Guide
Dataworkz Agents are LLM-powered, multi-step systems that combine reasoning, retrieval, and execution to automate business processes.
Dataworkz Agents plan actions, call Tools, and produce context-aware, personalized outputs. Unlike a conventional RAG pipeline that only retrieves relevant documents, an Agent orchestrates multiple steps across Tools and data sources.
Agents are best for business processes that require both knowledge retrieval and action — including customer support assistants, internal knowledge bots, and automated Workflow Agents.
💡 Note: Agents are designed so business users and developers can work together. You do not need deep ML expertise to create effective Agents. Clear personas, narrow Scenarios, and well-defined Tools go a long way.
Key Capabilities
Plan and orchestrate multi-step Workflows across Tools and systems.
Invoke external systems (databases, APIs, RAG Apps, Snowflake, MongoDB, and more).
Personalize responses using conversation context, variables, and metadata.
Handle failures gracefully with fallbacks, retries, and escalation paths.
Why Agents
A RAG-only approach answers questions by returning passages from documents. An Agent can act on those results: query a customer database, apply a policy, verify results, and respond with an action-oriented reply.
This translates to faster resolution, reduced manual work, consistent application of policies, and improved customer experience.
Example Use Cases
Enhanced Information Retrieval — Provide technicians or support staff with troubleshooting steps filtered by product model and firmware version.
Customer Chatbot — Answer order questions, validate returns eligibility using policy documents, and initiate refunds or escalation when needed.
Enterprise Knowledge Assistant — For HR and Finance teams: extract facts from policies, generate summaries, and link answers to official documentation.
How Agents Work
An Agent combines:
Reasoning LLM — builds a plan (sequence of steps).
Tools — actions the Agent can call (data queries, APIs, RAG lookups).
Scenarios — focused intents that restrict which Tools and behaviors apply.
A plan step invokes a Tool with parameters resolved from conversation context, environment, or prior step outputs. The Agent iterates until it can respond or determines a fallback path.
Example
AI: You have two orders — #232 (Yellow Shirt), #233 (White Socks).
User: What's the status of the order with the socks?
Agent plan:
FetchCustomerDetails(customer_id = current_user_id)FetchOrderDetails(customer_id, order_id = 233)FetchShippingDetails(tracking_no = $2.tracking_no)
Agent outcome: return shipping status or ask for missing inputs.
Accessing Agents
Navigate to AI Agents in the product menu. The primary UI layout is two-pane:
Left panel: Agent list (create / edit / run / delete).
Right panel: Tools Repository (browse, filter, create Tools).
Last updated