QuotePilotAI

• 

  ArhitectureDiagram

Inspiration

Small service businesses lose revenue every day because customer inquiries arrive through messy emails, forms, call notes, and direct messages. Owners often do not have enough time to qualify every lead, prepare a quote, write a professional response, update a CRM, and remember the follow-up context.

QuotePilot AI was created to show how a Qwen-powered Autopilot Agent can automate the lead-to-quote workflow while keeping a human in control before any client communication is finalized.

What it does

QuotePilot AI turns an unstructured customer inquiry into a complete business workflow.

A user can paste a messy inquiry, and the system automatically:

Extracts structured lead information. Identifies the customer’s pain points, urgency, location, and service request. Recalls business memory such as preferred pricing rules, target customer type, tone, and service packages. Recommends a suitable service package. Drafts a quote with assumptions and deliverables. Reviews the output for risk, missing information, and overpromising. Creates an approval-ready client email. Generates a 3-step follow-up sequence. Saves a CRM-style lead record after approval. Displays memory updates and a transparent agent audit trace.

QuotePilot AI does not blindly send quotes or client emails. It prepares business actions autonomously, but requires human approval before anything is finalized.

How we built it

QuotePilot AI was built as a production-style hackathon MVP using:

Qwen Cloud Alibaba Cloud Model Studio DashScope OpenAI-compatible API Next.js TypeScript Tailwind CSS Node.js API routes Local memory storage for MVP Human-in-the-loop approval workflow Agent trace and audit log system

The backend calls Qwen through the OpenAI-compatible API base URL:

https://dashscope-intl.aliyuncs.com/compatible-mode/v1

The system uses a multi-agent workflow with specialized agents:

Intake Agent Memory Agent Quote Agent Risk Review Agent Follow-up Agent Audit Agent

Each agent contributes to a specific part of the workflow, and the final output is returned as structured JSON for the dashboard.

Qwen Cloud usage

QuotePilot AI uses Qwen Cloud as the reasoning layer for the lead-to-quote workflow.

The Qwen model is responsible for:

Understanding ambiguous customer inquiries Extracting structured business data Recalling relevant memory Generating quote recommendations Drafting client-ready communication Reviewing risks and missing information Producing an agent trace for explainability

The project uses the OpenAI-compatible DashScope endpoint so the app can call Qwen models through a familiar SDK structure while still running on Qwen Cloud infrastructure.

Agent workflow

QuotePilot AI uses six coordinated agents:

1. Intake Agent

Reads the messy inquiry and extracts customer details, company information, location, requested service, urgency, budget clues, pain points, and missing information.

1. Memory Agent

Retrieves business context such as the user’s preferred target customers, service packages, pricing assumptions, tone, and rules about human approval.

1. Quote Agent

Recommends the best-fit service package, price range, scope of work, deliverables, timeline, and assumptions.

1. Risk Review Agent

Checks for missing information, vague assumptions, overpromising risk, and whether human approval is required.

1. Follow-up Agent

Creates a professional client email and a 3-step follow-up sequence.

1. Audit Agent

Records what each agent did and why, creating a transparent trace for review.

Human-in-the-loop design

QuotePilot AI is designed for real business use, where fully autonomous communication can be risky.

The system requires human approval before:

A quote is finalized. A client email is considered ready. A CRM record is marked approved. Follow-up actions are accepted.

This makes the system safer, more realistic, and more suitable for service businesses that need speed without losing control.

Memory design

QuotePilot AI includes persistent memory concepts for:

Business name Preferred customer type Pricing rules Service packages Communication tone Human approval requirements Past lead context Follow-up preferences

The Memory Agent uses this context to make each quote more aligned with the business owner’s preferences.

Why it matters

Small businesses often lose leads because they are slow to respond, inconsistent with quotes, or too busy to follow up. QuotePilot AI helps them move faster without replacing human judgment.

The product can help:

Local service businesses respond faster. Solo operators manage more leads. Agencies prepare quotes more consistently. Sales teams reduce manual admin work. Founders turn customer demand into approved revenue actions.

Challenges we ran into

The biggest challenge was making the project feel like a real workflow instead of a simple chatbot. We focused on structured outputs, agent roles, approval checkpoints, auditability, and memory.

Another challenge was keeping the workflow simple enough for a hackathon demo while still showing production-grade thinking. We solved this by building a focused lead-to-quote use case instead of trying to automate every business process.

Accomplishments that we're proud of

We are proud that QuotePilot AI demonstrates:

A real business workflow from inquiry to quote. Qwen-powered reasoning. Multi-agent task division. Persistent memory concepts. Human approval controls. Audit logs and explainability. A clean dashboard for judges to test. A project that could become a real product after the hackathon.

What we learned

We learned that useful AI agents need more than a model call. They need workflow design, memory, state, approval logic, structured outputs, and clear user interfaces.

We also learned that Qwen Cloud’s OpenAI-compatible API makes it fast to build with familiar developer patterns while still using Qwen models as the intelligence layer.

What's next for QuotePilotAI

Next, we want to add:

Email inbox integration Calendar booking CRM integrations Call transcript ingestion Industry-specific quote templates PDF quote generation SMS follow-ups Analytics for lead response time Revenue recovery dashboard Multi-user team accounts

The long-term vision is to turn QuotePilot AI into an AI operations assistant for small service businesses.

Built With

• alibaba
• api
• cloud
• css
• dashscope
• endpoint
• human-in-the-loop
• json-based
• memory
• model
• next.js
• node.js
• openai-compatible
• qwen
• store
• studio
• tailwind
• typescript
• workflow