AivahHelp CenterGo to Aivah
⌘K
No guide matched “”

Try fewer words or contact support for help.

Contact support

What Aivah does

Understand the parts of an Aivah AI employee and the safest way to begin.

Who this is forNew account owners, team leads, and anyone evaluating a first Aivah workflow

You will achieveChoose a focused first use case and understand how knowledge, instructions, appearance, actions, and results work together.

Before you beginReview the requirements before changing anything.

Last reviewed 10 July 2026Owner: Product EducationCurrent guide
On this page
  1. Before you begin
  2. Steps
  3. 1. Choose one purpose
  4. 2. Identify approved business knowledge
  5. 3. Define instructions and boundaries
  6. 4. Decide how it should appear and act
  7. 5. Test privately in Playground
  8. 6. Publish one experience and review the result
  9. Expected result
  10. Status meanings
  11. Usage credit impact
  12. Information stored or shared
  13. Limits and permanent actions
  14. Common problems and recovery
  15. Next step

Use this guide if you are new to Aivah or need to explain the product to your team. You will leave with a practical first project rather than a broad, difficult-to-test assistant.

Inside the workspace, an AI employee is usually labelled Agent. Agent is the on-screen workspace name for the same worker you create and manage.

Before you begin

Bring one repeated task that has a clear audience and result. Useful starting points include answering common support questions, qualifying visitors, guiding onboarding, handling basic phone enquiries, supporting training, or creating content from approved information.

Steps

1. Choose one purpose

Write down the job the AI employee should perform and the person it should help. A focused purpose is easier to teach, test, and improve.

2. Identify approved business knowledge

Choose one reliable source to begin with, such as a current product page, help article, policy, guide, presentation, or written note. Use information with a clear owner and review date.

3. Define instructions and boundaries

Decide how the AI employee should answer, what it must not guess, when it should ask a question, and when it should involve a person.

4. Decide how it should appear and act

Choose an appearance and voice that fit the job. Add connected tools only when an outside action is necessary, and require confirmation before sending messages, booking something, changing records, or taking another consequential action.

5. Test privately in Playground

Use Playground to check answers, boundaries, handoff, connected actions, voice, and appearance before anyone outside your team can use the AI employee.

6. Publish one experience and review the result

When the test passes, use Shared to publish in one place. Use Insights and the underlying records to review conversations, leads, calls, quizzes, and usage that are available for your setup.

Expected result

You can describe the AI employee through six parts:

PartWhat it means
PurposeThe job it performs and the result it should help produce.
Business knowledgeThe approved information it may use.
InstructionsIts behaviour, boundaries, actions, and handoff rules.
Appearance and voiceHow it looks and sounds in supported conversations.
Connected actionsThe outside services and approved actions it may use.
ResultsThe activity you can review and use to improve it.

Status meanings

This overview has no single product status. In task guides, pending or processing usually means wait, completed means the item is ready for its next test, and failed means open the details and recover before publishing. Always follow the exact status explanation on the relevant screen or guide.

Usage credit impact

Planning a first use case does not itself use Aivah activity. Supported conversations, content generation, voice, and other activity may use usage credits. Check the balance and choices shown in your workspace; this Help Center does not publish a fixed calculation.

Information stored or shared

Your plan is not stored in Aivah until you enter it during setup or employee creation. A saved AI employee can contain its name, purpose, instructions, knowledge sources, and selected preferences. Published experiences can show configured information to visitors, and connected services may receive information required for an enabled action. Review each experience, service, and permission before use.

Limits and permanent actions

An AI employee should not guess beyond approved knowledge or make high-stakes decisions without suitable human oversight. Do not use private customer records as general business knowledge. Give connected actions the smallest useful permission set.

Common problems and recovery

  • The job is too broad: reduce it to one audience, one repeated task, and one result.
  • Answers conflict: remove outdated or contradictory sources, then test again.
  • The employee acts too quickly: add a confirmation rule and a clear human handoff.
  • There is no obvious result: choose something you can observe, such as a completed setup step, a qualified lead, or a prepared handoff.

Next step

Turn the idea into a testable job in Choose your first job.

Keep going