01
How people decide
How your best people handle a variance, protect a margin, and decide what ships. Written down so the company can run it.
The Agentic Layer
Your company's knowledge becomes an operating engine.
Public models make generic knowledge cheap. The advantage is turning your data, systems of record, and judgment into an asset the company owns and runs.
In production
Sierra Nevada Brewing (~$420M). Named owner. Path the company kept. Not a chatbot beside the business.
The waves
Every ten-year wave was the same game. This one is not.
Mainframe. Client-server and ERP. Cloud and SaaS. A new platform, a long implementation, then keep-running. Fairly similar, every time.
Agents do not wait for a decade. They land on the stack you already have. The game is forever changed.
These are what you have to be able to see.
You cannot put agents into the work until this is visible. I come inside and take that on, one consequential workflow at a time.
01
What's already running?
Including the unofficial shadow AI and fragmented tools. If you cannot name it, you cannot govern it.
02
What can it see and do?
Not a policy PDF. What it can reach across ERP, WMS, and CRM, and what it cannot.
03
Who keeps the last move?
A named person is accountable when an agent acts, not when it drafts.
04
Can you prove what happened?
A record of what ran, what data was touched, and whether it finished.
The path
I name what moves, what stays human, and what waits.
Start with one consequential workflow already carrying the company. Connect the systems you already own. Set the baseline. Automate where the evidence can hold it. A change does not go live until the work still holds. Keep a record of what ran, who owned it, and whether it finished.
A person keeps the last move on anything that matters. People and agents in the same work — on the same six constellations, not in a seventh bucket. Then we reuse the pattern, or we stop. The people doing the work leave more able to run it.
The last move stays human
Agents assemble. A named person authorizes.
Agents gather across ERP, WMS, and MES, reconcile the gaps, and put a complete proposal in front of a person. No agent writes to the core systems in the dark. A named person inspects the receipts and authorizes the last move.
01 Read the estate
Live ERP, WMS, and plant signals
02 Agent drafts
Reconciles the gaps and writes the brief
03 Person authorizes
Inspects the receipts and keeps the last move
04 The write lands
The change runs with a named owner
No write without a person
Ledger balances, pricing, and production batches do not change unless a named person signs.
Prove what happened
A record of what ran, what was touched, who authorized it, and whether it finished.
The knowledge the company keepsOwned, not rented
Agentics flattens commodity knowledge. The moat is what your company knows.
Public models rent you the same answers everyone else can buy. The work is capturing how this company actually runs, and keeping that as an asset you own.
02
The data the company owns
Agents work against live master data and transaction records the company already owns. Not a rented pool beside the estate.
03
What you keep
Every finished workflow, exception, and human sign-off stays with the business. You keep what we build when I leave.
How I deliver
In productionThis is the build. Not software for sale.
I don't sell SaaS subscriptions or partner markups. I walk into your company with an earned mountain of production-tested templates, patterns, and receipts from day one. Whether you want all three systems working in concert or a single capability to solve an immediate bottleneck, you get an instant jump start. Your company keeps what we stand up.
PLAN · ORCHESTRATE · BUILD · SHIP
The Rig
The human-directed build platform for companies that want to build now without waiting for a lab. Gated plans, verifiable steps, and a complete audit trail where no change goes live until the work holds. Your team gets an immediate jump start on day one, and the company keeps the engine.
DISCOVER · MAP · TWIN · GOVERN
The BOSS
Thirty years of enterprise transformation craft distilled into a living digital twin, the tool of tools for business transformation. Maps your systems to the six process constellations, connects the people and agents in the Beehive, and proves every workflow through to green on green. Because it is a permanent living twin, the BOSS never goes away: it stays running as the real-time operating map of how your business actually runs.
COMPOSE · EXECUTE · MONITOR · GOVERN
FlowCraft
The 100% agentic workflow building machine, operating in production today on live firms like DigOps. Composes, executes, and governs autonomous AI workflows across your enterprise with deterministic human authorization on every consequential move. Not an automation overlay or rented subscription, but a full agentic operating system your company owns and runs.
See these systems in action.
Whether you need the full operating stack or a single capability, we can walk through live production architecture and receipts.
How the models sit
The cloud rented the stack. Enterprise is bringing it home.
Data is coming back in-house. Rightly so. That now includes the model. Frontier labs stay useful for open-ended reasoning. The work this company already knows how to do should run on a model it owns, on its estate.
The directionThomson Reuters, 2026
They specialized an open-weight model into one they keep.
They still rent a frontier lab for open-ended work. Domain work runs on the model they own. That is where enterprise agentics is heading.
Three seats. They are tools. They are not the owner.
Closed
You rent the model
Fast to start. The lab owns the weights and the path. Useful for open-ended work. Not yours.
Open weight
You specialize the model
Run the weights on your estate. Align them to your data. This is the move enterprise is making.
Open source
You can keep the stack
See it, run it, and own the path. Closest to a model the company actually keeps.
OpenAI
Anthropic
xAI
Thinking Machines
Frontier seats. Useful. Not the destination.
Bring one consequential workflow.
I will tell you whether it should become agentic, what it can change, and what must stay under human command.