A Role Assistant is a human-owned, role-scoped business capability generated and controlled in AI Fabrix.
A Business Role is the approved job function under which an identified person performs the work. AI Fabrix automatically generates a Role Assistant candidate from the Business Role and the approved business foundation around it. People validate and release it before it becomes available for operational work.
No arbitrary tools, independent identity or authority.
No hidden workflow and no replacement for the responsible employee.
The active Business Role determines what data may be supplied, which approved actions may be requested, where human approval is required and when work must wait or stop.
Enterprise Knowledge supplies context. Operational Trust enforces current authority. Enterprise Runtime owns execution and each next step. Evidence Fabrix owns the reusable Evidence lifecycle.
Who is responsible. The approved job function under which the identified person performs the work.
Which scope applies. Reusable boundaries such as region, legal entity, business unit, customer, project or amount.
What may be requested. Approved business actions rather than unrestricted vendor APIs or credentials.
Teams describe the outcome, Business Role, entities, authoritative information, permissions and Dimensions, governed operations, approvals, Evidence requirements and business measure.
AI Fabrix uses those approved enterprise definitions and certified capabilities to automatically generate the Role Assistant candidate that business and technical teams inspect and test before release.
The candidate does not start from an empty prompt or an arbitrary list of tools. Enterprise Knowledge supplies the reusable business model and Business Context Map. The Business Role determines the operating boundary, and the Role Assistant receives the approved capabilities applicable to that role.
Generation produces a versioned candidate configuration, not operational authority.
Avoids rebuilding the assistant for every user or interface.
Makes the proposed boundary visible.
Keeps generation separate from authority.
Outcome, sources, role, permissions and retrieval. Use representative business cases.
Operations, approval paths and Evidence requirements. Use representative business cases.
Missing information, denials, waits and safe stops. Use representative business cases.
Depth and approvals depend on organisational risk and policy.
Only an approved version is released. Availability never widens authority.
A person may use a Role Assistant through an approved customer application, Teams or another supported AI interface.
A Role Assistant works toward a defined outcome using current governed context rather than a hidden, fixed workflow.
It can ask, wait for human approval, preserve a denial or stop safely.
The account manager retains responsibility. The authorised approver retains authority for an exception.
Later Service and Finance Assistants can reference approved customer, agreement and authority definitions while validating their own purpose and scope.
Reuse occurs through the shared enterprise foundation and Business Context Map—not by copying prompts, tool lists and orchestration between agents.
Reveal missing information and data-quality problems.
Identify repeated exceptions and approval bottlenecks.
Material changes return through validation and approval.
Signals never expand authority or remove approvals.
Change the interface without rebuilding the foundation.
Move from generic AI interaction to repeatable business work.
