How Adanse works, end to end.
Reference material for the people who configure, govern and answer for an AI workforce.
Getting started
The operating layer between your business and the AI acting on its behalf.
Every record belongs to one workspace; membership decides access.
Connect a system, review Business DNA, hire one AI employee at R1.
Member, decider and owner — deciders can approve, reject and undo actions.
Understanding the business
Typed, sourced facts with freshness and confidence, linked to entities.
Declared behaviour compared against observed behaviour over time.
Entities, systems, employees and decisions as nodes and typed edges.
Curated document sets scoped to departments and employees.
Workforce
A named worker with a department, remit, skills and an autonomy boundary.
Reusable, versioned procedures an employee is permitted to run.
Hybrid human and AI teams with shared objectives and escalation paths.
Registered tools and actions an employee may call, with scopes.
Governance
From suggest-only to fully delegated, applied per employee and per action.
Conditions that must hold before an action executes, evaluated at decision time.
Reasoning, inputs, confidence basis, policy, approver, result and value.
Append-only evidence written as work happens, linked back to its decision.
Autonomy levels
Levels are enforced per action, not per employee alone — a single employee can be R4 for reporting and R1 for payments.
One unit of execution. No tokens to reason about.
Customers should not have to price model tokens, workflow steps, browser hours or sandbox CPU seconds. Execution is converted into a single unit, so the commercial contract does not change when the infrastructure underneath it does.
Every run shows the credits it used in your workspace, so you can see exactly where the allowance went.
Reasoning, workflow execution, API and MCP actions, knowledge and memory retrieval and evaluations draw on the standard allowance. Browser execution, code sandboxes, premium models and large document processing draw at higher rates or against their own allowance.
Cost is only meaningful next to outcome.
Billing in Adanse is not a credit meter. Spend is presented against the measured value the workforce produced in the same period, from the same records.
Illustrative figures. In your workspace these come from the outcome ledger.
Committed capacity, not a fixed ceiling.
Enterprise agreements are composed rather than listed, and priced annually against committed execution.