We built an automation workflow that watches the company Gmail inbox for new quote requests, hands each one to a carefully prompted OpenAI call that handles extraction, tier-based pricing, currency selection, and the full quote email body in a single pass, then routes the output through deterministic logic that either drafts a custom-quote response (for pools over 750,000 square inches) or calls the PayPal Orders API to generate a real, working checkout link before saving the finished quote, with payment link included, as a Gmail draft for the owner to review and send.
End result: the owner opens his inbox, sees a fully drafted, properly priced, currency-correct quote with a pay-now button already wired in, reads it, and hits send. Every quote leaves his outbox under his name, in his voice, with a customer-ready payment link, in under 3 minutes of his time.
We mapped the actual quoting workflow, taught one model the full pricing engine, then wired live PayPal payments directly into the output.
We sat down with Gerrit and walked through real quote requests as he would normally handle them. Where did he look up dimensions. Which tier did each pool size land in. How did currency get chosen. Where did GST get added. What happened when a pool was too big for the standard tiers. We catalogued every decision, then turned that catalogue into the rules the automation would follow. The pricing table, the currency logic, the custom-quote threshold, and the GST handling all came directly out of how he was already doing the work.
Instead of stringing together five or six separate LLM calls, we designed one tightly scoped GPT-5 prompt that handles extraction, tier lookup, currency selection, surface area math, and quote email writing in a single pass. The model receives a structured incoming email, applies the five-tier pricing table line by line, picks USD or CAD based on the shipping address, adds the GST note on Canadian orders, drops in a custom-quote response if the pool exceeds 750,000 square inches, and outputs a polished HTML email body ready for the next stage. One call, one job, well prompted.
The deterministic side of the workflow parses the cover cost and shipping cost back out of the LLM output, then makes two PayPal API calls per in-range quote. First an OAuth token request, then a Checkout Orders create call that builds a real PayPal order with the right amount, currency, and item description. The resulting checkout link is dropped into the Gmail draft alongside the quote, so the customer can read the price and click pay in the same email. Currency-aware: CAD orders include the 5 percent GST automatically. US orders do not. Out-of-range quotes get a custom-quote draft instead, with no payment link.
Every piece of this build serves the same goal. Every quote request leaves the inbox as a fully drafted, properly priced, payment-ready email in under five minutes of the owner's time.
.svg)
Custom software, high end websites, leads for your business. We can help...
.avif)
