User Manual · v1.0

Aizen Orchestrate

The AI-powered platform for building, managing, and deploying intelligent workflow agents.

Introduction

Welcome to Aizen Orchestrate — the AI-powered platform for building, managing, and deploying intelligent workflow agents. This manual walks you through every step of the workflow creation process, from the Workflows dashboard through to production deployment.

Whether you are creating a workflow using the SOP builder, importing an existing procedure document, or testing and deploying a generated workflow, this guide has you covered.

Who This Manual Is For

New users getting started with the Aizen Agent Studio
Team members responsible for creating and managing AI-powered workflows
Administrators deploying workflows to production environments

What You Will Accomplish

Navigate the Workflows dashboard and understand your workflow inventory
Create a new workflow using the Upload SOP method
Complete all steps of the SOP Configuration wizard
Review and generate your workflow from an enriched SOP document
Test your workflow in the Playground Console
Deploy your workflow to a live production environment

Agents

The Agents section is where you build, manage, and deploy the AI agents that power your workflows. Agents are the intelligent actors within your workflows — each one is configured with a specific role, personality, system prompt, and model, and assigned to workflows to handle tasks such as decision-making, conversation, or data analysis.

Agents Dashboard

The Agents dashboard gives you an at-a-glance overview of your agent inventory and serves as the central hub for all agent management.

Column	Description
Agent Name	The agent's unique identifier.
Type	The agent's functional category (e.g., `decision_maker`, `conversational_ensemble`, `content_analyzer`).
Description	A plain-language summary of what the agent does.
Project	The project this agent belongs to.
Owner	The user who owns this agent. Displays "You" if you are the owner.
Date Created	When the agent was first created.
Last Modified	How recently the agent was changed.

Create Agent

Click the + Create Agent button in the top-right corner. A modal will appear with three creation methods:

Method	Description
New Agent	Start from scratch using the Agent Builder — the AI-guided chat interface will walk you through defining all agent fields step by step.
Duplicate Agent	Copy an existing agent's full configuration as a starting point. Useful for creating variants of a proven agent.
Upload JSON	Import an agent from a previously exported JSON file. Used for migrating agents between environments or restoring from backup.

Agent Builder

Selecting New Agent opens the Agent Builder — a split-screen interface with a conversational AI interface on the left and live JSON output on the right.

Left Panel — Chat Interface

The AI guides you through building the agent step by step. Type your instructions and press Send. Additional controls:

Workflows — the workflow name. Click to open and edit.
Refresh Summary — reloads the agent summary based on current inputs
Clear Messages — clears the chat history for a fresh session
Close — exits the Agent Builder

Right Panel — Agent Configuration

Tab	What it contains
Overview	Agent Name, Description, and Type
Input/Output	Defines what data the agent accepts as input and what it returns as output
System Prompt	The core instruction set that governs the agent's behavior and decision-making
Model	The underlying AI model powering the agent
Personality	Tone, communication style, and persona settings for conversational agents

Agent Actions Menu

Click the ⋯ menu at the far right of any agent row to access these actions:

Action	Description
Test	Opens a test console to run the agent in isolation and validate its behavior before assigning it to a workflow.
Export	Download the full agent configuration as a JSON file for backup or migration.
Duplicate	Creates a copy of the agent with all its configurations intact.
Delete	Permanently removes the agent. This action cannot be undone.

Workflows Dashboard

The Workflows dashboard is your central hub for all workflow management. It provides a high-level overview of your workflow inventory and serves as the starting point for creating new workflows.

Summary Cards

Card	Description
Total Projects	The number of projects in your workspace that contain workflows.
Total Workflows	Total count of all workflows across all projects.
Workflows You Own	Workflows where you are the assigned owner.
Modified Today	Workflows edited or updated today.

Your Workflows Table

Workflows — the workflow name. Click to open and edit.
Project — the project folder this workflow belongs to.
Owner — the user who owns this workflow.
Date Created — when the workflow was first created.
Last Modified — how recently the workflow was changed.
Status — GENERATED (built, not deployed) or DEPLOYED (live in production).
Next Action — a quick link such as Review, Playground, or Status.

Filtering and Searching

Use the All / Project / Owner filter buttons to narrow the workflow list.
Use the Search workflows field to find a workflow by name.
Click the sort arrows on any column header to sort ascending or descending.

TipClick any workflow name in the table to open the workflow editor and Playground Console directly.

Creating a New Workflow

Click the + Create Workflow button in the top-right corner of the Workflows dashboard. A modal will appear with the following creation methods:

Method	Description
Upload SOP	Import an existing Standard Operating Procedure document (.pdf, .docx, .txt, .md) to quickly generate a workflow. This is the recommended path for most users.
Duplicate Workflow	Copy an existing workflow configuration as a starting point. Ideal for creating variants of proven workflows.
Upload JSON	Import a workflow from a previously exported JSON file. Used for migrating or restoring workflows between environments.

