2 ready-to-build workflows

AI agent workflows for Manufacturing & field service

Turn machine signals and complaints into dispatched work — before things break.

Each recipe below is expressed only in VegaDūta's real workflow building blocks — Trigger, Agent, Knowledge, Condition, Approval, Output, Tool (MCP), Voice, Device, Loop, Code and Parallel — so it maps 1:1 to something you can assemble in the Workflow designer. Consequential actions always pass a human Approval step.

1. Sensor alert → maintenance work order

A machine's readings drift out of range; today an operator notices late (or after a breakdown), then phones maintenance and fills a work order by hand.

Trigger: Webhook (IoT/SCADA sensor threshold)

How the workflow runs

  1. Trigger

    Sensor crosses a threshold. A Webhook trigger receives the telemetry event from your IoT/SCADA layer.

  2. Condition

    Confirm it's real. A Condition node filters noise (sustained vs momentary) so you don't cry wolf.

  3. Agent

    Diagnose + prioritize. An Agent node correlates the reading with the asset history and sets urgency.

  4. Device

    Safe-state the equipment. A Device Control node can trigger a safe stop / mode change where that's wired up.

  5. Output

    Dispatch the technician. An Output node raises the work order and messages the on-shift technician with the fault and part.

Channels & connectors

  • Webhook (IoT/SCADA)
  • Device Control
  • WhatsApp
  • Tool (MCP → CMMS)

Outcome

Drifting equipment is caught and dispatched automatically, often before failure.

Why it helps

Replaces late human noticing + manual work-order entry with an instant, prioritized dispatch — the core of predictive maintenance.

Build spec

1 agent — 1 diagnosis agent; noise filtering is a Condition node and the safe-stop is a Device Control node.

System prompt (paste-ready)

You are a maintenance-diagnosis agent. Given a confirmed sensor event {asset_id, metric, value, threshold, duration} and the asset's history, assess the likely fault, set urgency (low/medium/high/critical), and identify the probable part or check needed. Recommend whether the equipment should be safe-stopped. Base the diagnosis only on the telemetry and maintenance history provided — never fabricate a fault code or part number. Output {asset_id, likely_fault, urgency, recommended_action, part?}.

MCP connectors

  • IoT/SCADA telemetry — inbound via Webhook trigger
  • CMMS (work orders) — via a Tool (MCP) node
  • WhatsApp — technician channel

Built-in tools

  • knowledge_search (asset manuals + maintenance history)

Guardrails

  • The Condition node filters momentary noise — only sustained, confirmed threshold breaches reach the agent
  • Safe-stop via the Device Control node only where that interlock is explicitly wired — otherwise it advises, never forces
  • Diagnoses use real telemetry + history — no fabricated fault codes or parts

Output & delivery

On a confirmed breach, the agent diagnoses and prioritizes, a Device Control node can safe-state the equipment where wired, and an Output node raises the work order and messages the on-shift technician with the fault and part.

2. Field quality complaint → RCA intake

Customer quality complaints arrive by call/email and are logged inconsistently, delaying root-cause analysis.

Trigger: WhatsApp / Email / Call

How the workflow runs

  1. Trigger

    Complaint received. A WhatsApp/Email/Call trigger captures it in the customer's words.

  2. Agent

    Structure the report. An Agent node extracts product, batch, defect and severity into a standard RCA intake.

  3. Tool (MCP)

    Log to quality system. A Tool (MCP) node files it in your QMS and links the batch.

  4. Approval

    Quality engineer triages. An Approval node routes to the right engineer for disposition.

Channels & connectors

  • WhatsApp
  • Email
  • Voice / Call
  • Tool (MCP → QMS)
  • Approval / HITL

Outcome

Every complaint becomes a structured, batch-linked record the moment it arrives.

Why it helps

Removes inconsistent manual logging that slows recalls and root-cause analysis.

Build spec

1 agent — 1 intake agent; the QMS write is a Tool node and the triage is an Approval node.

System prompt (paste-ready)

You are a quality-complaint intake agent. From a customer complaint (voice or text, any language), extract a standard RCA intake {product, batch_or_lot, defect_description, severity, date_noticed, customer, quantity_affected}. Capture the customer's own description faithfully — don't diagnose the root cause yourself or assign blame. If the batch/lot is missing, ask for it. Flag safety-critical defects as high severity for immediate routing.

MCP connectors

  • QMS (quality management system) — via a Tool (MCP) node
  • WhatsApp / Email / Voice — intake channel
  • Sarvam — Indian-language support

Built-in tools

  • knowledge_search (product/batch catalog for validation)

Guardrails

  • The agent captures and structures — root-cause and disposition are decided by the quality engineer at the Approval node
  • Safety-critical defects are flagged high-severity for immediate human routing
  • Every complaint is logged to the QMS and linked to its batch/lot before triage

Output & delivery

The agent structures the complaint into a standard RCA record, a Tool node files it in the QMS linked to the batch, and the Approval node routes it to the right quality engineer for disposition.

Related industries

Build one of these in minutes

The sandbox gives you a live agent workspace — no account, no card. Or head back to the full catalogue and compare patterns across every industry.

Try this in the sandbox — no signupBrowse all industries
See how VegaDūta compares to n8n, Dify and BotpressWhy VegaDūta's architectureEstimate your WhatsApp API costs