AI systems · operating leadership
Most people sell you an AI tool and wave goodbye. I read how an organization actually runs, design the system that fits it, then stay until people adopt it. I know it works because I live on it: four companies, largely solo, on an operating system I built and use every day.
The problem it answers
Organizations buy tools that turn into shelfware inside a quarter. The model was never the missing piece. What's missing is the architecture that fits how the place actually runs, and the leadership to make people adopt it. I do both, because one without the other doesn't hold.
Pillar one
Real, deployed, and checkable. Each one runs in production today, with an honest status on what's live and what isn't.
The control plane for the whole portfolio: clients, projects, and a ranked do-now queue, fed from the source files instead of hand-maintained. The live database is the source of truth, and edits flow back to the repo on their own.
boringos.online
Seventy-plus role-scoped agents organized into per-brand teams, built on a five-primitive engineering layer with evals and a banned-patterns rubric. An org chart rendered as software under test. Invoked on demand, not running the company on autopilot.
Agent · Workflow · Eval · Registry
A governed back-office platform deployed for a real church: roles, an audit log, events approval, and two-person giving governance. My own adversarial audit caught a real bypass before it mattered.
Friends Church Orange
A 13-domain knowledge base I query instead of browse, kept clean by an ingest-and-lint discipline. Institutional memory that doesn't walk out the door when someone leaves.
second-brain · query · lint
WiFi-based volumetric sensing: a neural decoder reading occupancy off ordinary signals, no cameras. Genuine edge-AI and RF research, simulation-validated today and honest about it.
Portillo Technologies
Pillar two
I know why most systems become shelfware, and I have the rhythm and judgment to keep these from rotting.
A daily digest capped at three ranked actions, each tied to a decision. The rhythm that keeps a system in use instead of on a shelf.
Every Friday, something gets cut. "Lots of momentum" without a number doesn't count. Subtractive management, made routine.
One-way doors get the weight they deserve; two-way doors get speed. Commit or defer, never drift.
A leadership-coaching engine, a decision coach, and a CFO-in-a-file that reduces multi-entity finance to one default-alive or default-dead call. Automation is the easy half. This is the other half.
The whole claim in one fact. The architecture is good enough that attention scales without bodies, and I'm the proof of concept.
Capacity, RevOps, process docs, runbooks. The unglamorous scaffolding that turns a good system into one a team can actually run.
Why the two hold together
The system surfaces what needs deciding.
The cadence forces the decision.
The decision changes the work.
The work updates the system.
And back to the top, on its own. Nobody hand-maintains the middle. That loop is the product.
The receipts
Every number on this page traces to a live URL, a test count, or a git history. The detailed receipts live in Boring Proof.
The live control plane running the whole portfolio.
boringos.onlineThe full dossier, every claim backed by a gate, a test count, or a URL.
proof.boringlabs.onlineThe same approach aimed at a single industry: a toolkit for architecture firms.
boring-for-architects.pages.devA governed platform live for a real church. Admin and public forms held fail-closed until access control is wired.
In an org · status honestOne org, led the change
I designed and deployed Church OS there: a governed platform with roles, an audit trail, events approval, and two-person giving governance, built for an organization I'm responsible for week to week.
During my own adversarial audit I caught a real two-person-giving bypass and closed it before it could be used. That's the difference between shipping a tool and owning the outcome.
Honest status: it's deployed and running, with admin and public forms held fail-closed until we wire Cloudflare Access and Turnstile. This is the clearest case of me leading a system into an organization, not just handing one over.
The honest part
Credibility is the feature. These stay on the page on purpose.
I run solo, with agents and a few client engagements. The honest unit is my own portfolio and Friends Church Orange as TD. No enterprise headcount, no claim of one.
No PROSCI, no Prince2. The change-management claim is practitioner judgment from doing it, not a certificate.
Governed and tested, invoked on demand and inside workflows. Not robots running the company on a timer.
Simulation-validated, planar only, pre-provisional. Real ML and RF research, never sold as a deployed product.
It proves the operating model and it caps the leadership claim. The orgs I've most changed are my own and the church I run tech for. I won't imply a record of turning around outside companies.
How to work with me
This isn't "brief the maker and get a tool back." I read how your organization runs, design the system that fits it, and stay embedded long enough that your people own it and the change holds. Fractional or project-based. The goal is an organization that runs better without me in the room.
Where this starts
Tell me where the organization loses time or drops the ball. I'll tell you what I'd architect first, how I'd make it stick, and what it would take. No deck.
Jason@boringstories.io