SOP Configuration Wizard

The SOP Configuration Wizard is a guided multi-step process that walks you through every aspect of setting up your workflow. The left panel always shows your current progress, with a progress bar at the bottom.

Step 1 of 7 — Session Settings

After selecting Upload SOP, the wizard opens Step 1: Session Settings. This step establishes the identity of your new workflow configuration.

Field	Description
Configuration Type	Select from the dropdown. Choose 'Create New Configuration' to start fresh or select an existing configuration to reuse its base settings.
Configuration ID	Enter a unique alphanumeric ID (e.g., `W001` or `sales_onboarding_v1`). This ID cannot be changed after creation.

TipUse a consistent naming convention — for example: dept_process_version (e.g., sales_onboarding_v1). This makes configurations easy to identify across large workspaces.

Once the Configuration ID is filled in, click Next Step to advance to Step 2.

Step 2 of 7 — Upload SOP

Step 2 is where you provide the Standard Operating Procedure document. The platform analyzes this document and uses it as the foundation for generating your workflow.

Supported File Formats

.pdf — PDF documents
.docx — Microsoft Word documents
.txt — Plain text files
.md — Markdown files

NoteOnly one SOP file is allowed per configuration. Uploading a new file will replace the existing one. Ensure your SOP is complete before proceeding.

Step 3 of 7 — Optional Integrations

Step 3 allows you to enhance your workflow with optional tool integrations. All three types are optional.

Type	Description
OpenAPI Specification	Upload one or more OpenAPI Spec files (.json, .yaml, .yml). Defines API endpoints and schemas the workflow can interact with.
MCP Configuration	Upload a Model Context Protocol (MCP) config file. Enables structured AI interactions and context passing between workflow steps.
Webhook Configuration	Upload a Webhook config file. Allows your workflow to trigger external services or receive real-time event notifications.

NoteThese integrations are optional. You can skip this step and add integrations after your workflow has been generated.

Step 4 of 7 — Select Your Agent

Step 4 is where you assign AI agents to execute and manage your workflow. Agents handle decision-making, conversation, task orchestration, and tool execution.

Selecting Agents

Click an agent in the Available Agents list to select it — it moves to the Selected Agents section.
To remove an agent, click the X next to its name in Selected Agents.
Use the Search agents field to filter by name.

Agent Types

Type	Description
Decision Maker	Analyzes data and makes routing or classification decisions at key workflow branch points.
Conversational	Transforms SOPs into autonomous agents that guide customers through complex workflows across voice, SMS, WhatsApp, and webchat.
Judge	Evaluates content against multiple criteria with weighted scores and pass/fail determinations.
Guardrail	Enforces safety and policy through multi-layer validation — PII detection, content filtering, LLM moderation, and custom checks.
Content Analyzer	Processes multi-modal content through configurable pipelines. Images, documents, audio — extract intelligence from any format.
Content Generator	Creates text, images, audio, and documents using LLM and specialized models.

Use Third Party Agent

Check the 'Use Third Party Agent' box to integrate an external AI agent not hosted on the Aizen platform.

Step 5 of 7 — Add Predictive Models

Step 5 allows you to attach one or more predictive ML models to your workflow, enabling agents to call trained models as part of their decision-making logic.

Column	Description
Model Name	The unique identifier for the model.
Project	The project the model belongs to.
Model	The underlying model object being referenced.
Version	The specific version of the model to use.

TipClicking 'Analyze SOP' adds an additional Step 6: Analyze & Resolve to the wizard, letting you review and fix issues the AI identifies before final workflow generation.

Step 6 of 7 — Analyze & Resolve (Optional)

If you clicked Analyze SOP, the wizard adds this step. The AI scans your SOP for issues — missing information, ambiguous steps, undefined actors, or incomplete decision logic — and proposes fixes.

What Happens in This Step

The AI analyzes your SOP for gaps, ambiguities, and unresolved references.
Identified issues are displayed with proposed resolution suggestions.
You can review each issue, accept or modify the AI's suggestion, and mark it resolved.
Once all issues are resolved, the platform generates an Enriched SOP.

NoteResolving all issues in this step produces a higher-quality generated workflow. We recommend addressing every identified issue before proceeding.

Step 7 of 7 — Review & Generate Workflow

The final step presents the fully Enriched SOP in an editable text view. Review carefully before generating.

What to Review

The scope and purpose sections — confirm they reflect your intended workflow.
Actor definitions — verify all roles (agents, systems, tools) are correctly described.
Process steps — check the sequence and logic of all operations.
Decision points and conditions — ensure all branching logic is correct.

Editing and Saving

Click Edit to enter edit mode and make direct changes to the SOP text.
Click Save to File to download a copy of the Enriched SOP for your records.

