I build AI systems that run real businesses.
Behind a $10 million-a-year CPG portfolio, a fleet of governed agents does the work of three full-time roles — sourcing, procurement, and planning — in production, every day. That playbook is now becoming ROST: an operating system for companies staffed by humans and AI agents together.
ROST — the agentic operating system for hybrid companies.
ROST takes an EOS/Traction-style management operating system and rebuilds it so AI agents are first-class staff. Every seat — human or agent — has a charter that defines what it owns, what it may decide on its own, and when it must escalate to a human. Goals cascade down the org chart, metrics flow back up automatically, and every action an agent takes is permission-checked and logged.
Three roles a company would hire for — run by agents instead.
Each of these is work previously done by a remote hire. Today a governed agent owns the seat: it does the daily job unattended and asks a human only for the decisions that commit money or risk.
Runs the purchase-order desk
Owns the purchase-order lifecycle end to end — creating orders, booking CH Robinson freight, filing documents, and reconciling systems — then holds the few money-moving steps for a signature.
3 unattended runs a dayRuns supplier discovery
Discovers candidate suppliers, posts RFQs, extracts every quote no matter the format, tracks samples, and hands procurement a clean, ranked supplier package.
runs daily, quotes checked twice a dayRuns production planning
Turns live demand, inventory, bills of materials, capacity, and cash into an executable 8-week plan of what to make, ship, and buy — written straight back into the operating tables.
solves on demand in about five minutesThe connected system that runs the portfolio today.
Sales channels feed a sync engine, which lands clean data in one shared backbone. The operating fleet reads that backbone and writes its decisions back — so every app and agent works from the same source of truth, not spreadsheet fragments.
scheduled jobs · health logs
system of record · ops tables
analytics · sales · COGS
company brain memory
Custom ERP
The single screen the business runs from — inventory, purchasing, manufacturing, fulfillment, and the production plan in one place.
in service dailyMRP Planner
Turns live demand, inventory, bills of materials, capacity, and cash into an executable 8-week production and purchasing plan.
runs on demand · ~5-minute solveNovi
A governed company memory and AI teammate that answers with citations, remembers with permission awareness, and takes action across the tools the team already uses.
live · in daily useProcurement Agent
Owns the purchase-order lifecycle end to end and books freight — from parsing a supplier PDF to filing the shipping documents.
runs 3× a day unattendedSourcing Agent
Discovers suppliers, posts RFQs, extracts every quote, tracks samples, and hands procurement a clean, ranked supplier package.
runs daily · quotes checked 2×/dayHermes
An authenticated build-and-improve agent embedded in the ERP — scoped work in, reviewed pull requests out, risky changes held behind approval.
in service · gatedShippable apps built on the same playbook.
Products that stand on their own, built with the same agent-assisted engineering process as the fleet above.
Trend Intelligence Platform
Pulls signals from Reddit, YouTube, X, news, and marketplace data, then scores emerging product trends so buyers see what is moving before it peaks.
live · login-gatedCashflowSim
A 730-day Amazon FBA cash-flow simulator with competitor research that reads cost and spec data straight from product images.
liveElevatez Platform
A medspa and clinical assessment platform — lead-generation and assessment flows for a real vertical, with SMS alerting.
liveThe output behind all of it.
Everything above was designed, built, and is operated by one person using a repeatable, agent-assisted engineering process. The numbers below are the measurable trace of that process.
find the signal
scope · source of truth
Codex + Claude Code
test like a user
deploy · monitor
turn work into leverage
Derrek Wiedeman
I'm an operator who learned to build. I run a CPG portfolio doing about $10 million a year across multiple channels, and over the past couple of years I designed, built, and now operate the fleet of governed agents that runs its sourcing, procurement, and planning day to day.
That work convinced me the same pattern belongs in every company — not as a demo, but as the system the business actually runs on. So I'm now building ROST: an operating system for companies that run on humans and AI agents together.
Tell me what you want AI to do in your business.
I'll tell you straight whether I'd build it, and how. 30 minutes, no deck.