EmployAIQ

Governance

AI Employees that answer to people

EmployAIQ is built around one idea: an AI Employee should only do the work it has been given, with the tools it has been allowed, at the level of independence it has earned — and a person can review its recorded actions and stop it at any time.

Book Your AI Workforce Blueprint(opens in a new tab)Request Sandbox Access

Reasoning and control

Cognition guides the reasoning. Governance is the code-enforced control layer around tool use.

However an AI Employee reasons, a requested tool action is checked by the platform against its permissions, trust level and applicable approval requirements before the tool executes.

  1. Step 1: Role

    Sets the job

    The role becomes the AI Employee's operating brief.

    • Responsibilities and operating instructions
    • Capabilities and skills
    • The tools the role is approved to use
    • Reasoning guidance for the role, where configured

  2. Step 2: Context and memory

    Informs the reasoning

    What the AI model can draw on for the current piece of work.

    • Recent conversation context
    • Relevant business knowledge from your workspace
    • Lessons from human feedback, where available

  3. Step 3: Cognition

    Guides the reasoning

    The AI model interprets the situation within the operating brief.

    • Decides whether to reply or to request an action
    • Can only request tools the role makes available

  4. Step 4: Governance

    Controls the action

    Separate platform code checks each requested tool action before it runs.

    • Allowed tools and the AI Employee's trust level
    • Human approval requirements and any approval rules you have saved
    • Spending caps on actions that move money
    • Result: the action runs, is held for approval, or is refused — and is recorded

  5. Step 5: You

    Stay in control

    People oversee the work and set the boundaries.

    • Approve or reject held tool actions
    • Set or change trust levels
    • Pause workspace autonomy or deactivate an AI Employee
    • Review action and approval history

  6. Step 6: Outcomes

    Measured and reported

    Results can be reported on your dashboard for human review.

    • Evidence for people to review — results do not change the AI Employee's decisions on their own
Role-defined, model-reasoned, governance-enforced. Conversational replies are guided by the role and your content; they are not approved message by message. Approval applies to tool actions.

What this means in practice

  • Conversational replies are not approved message by message; they follow the role's instructions and your content.
  • Applicable tool actions can be held for a person to approve or reject before they run.
  • Higher trust does not mean unlimited autonomy: the role's tool permissions always apply, and actions that move money need spending caps.
  • Pausing workspace autonomy holds autonomous replies and outbound email for review; a deactivated AI Employee cannot use tools or act on its own.
  • Tool actions and approval decisions are recorded and shown in your workspace's activity history.

For technical teams

EmployAIQ composes a role-specific operating brief and runs an AI-model tool-calling loop within a capability-derived tool allowlist. A separate, code-enforced governance layer applies trust and approval decisions before tools execute, while action and approval records provide operational visibility.

Cognition is the role-specific reasoning guidance an EmployAIQ AI Employee uses when interpreting its work. An AI model applies that guidance to the current conversation, relevant business knowledge and available lessons, then decides whether to respond or request an approved tool.

Roles and responsibilities

Who does what

Governance starts with clear responsibilities — for the AI Employee and for the people who oversee it.

The AI Employee

Does the work its role defines with the tools its role allows. When a tool action needs approval, the platform holds it — that decision is not left to the AI.

Approvers

People you designate review gated actions, approve or reject them, and decide which routine decisions become standing rules.

The workspace owner

Chooses roles, sets trust levels, confirms promotions, connects business systems, and can pause or deactivate any AI Employee.

Capabilities and skills

A defined role, not open-ended access

Each AI Employee is assembled from a role template, the capabilities and skills that role needs, and only the tools you approve.

Role

Each AI Employee starts from a role template that defines its job, its responsibilities and the boundaries of what it should and should not do.

Capabilities

Capabilities are the areas of work the role is approved for — for example email outreach, content publishing or knowledge lookup.

Skills

Skills are the working methods attached to a role: how it qualifies a lead, structures an article or handles a support question.

