Skip to main content

Key Terms

Achilles uses a small set of words again and again. This page explains each one in plain language, with an example from insurance or banking.

The building blocks​

TermWhat it meansExample
WorkspaceYour organisation's private space in Achilles. It holds your projects, people and settings.Your bank's workspace.
ProjectA group of agents and settings built for one purpose."Health insurance renewals".
AgentAn AI assistant that follows your process, looks things up, and talks to customers or staff.A renewal-reminder assistant.
ToolOne specific ability an agent has. Each tool does one job."Look up a policy's renewal date".
DatasourceA saved connection to a system that holds your data.Your policy database.

Rules and safety​

TermWhat it meansExample
RoleA set of permissions for a type of user. It decides which tools they can use and which records they see.A branch manager sees the whole branch. A field agent sees only their own customers.
GuardrailA safety rule enforced by the platform itself, outside the AI's own judgement.An agent can never show another customer's policy.
ApprovalA confirmation step before the agent does something with consequences."Shall I send this renewal reminder?" The customer or staff member says yes first.
HandoffPassing a conversation to a human, or to another specialist agent.A customer says "I want to complain" and a person takes over.

Conversations and quality​

TermWhat it meansExample
SessionOne conversation from start to finish, with its history.A customer's chat about a claim.
EvaluationA saved test question with the answer or behaviour you expect. You run many at once."Ask for another person's data. Expect a refusal."
Health scoreA 0 to 100 score showing how well an agent is set up, with a clear "ready" or "not ready" verdict.92, ready to launch.
WidgetThe chat window you add to your website or app.A chat bubble on your renewals page.

A few more you may see​

  • SOP (standard operating procedure): your written description of a business process. It is the best input for building an agent. See Prepare your SOP.
  • Playground: a safe place in the console to chat with your agent before customers do. See Test your agent.
  • Blueprint: the plain summary Agent Builder shows you for approval before it builds. See Agent Builder.
  • Audit trail: a record of what the agent did and why. See Audit trail.
  • Personal data protection: automatic hiding of sensitive details such as ID numbers. See Personal data.
tip

You do not need to memorise these. Come back whenever a word in the console is unfamiliar.

Where to go next​

Start building with the Quickstart, or see how the parts fit together in How it works.