Customer
Glossary.
We build AI tools for your business. This glossary defines what we mean — no hype, no marketing fluff, just what they are.
A
Agent
See AI Agent.
Agent Harness
The runtime system that wraps an AI model with safety rails, tool access, and human checkpoints. The harness is where we enforce limits — not the model. Every Automata includes a harness.
AI Agent
An automated system that follows instructions, uses tools, and completes tasks within defined boundaries. It is not a person, not a chatbot, not magic. It is a machine built for work — composed of a model, a harness, tools, and a working environment. It handles tasks, not jobs.
Assembly
Multi-Automata orchestration. When you need several Automata working together across end-to-end processes. See also Directorate, Bureau.
Automata (Ax)
Our brand for autonomous AI agents. The name comes from "automaton" — a machine that follows a set of instructions. Not a person, not an employee. Automata handle invoicing, scheduling, procurement, and customer resolution. Self-hosted on Jamaican infrastructure.
B
Bureau
The back-office category of Automata. A Bureau contains specific Desks — the actual workflows. See Collections Desk, Schedule Desk, Procurement Desk, Resolution Desk.
Bootstrapped Agent
A pre-configured AI Agent for a common business function. Delivered in 2 weeks at a fixed price. Includes working environment, harness, and one trained task loop.
C
Collections Desk
An Automaton that handles invoicing, payment chasing, reminders, and reconciliation. Your collections process, automated end-to-end.
D
Desk
A specific workflow within a Bureau. Each Desk is one task-focused Automaton. Example: Collections Desk, Schedule Desk.
Directorate
The coordination Automaton. It doesn't do the work itself — it monitors your Bureau Desks, escalates exceptions, and reports to your team. Think of it as the supervisor that never sleeps.
F
Factory
Our build and deployment environment. Automata are constructed here, tested here, and shipped from here to your workspace.
G
Guided Execution
A workflow where the agent proposes a plan, a human approves it, the agent executes, and a human verifies the result. No autonomy ambiguity — you stay in control.
H
Harness
See Agent Harness.
Harness Engineering
The discipline of designing environments, constraints, and feedback loops that make AI agents reliable in production. Distinct from prompt engineering. This is what we do.
Hybrid Inference
Running AI models across both local (self-hosted) and cloud infrastructure, selected per-task based on sensitivity and speed requirements.
K
Knowledge Agent
An AI Agent specialized in finding and explaining information from a defined knowledge base. Read-only by default — it answers questions but does not take action.
L
Laboratory
Our research, training, and validation environment. Automata are tested here before they reach your Bureau.
M
Monitor Agent
An AI Agent that watches systems, data, or processes and alerts on anomalies. Watch only — it does not make changes.
O
Orchestration Agent
An agent that coordinates other agents. Delegates tasks, collects results, routes problems. Does not execute tasks directly.
P
Process Agent
An AI Agent that spans an end-to-end business process (e.g. procurement, order-to-cash). Maintains state across steps and hands off between sub-agents.
Procurement Desk
An Automaton that handles vendor selection, purchase order creation, and delivery tracking. End-to-end procurement automation.
R
Resolution Desk
An Automaton that handles customer inquiries, support tickets, and escalation. Keeps your customers informed while you handle the complex cases.
S
Schedule Desk
An Automaton that manages appointments, calendar booking, and reminders. No more back-and-forth emails to find a time.
Sovereign Deploy
AI infrastructure deployed on customer-owned hardware or dedicated tenancy in Jamaican data centres. Your data never leaves the country. This is our default deployment model, not a premium add-on.
Stack Ownership
Owning your full AI stack — inference, memory, tooling — rather than renting from a single vendor. Prevents lock-in and vendor dependency.
T
Task Agent
A focused AI Agent that executes one function well. The basic unit of deployment. A Task Agent handles a defined sequence of steps with tool access and human milestone check-ins.
Tooling Layer
The set of capabilities an agent can call: skills (procedures), plugins (third-party integrations), and MCP servers (external tools). This defines what the agent is allowed to do.
V
Verification Gate
A mandatory human review step before agent output becomes action. Built into every Automata. Not a limitation — a design principle. You see what the agent produced before it takes effect.
How We Use These Terms
- On this site — We use the Automata brand (Directorate, Bureau, Desk) because it's clear and honest about what you're buying.
- In our blogs — We use the engineering taxonomy (Task Agent, Process Agent, Orchestration Agent) when discussing architecture.
- In your contract — We use the delivery form (Single Desk, Multi-Desk, Assembly) because it maps to what you pay for.
They describe the same thing at different levels of detail. This glossary connects them.
Last updated: July 2026