Skip to main content

Getting Started Overview

Subbasis helps you move from idea to governed execution in practical, controlled steps. This guide is the best entry point for turning a business operation into a durable workflow.

Overview

You will use this path to:

  • choose one operation with clear value
  • model it as a workflow
  • assign agents with explicit boundaries
  • connect required systems with governed connector usage
  • run with approvals, evidence, and iterative improvement

Why it matters

Teams often start AI initiatives with chat experiments and isolated automation. That is useful for exploration, but operational reliability needs a stronger model.

Subbasis adds runtime governance so execution remains reviewable when conditions change: users intervene, connectors fail, policies evolve, or workflows pause and resume.

How it works in Subbasis

Subbasis implementation usually follows five phases:

  1. Define one business operation as a Workflow.
  2. Assign one or more Agents with limits and permissions.
  3. Connect required systems through governed connectors.
  4. Run with approvals and evidence.
  5. Review outcomes and improve.

Main concepts to keep in mind

  • Workflow: durable operation model, not only trigger automation.
  • Agent: role-based execution participant.
  • Approval gate: explicit checkpoint before sensitive actions.
  • Context Intelligence: operational state continuity.
  • Evidence: reviewable execution-grade history.

Example scenario

Use case: support escalation.

  • Intake agent triages request.
  • Specialist agent proposes next action.
  • Approval gate validates external action if required.
  • Connector executes update in ticketing/CRM system.
  • Evidence captures path and outcomes for review.

Who this path is for

  • Developers building workflow-driven operations
  • Consultants packaging reusable execution patterns
  • Teams testing governed agentic execution in production-like environments

What to configure early

  • one workflow with 3-7 explicit steps
  • one or two agent roles
  • one approval condition for high-impact actions
  • one connector path
  • one evidence review routine

Limits and deployment notes

  • Start with small scope; validate governance behavior first.
  • Feature depth depends on plan and deployment setup.
  • Avoid assuming advanced distribution/local patterns until explicitly configured.