Autonomy

Define how faryour AI Agent can go.

Choose the autonomy level of each Secure AI Agent. Determine what it can do on its own, which actions require approval, and what stays blocked.

Autonomy levels

You decide how much control to delegate.

Not every task needs the same autonomy level. Configure each agent according to the process, the risk, and your company's policies.

  1. Query

    The agent accesses information and answers requests, without making changes to the connected systems.

    Example

    Check the status of an opportunity in the CRM.

  2. Recommend

    The agent analyzes the context and prepares an action, but the decision stays with a person.

    Example

    Prepare a reply to a customer before sending it.

  3. Execute with approval

    The agent prepares and executes tasks after receiving human authorization.

    Example

    Request approval before sending a proposal.

  4. Execute autonomously

    The agent can execute previously authorized actions without approval for each task.

    Example

    Automatically update records within the defined rules.

Control per action

Autonomy does not have to be all or nothing.

Within the same Secure AI Agent, different actions can follow different rules. Routine tasks can be automated while more sensitive operations still depend on human approval.

Explore governance
  • Query financial dataRead-only accessAutomatic
  • Update informationWithin the configured rulesAutomatic
  • Send external communicationHuman confirmation requiredRequires approval
  • Change critical configurationAction not permitted for this agentBlocked
Autonomy with control

Automate what makes sense. Keep people where it matters.

The goal is not to remove people from every decision. It is to give each Secure AI Agent exactly the autonomy it needs to do its work.