Jared Brown

Work

Case study

TRS: one reconciliation tool to an AI-powered operations stack

A family-owned brand asked how AI could make its Amazon operation more efficient. Two years later it runs on a stack of tools I built, and the business has grown.

The client is TRS, a family-owned brand selling SDS binders and estate-planning binders on Amazon, with no in-house technical staff. I mapped the workflow, directed the builds, reviewed and shipped them, and run the ad operations.

Key results

  • ~3 days of manual work removed each month
  • $2–3K a month in ad spend saved
  • +28% revenue, year over year September 2026 vs. September 2025
  • 15 of 15 pilot customers rated it “very useful” Safety resource centers
  • 2 years engagement

Problem

TRS sells safety data sheet (SDS) binders and estate-planning binders on Amazon.

Monthly settlement reconciliation, matching Amazon’s payouts to the orders and fees behind them, was manual and error-prone. Reporting took days. Ad spend was rising without a clear link to return. The company had no in-house technical staff.

Who it was for

The owner and a small operations team.

What I built

One tool at a time, in the order it happened.

The stack, built from the bottom up.
  1. Discovery. Mapped the monthly workflow and found the two most expensive manual tasks: payment reconciliation and monthly reporting.
  2. Settlement Reconciliation Tool. Ingests transaction CSVs, Orders Detail reports, cost-of-goods data, and payment PDFs, and produces per-product (per-ASIN) profitability and reconciliation checks. It removed about 3 days of manual work a month.
  3. KPI Dashboard with Claude. Sales, PPC, margins, SKU performance, and B2B revenue in one view (PPC is the industry’s common name for cost-per-click advertising). A Netlify serverless function sends structured prompts to Claude (Haiku) through the Anthropic API, which explains month-over-month changes in plain English.
  4. Advertising operations. I took over 20+ Sponsored Products campaigns: cost-per-click (CPC) ads for individual product listings, shown in Amazon shopping results and on product pages; you pay only when a shopper clicks. I restructured targeting, bids, negatives, and budgets. Spend is down $2–3K a month.
  5. Digital Resource Centers. A no-cost product enhancement. I used AI to research federal and state OSHA hazard-communication requirements, then built QR-linked companion sites on Cloudflare Pages with fillable worksheets (chemical inventories, training logs, emergency contacts), in English and Spanish, with state-specific modules.
  6. P11. A Python engine that audits PPC campaign structure and flags waste, so the diagnosis repeats every month.

Built with

ChatGPT for planning, Codex and Claude Code for builds, Claude as reviewer. The Anthropic API, Netlify Functions, Cloudflare Pages, Python, and PapaParse.

My role: scoping with the owner and the operations team, directing the builds, reviewing, and shipping. I also run the ad operations.

Demo

  • KPI dashboard on an invented brand with synthetic data, with Claude diagnostics liveComing
  • Live link to the TRS Resource CenterComing

Demo environment with synthetic data. Production runs on live data that cannot be shown.

What happened

  • Reconciliation: about 3 days a month of manual work removed.
  • Advertising: spend down $2–3K a month.
  • Revenue: up 28% year over year, September 2026 vs. September 2025.
  • Resource centers: piloted with 15 existing customers. All 15 rated them “very useful,” the top choice on a three-point scale.

What I’d change

  • A shared data model, so the reconciliation tool and the dashboard don’t each parse the same reports.