Key Concepts

Overview

Before building your first Agent in Arahi AI, it’s important to understand a few key concepts.
These are the foundational building blocks that define how Agents think, act, and collaborate within the platform.

1. Agent

An Agent is an intelligent AI entity that can think, reason, and perform specific business tasks.
Each Agent is made up of:

A Prompt (the Agent’s core instructions and behaviour).
Tools (actions it can perform).
Triggers (events that start its work).
Memory (context and information it retains).
Variables (dynamic placeholders for flexible logic).
Human Approval (a human-in-the-loop safety step for oversight).

Agents can work individually or as part of larger workflows involving multiple Agents.

2. Task

A Task represents one instance of an Agent doing work — such as analysing a document, drafting an email, or updating a database.
Each task includes:

Input data.
The Agent’s reasoning and output.
Logs, status, and metrics.
Optional chat interaction and approval steps.

Tasks are how you monitor and manage Agent activity.

3. Tool

A Tool gives your Agent the ability to take real actions in the world.
Examples include:

Sending an email.
Searching a document database.
Running a custom API call.
Tools connect your Agents to your existing systems and data sources.

4. Trigger

A Trigger determines when an Agent starts a new task.
Triggers can be:

Manual: started directly by a user.
Automated: scheduled or event-based.
Integrated: connected through Slack, Gmail, CRMs, or webhooks.

5. Memory

Memory allows an Agent to store useful information from past interactions — improving its context awareness and consistency.
For example, an Agent can remember previous client details, preferences, or task outcomes to provide better responses next time.

6. Variables

Variables are reusable data placeholders (e.g., {{company_name}}, {{api_key}}).
They help you maintain consistency and update values globally without editing each prompt or tool manually.

7. Human Approval

Human Approval introduces oversight into your automation.
When an Agent encounters uncertainty or high-risk scenarios, it can pause and request review or approval before proceeding.
This ensures accuracy, compliance, and accountability in critical workflows.

8. Workflow

A Workflow is a chain of Agents working together to complete multi-step processes.
For example:

Lead Capture → Qualification → Follow-Up → Report Generation

Workflows can run automatically, triggered by external events, or manually launched by your team.

9. Analytics

Analytics in Arahi AI provide visibility into how your Agents are performing.
You can track metrics such as:

Task success rate.
Time saved per automation.
Escalation and approval frequency.
Credit and cost usage.

Analytics help you measure ROI and identify opportunities for optimisation.

Summary

Concept	Description	Example
Agent	The AI entity that performs tasks	“InvoiceProcessor Agent”
Task	A single execution of an Agent’s job	Processing one invoice
Tool	Actions the Agent can perform	“Send Email”, “Parse Document”
Trigger	Event that starts an Agent	“New email received”
Memory	Stores context from previous runs	Client preferences
Variables	Reusable placeholders	`{{company_name}}`
Human Approval	Human review checkpoints	“Flag for manager approval”
Workflow	Multi-step automation	“Lead → Email → Report”
Analytics	Track and measure performance	Success rate, cost per task

Understanding these key concepts will help you build, manage, and scale your Agents efficiently within Arahi AI.