A model can recommend a route or wording; organization permissions still determine which actions may run.
If a provider is not configured or a budget is exhausted, the available rules, approved answers, clarification, and staff handoff remain the fallback. Recording staff feedback helps review the workflow; it does not automatically change its rules.
An appointment request
A caller asks to book a visit
- The caller contacts a configured line assigned to an active scheduling agent.
- The agent identifies the booking request and asks for missing date, time, or service information.
- Identity and data-access checks run according to the rules and action requirements.
- If the booking needs approval, staff see it in the Review queue.
- A permitted request reaches the configured calendar. A confirmed response produces an Executed action and provider reference.
- If the provider result is uncertain, staff check the calendar before submitting another request.
Several specialists, one conversation
An agent team connects a coordinator with selected specialists. Phone numbers can inherit the organization team, use an individual override, or require the same team on every number. Korenact AI selects the relevant specialist for each supported request while preserving the conversation and that specialist's policy checks.