Acudesk’s Workflow agents are AI-driven automation units that connect directly to your systems, not your users. They interpret incoming API requests, read and update data, call your endpoints, trigger workflows, and return clean, structured outputs — all without a single conversation.
These agents operate behind the scenes with speed and precision, making decisions, applying your business rules, and orchestrating multi-step processes exactly as you define them.
Unlike conversational agents, these agents work entirely behind the scenes, making decisions, transforming data, and orchestrating workflows with speed, accuracy, and consistency.
These agents are ideal for scenarios where accuracy, consistency, and automation matter more than user interaction.
Instead of relying on scripts or custom integrations, a workflow agent understands your data structures, follows your validation logic, and performs tasks consistently across your environment. They can call internal or external APIs, update records, read from datasets, return JSON outputs, and automate complex workflows end-to-end.
With Acudesk, workflow agents become the intelligent layer between your systems — reliable, secure, and fully aligned with your business logic — automating work that previously required manual effort or complicated engineering.
1. You define the instructions
Each agent has its own role, purpose, and behaviour. You describe what it does, how it should interpret inputs, and what outcomes it should produce.
2. You connect datasets and actions
The agent uses datasets to retrieve information, and actions to make API calls or update systems. These building blocks define what your agent knows and what it can do.
3. You define the output
Workflow agents always return structured JSON.
You can define custom output schemas, examples, validation rules, and error-handling instructions to ensure consistency for your consuming applications.
4. Your systems call the agent
Workflow agents receive data through API calls from your own software, tools, or workflows. They interpret the request, perform the required tasks, and respond with the correct structured output.
Workflow agents are designed for work that requires precision and predictable output. They interpret incoming requests, apply your business rules, and interact with your systems, making them ideal for automation tasks that demand consistency across departments and platforms.
Datasets are the foundation of every workflow agent. They give the agent access to the real information that drives your organisation — whether it’s live SQL data, API-powered metrics, internal documents, or contextual grounding entries. By connecting these sources, your agent can understand your business environment, retrieve accurate details, and make informed decisions without relying on generic AI assumptions.
With the right datasets in place, workflow agents become capable of answering system-level requests with precision. They can validate information, apply rules, enrich inputs, and return structured outputs that align perfectly with your internal systems. This ensures that every automation, transformation, or workflow orchestration is based on the most current and relevant data available.
Workflow agents rely on datasets for context:
- SQL datasets for live business data
- API datasets for real-time system metrics
- document collections for RAG-style information
- image/document processing tools
- grounding data for correct interpretation of entities
This allows workflow agents to answer internal API requests with real, accurate, up-to-date information pulled from your own systems.
Actions turn workflow agents into true automation engines. While datasets supply the information an agent needs, actions define what the agent can actually do — from submitting requests to updating records or triggering multi-step workflows. Each action encapsulates a specific system operation, ensuring the agent performs tasks exactly according to your rules and API structures.
Because actions reflect your own business logic, the agent can make decisions and execute processes consistently across all your systems. Instead of maintaining scripts or custom integrations, you simply configure actions once and let the agent determine when they should be used. This creates predictable, accurate, and repeatable behaviour, even in complex environments.
When a request comes in, the agent automatically selects the appropriate action, fills in the parameters, and executes it safely based on your access controls. This makes actions the core mechanism that enables workflow agents to automate work reliably, intelligently, and without unnecessary complexity.
An action includes:
- endpoint URL
- required and optional parameters
- expected response structure
- human-friendly description
- agent-specific instructions
- example payloads
- intent or goal
Just like conversational agents, workflow agents use grounding to correctly interpret the terminology and structures unique to your organisation.
This allows the agent to understand that a reference such as “clientId” points to a specific database column, that “vehicleType” maps to your internal fleet taxonomy, or that “Unit A” refers to a particular branch such as your Johannesburg location.
By linking incoming requests to the right business concepts, grounding ensures that the agent interprets internal references accurately and maintains consistent behaviour across all automated workflows.
Workflow agents follow the same strict security model used throughout Acudesk.
Each agent has its own set of access permissions, every dataset has defined access rules, and individual actions can be limited to specific users or clients.
When an agent receives a request, it will only perform the operation if the caller has the required permissions. This built-in enforcement guarantees that all automations respect your security boundaries, protecting sensitive information and ensuring compliance across your systems.
Workflow agents fit naturally into almost any part of your technology stack.
They act as an intelligent automation layer that receives requests, applies your business logic, and returns structured results, making it easy to extend your existing systems without rewriting them. Any environment that can send an API request can instantly benefit from a workflow agent.
Any system that can make an API request can use a workflow agent, making them a highly flexible and universal automation component within your organisation.
These examples demonstrate how workflow agents use datasets, grounding, and actions to process information and perform tasks reliably. From orchestrating workflows to making rule-based decisions, each agent follows a predictable, configurable sequence that aligns with your internal logic.
Maintenance workflow
The agent detects equipment damage
Evaluates severity
Calls your maintenance API
Returns a structured JSON report with recommended actions.
Cross-system workflow orchestration
The agent fetches data from datasets, enriches it, calls multiple downstream APIs in sequence, handles errors, and returns a consolidated structured result — all without you writing orchestration code.
Decision-making logic
The agent checks datasets, business rules, thresholds, and history
Returns a clear approval or decline instruction with reasoning
Service request routing with escalation
The agent classifies the issue using your grounding data
Determines whether the request must be routed to IT, operations, or finance
Checks workload or availability datasets to pick the right assignee
Triggers an action to create the ticket in the correct system
And returns a routed, priority-tagged, fully structured ticket summary.