2 ready-to-build workflows

AI agent workflows for Retail & e-commerce

Order tracking, cart recovery and support on the channels customers already use.

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. Self-serve order tracking on WhatsApp

Most support tickets are just 'where is my order?' — cheap to answer, expensive at volume.

Trigger: WhatsApp (inbound message)

How the workflow runs

  1. Trigger

    Customer messages. A WhatsApp trigger starts on any inbound message.

  2. Agent

    Identify the order. An Agent node extracts the order id / phone and confirms identity.

  3. Tool (MCP)

    Look up live status. A Tool (MCP) node fetches real status from your OMS/carrier.

  4. Output

    Reply with the answer. An Output node sends the status and expected delivery.

Channels & connectors

  • WhatsApp
  • Tool (MCP → OMS/carrier)

Outcome

The commonest question answers itself, instantly, in any language.

Why it helps

Removes the highest-volume, lowest-value ticket type from the human queue.

Build spec

1 agent — 1 order-status agent; the OMS lookup is a workflow Tool node.

System prompt (paste-ready)

You are an order-status assistant for {{store}}. From an inbound WhatsApp message, extract the order id (or phone/email) and verify it belongs to the sender before revealing anything. Call the order-lookup tool, then reply in the customer's language with the current status and expected delivery date, exactly as returned. Never guess a status or date — if the lookup fails or the order can't be matched to the sender, say so and offer a human.

MCP connectors

  • OMS / carrier — via a Tool (MCP) node (no first-party connector)
  • WhatsApp — trigger + reply channel

Built-in tools

  • http_request (order/tracking lookup)

Guardrails

  • Identity check before disclosing any order detail — the order must match the sender
  • Status and dates are read from the OMS only, never inferred or estimated
  • On a lookup failure, defer to a human rather than guess

Output & delivery

The agent reads live status via the OMS/carrier Tool node and replies on WhatsApp with the status and expected delivery, in the customer's language, within seconds.

2. Abandoned-cart recovery with a real conversation

Abandoned carts get a generic email that gets ignored.

Trigger: Webhook (cart-abandoned event)

How the workflow runs

  1. Trigger

    Cart abandoned. A Webhook trigger fires after the inactivity window.

  2. Agent

    Personalize the nudge. An Agent node references the exact items and any objection it can infer.

  3. Condition

    Respect frequency caps. A Condition node ensures you don't over-message the same customer.

  4. Output

    WhatsApp nudge + offer. An Output node sends a helpful, on-brand message with a one-tap return link.

Channels & connectors

  • Webhook
  • WhatsApp

Outcome

Carts are recovered through a real conversation the customer can reply to, not a dead-end email.

Why it helps

Meets shoppers on the channel they read, with context — the mechanism that lifts recovery over generic email.

Build spec

1 agent — 1 recovery agent; the frequency-cap check is a workflow Condition node.

System prompt (paste-ready)

You are a cart-recovery agent for {{store}}. Given an abandoned cart {customer, items[], cart_value, last_seen}, write ONE helpful, on-brand WhatsApp nudge that names the exact items left behind and addresses a likely objection (sizing, shipping, price). Include a one-tap return link. Do not invent a discount or offer unless one is supplied in the input. Keep it friendly and brief, in the customer's language.

MCP connectors

  • WhatsApp — send channel
  • Store cart events — inbound via Webhook trigger

Built-in tools

  • knowledge_search (active promotions / shipping policy)

Guardrails

  • Frequency cap enforced at the Condition node — a customer is never over-messaged on the same cart
  • Discounts/offers only when supplied in the input — the agent never fabricates a promo
  • One nudge per abandonment window, on the channel the customer opted into

Output & delivery

After the Condition node clears the frequency cap, the agent's personalized nudge is sent on WhatsApp with a one-tap link back to the exact cart.

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