Controlled tools

Tools reach an AI Employee only through approved capabilities and connected integrations. If a tool is not approved for the role, the employee cannot use it.

Your knowledge

Employees answer from your own business content — website, documents and policies — added to the workspace knowledge base.

Human oversight

A trust level, approval rules and designated approvers decide which tool actions run on their own and which wait for a person.

Trust levels

Autonomy is earned, one level at a time

Each AI Employee works at one of four trust levels. The level decides which tool actions run on their own and which wait for a person. Higher trust is never unlimited autonomy.

  1. L1: Supervised

    Applicable tool actions — sending an email, publishing content, updating a record — wait for a person to approve them.

  2. L2: Rule-assisted

    Works like L1, except where you have saved an approval as a standing rule: matching routine actions can then proceed. Actions that move money honour only narrowly scoped rules.

  3. L3: Trusted

    Routine tool actions run without waiting for approval. Actions that move money still wait for a person.

  4. L4: Autonomous

    Tool actions run within the role's permissions; actions that move money run only within spending caps you set. Autonomous does not mean unrestricted.

  • Every AI Employee starts at L1.
  • The system recommends a promotion; a person confirms it.
  • The owner can lower an employee's trust level at any time.
  • Higher trust never removes the role's tool permissions.
  • Conversational replies are not approved message by message; approval applies to tool actions.
  • The approval check runs in platform code where tools are executed, not in the AI model.

Controls

What operates today, what is configured with you, and what is in development

Each control is labelled so you know exactly where it stands.
  • Operates today: available in the platform now.
  • Configured per customer: set up for each customer during guided onboarding.
  • In development: not yet available.
Operates today

Controlled tool access

Tools reach an AI Employee only through the capabilities approved for its role and the integrations connected in your workspace. A tool that is not approved for the role is not available to it.
Operates today

Human approval

Applicable tool actions — for example sending an email, publishing content or changing a record — can be held in an approval queue for a designated approver to approve or reject. The check runs in platform code where tools execute, not in the AI model. Conversational replies are guided by the role's instructions and your content; they are not approved message by message.
Operates today

Approval rules from your decisions

Supported in the platform: when you approve a routine action, you can save that decision as a standing rule for that AI Employee. Workspaces start without rules; you add them as you go.
Operates today

Action and approval history

Tool actions, approval decisions and trust-level changes are recorded, and your workspace shows an activity history of what each AI Employee did.
Operates today

Escalation to a person

Where the role includes it, an AI Employee can hand a conversation to a person on your team, who is notified by email and takes over. The AI Employee stands down on that conversation.
Operates today

Stop switches

Pause autonomy across the whole workspace with one switch — autonomous replies and outbound email are then held for a person — or deactivate an individual AI Employee. Lowering an employee's trust level is also available at any time.
Operates today

Workspace data separation

Each business works in its own workspace. Data access is separated per workspace at the application level.
Operates today

Encrypted credentials

Stored credentials — such as AI model keys and integration secrets — are stored encrypted.
Configured per customer

Policies, limits and connected systems

Approval policies, approvers, usage limits, outbound email authorisation and which business systems an employee may use are set up with you during guided onboarding.
Configured per customer

Additional channels

Website chat and email are available today. Additional channels are configured case by case during guided onboarding, where supported.
In development

Automatic trust adjustment based on performance

Adjusting an employee's trust level automatically from its performance is in development. Today, trust changes are made by people.

Guided production configuration

Production is configured with you

EmployAIQ is live and available through Founding Access. Public starting prices are visible, while production deployment and sandbox access currently use guided onboarding so that governance policies, usage limits and approved business systems can be configured safely.

If your organisation has specific security, identity, data-handling or approval requirements, we discuss them during scoping. Enterprise deployments

Plan your AI workforce

See which AI Employees fit your business

The governed AI workforce platform from AIToken Labs.

Book Your AI Workforce Blueprint(opens in a new tab)Request Sandbox Access

Already using EmployAIQ? Sign In

Founding Access — Live Platform with Guided Onboarding