Skip to content
Allan Henslee, PhDResume (PDF)

Case 03Premier Party Cruises · 2026

Custom quotes in minutes, not hours or days

A proposal builder, an instant quote page, a lead dashboard and one customer history for an Austin party-boat company, with prices set by people and a price sheet, never by the AI.

Year
2026
Client
Premier Party Cruises, through Robots & Bacon
Role
Built the proposal builder, the quote page, the lead dashboard and the customer history
§ 01

The problem

Custom event quotes were built by hand, so a customer could wait hours or days. Website inquiries came in through several forms and the chat and were copied into different tools, with no single record of who asked for what.

§ 02

What I built

  • Proposal builder. A rep fills one form and gets a branded two-page proposal and PDF on the spot, then edits it in plain English ("change the DJ to TBD").
  • Instant quotes. The website quote page gives an all-in estimate and opens the booking system's live calendar with the date filled in. Every request gets a quote email within seconds.
  • Lead dashboard. Who came in, from which channel, and what they booked; click any lead or booking count to see the people behind it.
  • One customer history. Every website inquiry, from any form or the chat, lands in one history per customer in Premier's own database, next to their bookings and quote emails.
§ 03

My part, and who else

These four are mine, built with Claude Code. They sit inside Premier's own site and back office, which the owner's team wrote.

§ 04

How it works

  1. 01Rep's requestPlain English
  2. 02AI drafts the edit
  3. 03Checks in codePrice, policy, season, photos, fit
  4. 04Rep reviewsCan undo AI edits
  5. 05Proposal updated
Fig. 1 Every AI edit is checked in code before the rep sees it.

How I know it works

Tests, review, measured results

CheckHowResult

Price attacks

Deliberate attempts to slip a made-up price past the checks, inside a 21-case test run

6 of 6 refused

Pass

Custom quote time

From my work with the sales team; not formally timed

Hours or days → minutes

Quote email speed

Time from a website quote request to the emailed all-in quote, Sep 4–18, 2026

1–2 seconds (median)

§ 05

Decisions and what went wrong

  • Availability comes only from the booking system, and prices from the rep or the company's price sheet, never from the AI. A price on a proposal can only be TBD, a number the rep gave, or one already on the proposal.
  • Checks in code, not in the prompt, because a prompt is only probably obeyed.

What went wrong: Before anyone used it, a separate, skeptical Claude audit tried to break it and found 13 issues. Seven were fixed, including two ways a wrong price could slip in. One limit remains by design: any number in the rep's message counts as approval, even a year, so the rep still checks every price.

Why it matters in healthcare

It's the core problem of AI in regulated work: let the model help, but make certain mistakes impossible, not just unlikely.

§ 06

What changed

A custom quote now takes a rep minutes instead of hours or days by hand. Every website quote request gets an emailed quote within seconds, and the team can open any website inquiry and see that person's bookings and quote emails alongside it.

Under the hood

8

checks in code on every AI edit

6 of 6

price attacks refused

1–2 s

median quote email

  • Claude
  • React
  • Supabase
  • Postgres
  • Netlify