---
title: "A Singapore AI studio cut vessel-repair quoting from 2 to 4 hours down to about 8 minutes with a Claude agent"
person: "Leon Lee (Connective Labs)"
profile: "https://dev.to/leon_from_connective_labs"
tools:
  - "Claude"
  - "LangChain"
  - "Supabase"
  - "Next.js"
  - "Vercel"
  - "Groq"
  - "OpenAI"
country: "Singapore"
category: "Get tasks done"
category_slug: get-tasks-done
proof:
  - label: "Read his post on DEV"
    url: "https://dev.to/leon_from_connective_labs/what-we-mean-by-ai-worker-and-how-we-actually-build-them-5ad9"
date: 2026-11-07
date_posted: 2026-05-14
as_claimed: "The result is shown as claimed in the public source linked here. It has not been checked and is not a typical outcome."
url: "https://agentstories.org/stories/leon-lee-connective-labs-maritime-quote-agent.html"
markdown_url: "https://agentstories.org/stories/leon-lee-connective-labs-maritime-quote-agent.md"
site: "Agent Stories by Solopreneur Code"
---

# A Singapore AI studio cut vessel-repair quoting from 2 to 4 hours down to about 8 minutes with a Claude agent

Leon Lee (Connective Labs) · Singapore · Claude, LangChain, Supabase, Next.js, Vercel, Groq, OpenAI

## What the agent did

Leon’s studio built an AI Worker for a maritime services client. It watches the inbox for job specs, extracts line items with Claude, matches them to a rate card in Supabase, drafts a quotation, flags unsure items and waits for human approval.

## Result, as claimed

For a maritime services client, quoting vessel repairs used to take 2 to 4 hours while an ops manager checked 200+ line items against a rate card in Excel. The AI Worker Connective Labs built watches the inbox, extracts line items with Claude, matches them to the rate card in Supabase, drafts a quotation, flags uncertain items, and sends the draft for one-click approval. Leon says the draft cycle is 8 minutes including human review.

> The result is shown as claimed in the public source linked here. It has not been checked and is not a typical outcome.

## Proof

- [Read his post on DEV](https://dev.to/leon_from_connective_labs/what-we-mean-by-ai-worker-and-how-we-actually-build-them-5ad9)

## Try it yourself

Turn one job-spec email and your rate card into a draft quote with uncertain lines flagged for you.

Prompt:

```text
You help me quote a service job. Here is the customer job spec: [paste email or PDF text]. Here is my rate card: [paste line items and prices]. Extract every line the customer asked for. Match each line to the closest rate-card item. If the match is weak, mark it [unsure] and say why. Build a draft quote with item, matched rate, quantity if given, and a total. List the unsure lines at the top for my review. Do not invent prices. Do not send anything.
```

Tools:

- [Claude](https://claude.ai): Named for extracting line items from messy PDFs and free text
- [Supabase](https://supabase.com): Named as the rate-card and data store in the write-up

Steps:

1. Paste one real past job spec and your current rate card into the prompt.
2. Check every [unsure] line yourself before you trust the total.
3. When the draft looks right, decide whether to connect inbox watch later.
4. Keep a human approval step before any quote reaches a customer.

Watch out: A fuzzy match can pick the wrong rate. You own every price that goes out.

## More

- [This story as a web page](https://agentstories.org/stories/leon-lee-connective-labs-maritime-quote-agent.html)
- [More stories: Get tasks done](https://agentstories.org/categories/get-tasks-done.html)
- [More stories from Singapore](https://agentstories.org/countries/singapore.html)
- [More stories that use Claude](https://agentstories.org/tools/claude.html)
- [More stories that use Supabase](https://agentstories.org/tools/supabase.html)
- [More stories that use Next.js](https://agentstories.org/tools/next-js.html)
- [More stories that use Vercel](https://agentstories.org/tools/vercel.html)
- [Every story, indexed for AI assistants](https://agentstories.org/llms.txt)