Once satisfied, click Generate Workflow. The platform processes the SOP and automatically creates your workflow. You are then redirected to the Human-in-the-Loop Review screen.

Reviewing the Generated Workflow

After generation, you arrive at the Human-in-the-Loop Review screen — an interactive AI-powered chat interface that lets you inspect, modify, and validate every aspect of your workflow before it goes to production.

Think of it as having a conversation with your workflow before you ship it.

Interface & Navigation

Top Navigation Bar

Back — returns you to the Workflows dashboard without saving changes
Unified Workflow Graph — opens the complete multi-agent visual diagram
Open Playground — launches the Playground Console
Open Table Reader — opens a structured tabular view of the workflow data

Workflow Header Panel

Review Features checkbox — enables step-level feature annotation and review mode
Save & Exit — saves any changes and returns you to the Workflows dashboard

Chat Canvas & Input

The large center area is your conversation space. A confirmation message appears on session creation showing step and group counts. Type instructions in the input bar and press Enter to send.

TipUse the chat input to ask the AI to make changes — for example: 'Add a step, show diagram'.

Playground Console

The Playground Console is your testing environment for running, debugging, and validating your workflow before deploying it to production.

Left Panel — Available Sequences

Column	Description
ID	The unique test sequence identifier.
Type	Scenario type — `happy_path` (normal flow) or `error_path` (failure scenarios).
Len	Number of steps pre-loaded in the sequence payload.
Actions	Open (inspect the sequence) or Run (execute immediately).

Filtering and Configuration

Filter — toggle between All, Happy, or Error test sequences.
Delay (ms) — delay between step executions (default: 500ms).
Preserve Session — optionally enter a session_id to persist state across runs.

Running Tests

Check the checkbox next to one or more sequences to run.
Click Run Selected to execute all checked sequences.
Monitor the Workflow Debugger panel on the right for real-time results.

Workflow Debugger Panel

Metric	Description
Execution Time	Total time to complete all steps (e.g., 9.0s).
Steps Executed	Number of workflow steps that ran.
Activities Completed	Successful activities within those steps.
Failed	Count of steps or activities that encountered errors.
Steps Executed List	Chronological list of every step that ran.

Debugger Quick Actions

Show Summary — display the full execution summary.
Show Errors — highlight only the steps that failed.
Timeline — view a time-based execution trace.
Slow Steps — identify bottleneck steps with high execution times.
Final Context — inspect the full context object at end of workflow.

TipAlways run both happy_path and error_path test sequences before deploying. This ensures your workflow handles both normal operations and edge case failures correctly.

AI Debugger Chat

The chat input lets you query the AI about the execution. Examples: "Why did process_payment fail?" or "Show me the timeline for the last run".

Deploying Your Workflow

Once testing is complete, click the Start button (top-right of the Playground Console) to deploy your workflow to a production environment.

Start Workflow Modal

Field	Description
Version Description	A description of this version being deployed (e.g., 'Initial production release v1.0').
Deployment	Select an existing deployment or click '+ Create Deployment' to create a new one.
OpenAPI Servers	Configure API server endpoints the workflow will call. Each requires a Base URL and Timeout.
Webhooks	For webhook integrations, enter the Endpoint URL that will receive webhook events.
Export Configuration	Export the full deployment config as YAML or JSON via Preview and Copy.

Creating a New Deployment

Field	Description
Deployment Name *	Lowercase name using letters, numbers, hyphens, and dots only (e.g., `my-deployment`). Required.
Description	Optional description of this deployment environment.
CPU Count *	CPU cores to allocate (default: 1). Increase for high-throughput workflows.
Memory Limit (MB) *	RAM allocated to this deployment (default: 4096 MB).

Click Create to provision the environment, then return to the Start Workflow modal to complete configuration.

Finalizing Deployment

Once all required fields (marked *) are completed, click Start. The workflow is versioned and deployed to the selected environment.

NoteAfter successful deployment, the workflow's status updates from GENERATED to DEPLOYED, and the Next Action column shows 'Status' instead of 'Playground'.

TipUse a versioning convention for deployment names such as workflow-name-v1 to make it easy to manage multiple versions of the same workflow over time.

Managing Your Workflows

After deployment, monitor and manage all workflows from the main Workflows dashboard. The Status column clearly differentiates generated from deployed workflows.

Workflow Status Values

Status	Description
`GENERATED`	The workflow has been created and is available for Playground testing but has not yet been deployed to production.
`DEPLOYED`	The workflow is live in a production deployment and actively accessible via its API endpoints.

Managing Workflows

Click a workflow name to open the editor and Human-in-the-Loop review screen.
Use the '...' menu at the far right of any row for options such as Edit, Duplicate, or Delete.
Click 'Playground' for any GENERATED workflow to open the test console.

TipRegularly review your Workflows dashboard to identify stale or unused workflows. Keeping the workspace clean makes it easier to find and maintain active workflows.