Skip to content
← Resources

Agent Types &
Brand Mapping.

Preamble

What Is an AI Agent?

An AI Agent is not a person. It is not a chatbot. It is not magic.

Working Environment (homedir, data, permissions)
+ LLM Model (inference engine)
+ Agent Harness (agent loop + guardrails + system prompt + core tools)
+ Tooling Layer (skills, plugins, MCP servers)

Remove any one of these components and what remains is not an agent — it is a language model generating text in a vacuum. This definition matters because the industry has stretched "AI Agent" to cover everything from a ChatGPT wrapper to a full production system. We do not.

Layer 1

Engineering Types — What It Is

Task Agent

Bounded, one-function agent. Executes a defined sequence with tool access and human milestone check-ins.

Example: A Collections Desk automaton that chases overdue invoices — one workflow, clear exit criteria.

Process Agent

Spans an end-to-end business process (e.g. procure-to-pay). Maintains state across steps and hands off between sub-agents.

Example: A Procurement Desk that handles vendor selection → PO creation → delivery tracking → invoice matching.

Orchestration Agent

Coordinates other agents. Delegates subtasks, aggregates results, routes edge cases. Does not execute tasks directly.

Example: A Directorate automaton that monitors Bureau Desks, escalates exceptions, and reports to the human team lead.

Knowledge Agent

Specialized in retrieval, synthesis, and explanation from a defined knowledge base. Read-only by default.

Example: An agent that answers employee policy questions from a company handbook — it knows, it doesn't act.

Monitor Agent

Observes systems, data, or processes and alerts on anomalies. Read-only by design.

Example: An agent that watches transaction volumes and flags unusual patterns before they become problems.

Layer 2

Commercial Brand — What Customers Buy

We market agents under the Automata (Ax) brand — workshop and industrial terminology that is precise, non-anthropomorphic, and honest about what the technology delivers.

Brand Category Engineering Equivalent What It Does
Directorate Orchestration Agent Strategic coordination. Sets direction, monitors operations, reports.
Bureau Task Agent / Process Agent Back-office operations umbrella. Contains specific Desks.
├── Collections Desk Task Agent (invoicing) Payment chasing, reconciliation, reminders.
├── Schedule Desk Task Agent (scheduling) Appointment booking, calendar management.
├── Procurement Desk Process Agent (procure-to-pay) Vendor selection, PO creation, delivery tracking.
└── Resolution Desk Task Agent (support) Customer inquiries, escalation, resolution.
Laboratory Environment (not an agent type) Research, training, validation before deployment.
Factory Environment (not an agent type) Build and deploy pipelines.
Assembly Orchestration Agent (multi-agent) Coordinate multiple Automata across end-to-end workflows.
Layer 3

Delivery Form — How It's Packaged

Offering Engineering Basis Who It's For
Single Desk Automata One Task Agent, one workflow SMEs with a single pain point
Multi-Desk Automata 2–3 Task/Process Agents, shared Bureau Growing businesses, 2-3 workflows
Full Assembly Multiple agents + Orchestration Agent Enterprise, multi-process automation
Hermes Agent (Ha) Full agent stack, self-hosted Organizations running their own agent infrastructure
Why This Vocabulary

Why We Don't Say "Digital Employee"

We do not use "Digital Employee," "AI Colleague," or "Digital Staff." These terms are imprecise and, per the BCG Henderson Institute, actively harmful — they reduce human accountability, lower review quality, and over-promise what the technology delivers.

Instead, we use:

  • Workshop language (Bureau, Desk, Factory, Assembly) — implies machines and process, not people
  • Capability-prefixed engineering types (Task Agent, Process Agent, Orchestration Agent) — says what it does, not what it pretends to be
  • Element symbols (Ax for Automata) — fits the Periodic Table of Elements brand without collision

When you see "Automata" on this site, read: an AI Agent built on open infrastructure, self-hosted in Jamaica, with a defined task scope and human verification gates.