Qualification without a form
HEA can ask who the visitor is and use that answer to bias timing, support tone, and next-step relevance for the rest of the session.
HEA-World gives each Human-Enhanced Agent a governed action layer: skills for what the agent can do, promotions for what it should drive, and safeguards that keep the conversation helpful before it becomes operational.
A normal website waits for visitors to find the right page. A governed HEA listens for intent, answers first, and then decides whether a skill or promotion is useful in that moment. The result is still a conversation, but it can now carry context, outcomes, and clear handoffs.
Skills are governed capabilities selected by the owner. They can collect missing fields, respect the visitor's active language, ask for explicit confirmation when needed, and keep the final action separate from the visible assistant text.
HEA can ask who the visitor is and use that answer to bias timing, support tone, and next-step relevance for the rest of the session.
A skill such as qualification_contact_capture_v1 can gather
contact fields, sanitize them, and ask before sending anything.
Support intake can gather email, reason, and details, then keep the request in a pending state until the visitor confirms the send.
Promotions can appear as cards or direct links when they are relevant, with repetition controls so the HEA stays useful instead of pushy.
When a skill has a safe backend adapter, HEA can move from collected fields to a bounded action such as a read-only lookup or owner email.
If a request is unsafe, incomplete, or better handled by a person, the HEA can fall back to a support path instead of pretending it acted.
HEA keeps the answer-first personality path separate from the operational action layer. Skills and promotions rank what might help next, but they do not replace the conversation or silently execute sensitive actions.
The visible conversation stays natural while hidden state tracks durable facts.
Audience, topic, timing, and repetition rules decide which skill or promotion fits.
The HEA asks for the minimum missing information in the active conversation language.
Execution-backed requests wait for a clear yes/no confirmation before acting.
Outcome states such as sent, failed, declined, or looked up stay visible to operators.
A promotion can invite a visitor to book, watch, download, buy, or learn. A skill can collect the details that make that next step useful. HEA uses both as governed suggestions, not as a hidden branching workflow engine.
Owners choose the skills and promotions available to a HEA, then keep rollout under the same train, test, publish, and review discipline as the rest of the platform.
Agentic conversation is only useful if the business can inspect what happened. HEA keeps the setup, runtime state, and review surfaces aligned so teams can trust the system before scaling it.
The editor exposes audience profiles, selected skills, promotion content, setup fields, timing, repetition, and draft status in one place.
Structural audience changes require Train skills before publish, while ordinary copy or ordering updates still follow the draft/publish path.
Review surfaces show captured facts, active request states, and Skills & promotions used so operators can spot what the HEA did.
Start with a HEA trained on your knowledge, then add the skills and promotions that match your real visitor journeys: qualify, support, hand over, share the right asset, or execute a safe confirmed action.