Conrad B.

I ran two product brands for 20+ years ($30M+ in sales), then rebuilt my own back office on AI agents. A full year of books closed through the QuickBooks API, one inventory record across multiple 3PLs and Amazon, and every action logged. Now I build the same thing for other businesses: AI agents and automations for bookkeeping, inventory, fulfillment, reporting, and the handoffs between your tools. The system I run covers most of what used to need a full-time operations manager. That is the difference between me and most AI developers you will talk to. I have sat in both chairs. I know what an owner needs to see before trusting a system with their money, and I know how to build the thing that earns that trust. WHAT I BUILD Process discovery and workflow mapping. Before any code, I map how your business actually runs and find the work worth automating. Most projects fail here, not in the build. Agentic workflows. Multi-step agents pull the context they need, use your tools, update your systems, and hand decisions back to a human when judgment is required. Built with Claude Code, ChatGPT Codex, or whatever the job calls for, with MCP tool integration and Python. Claude and Claude Code setups for owners and teams. Skills, memory, connectors, and guardrails configured around how your business works, plus the training to run it yourself. Back office automation. Bookkeeping and reconciliation via the QuickBooks Online API, inventory across multiple 3PLs and Amazon FBA, order and exception triage, recurring reporting. Knowledge bases and retrieval. The context layer that makes an agent useful inside your business instead of generically smart. Ecommerce and marketing systems. Amazon SP-API (approved developer), Shopify, Meta Marketing API, cold email infrastructure with deliverability gating. WHAT I HAVE SHIPPED A full year of books closed entirely through agents over the QuickBooks Online API. Roughly ten times faster than the process it replaced, statement-verified against bank records, human-approved at every posting, and it root-caused errors an outside bookkeeper had missed. An inventory system of record spanning multiple 3PLs plus Amazon FBA, with autonomous daily syncs and replenishment orders staged for human review. A five-month international fulfillment backlog cleared in a single session by triangulating four disconnected systems into one actionable board. Exploded View, an open-source Claude Code plugin on GitHub that maps any codebase or agent setup into an interactive diagram you can click through. A self-serve AI adoption assessment for businesses, live on the web, built on Ann Miura-Ko's adoption levels. HOW I WORK Fixed scope, clear checkpoints, and a working system you own at the end. No black boxes and no dependency on me to keep it running. Anything that moves money or talks to your customers waits for a human yes, and every action the system takes is logged. If you are not yet sure what to automate, I can run a paid discovery sprint first and give you the map whether or not you hire me to build it. Bring me your problems. Let's solve them.