Role Assistants turn business roles into governed AI assistants — Sales Assistant, Project Assistant, Finance Assistant — that help employees complete recurring work using trusted business context, governed capabilities, and evidence from completed tasks.
This is the third pillar of AI Fabrix. Role Assistants are human-owned and role-scoped. They are not a generic agent platform and not prompt orchestration. They execute inside Operational Trust and Enterprise Knowledge.
How the platform works (developers): The architecture tier uses Digital Worker (
systemKind: digitalWorker) for generated metadata, runtime, MCP, and OpenAPI. Customer UX and adopt docs use Role Assistants and{Role} Assistantnames only. See Platform architecture.
Why it matters
Most AI agent platforms start from prompts. AI Fabrix starts from the business role. A role defines what the user is allowed and expected to do — not only what they typed in a chat box.
A Role Assistant understands which role is active, which business context matters, which capabilities are available, which approvals are needed, and which outcomes are expected. The active role is the operating boundary.
Without Enterprise Knowledge, an assistant would lack business context. Without Operational Trust, it would lack authority. Without Evidence Fabrix, the organization could not learn from completed work — see Learning from completed work and How assistants are created.
How assistants are created
A Role Assistant is derived from business context:
Role + Enterprise Knowledge + Governed Capabilities + Operational Trust + Evidence = Role Assistant
The platform already knows what the role does, which business entities it works with, which capabilities apply, and which rules and evidence improve prior work. Employees do not hand-orchestrate every assistant from scratch.
Assistants are assigned to people and scoped to roles from enterprise identity — not anonymous chat sessions.
Business context replaces prompt engineering
Assistants are guided by role, task type, resource type, and governed capabilities — not by perfect prompts. A Sales Manager gets a Sales Assistant; a Project Manager gets a Project Assistant. The role defines the work; the assistant helps perform it.
When business metadata and cross-system relationships are modeled well, assistants need less prompt tuning because context is already governed.
Tasks, outcomes, and evidence
Role → Role Assistant → Task → Governed Capability → Outcome → Evidence → Skill Growth
Assistants request capabilities; the platform executes through the approved path and records evidence. They do not bypass approval, override governance, or silently change business rules.
Tasks have objectives and expected outcomes. Evidence captures what was used, what ran, and what resulted — feeding operational memory and skill growth.
Operators who run Role Assistants day to day rely on the same role boundaries and certification state as developers configure during integration.
Business owners can see which capabilities an assistant may request and which outcomes require human approval before execution proceeds.
Skill growth
Role assistants become more reliable as they complete work. Skill levels (Trainee → Capable → Trusted → Expert) show maturity internally; customer-facing cards emphasize Trusted and Expert where shown. Promotion reflects useful outcomes and confirmed impact — not expanded authority or fewer approvals.
Limits
Full Role Assistant task flows, skill promotion, and impact dashboards vary by deployment. See Evidence Fabrix for what evidence and learning features are live in your environment.
Example
A Sales Manager starts a weekly pipeline review. The Sales Assistant uses in-scope customers, open deals, renewal signals, missing data, and prior review evidence to prepare the review and recommend follow-ups — requesting approval before any governed action runs.
Business value
Role Assistants help enterprises deliver role-based AI assistance without building agents from scratch, reduce prompt dependency, keep humans in authority, and turn business context into repeatable, governable work.
One-line summary
Role Assistants turn business roles into governed AI assistants that complete recurring work through approved capabilities — not generic agents.