tier 3 — bespoke engagement
the sovereign stack pilot
a hand-built sovereign ai stack, deployed on your hardware, proven against your real workflows. for high-trust operators who can't put their data into someone else's training set.
why we know what good looks like
in 2007, the first iphones came across my counter as a wireless sales manager during the launch. half the customers asked for a flip phone instead. the other half changed how they worked inside a month. i watched the rejection-then-mass-adoption arc play out in real time, and learned what the early phase of a real shift actually looks like from the operator side.
a decade later, i was selling sovereign compute infrastructure pointed at digital assets to family offices out of a small oregon shop. clients bought the hardware outright. we hosted, maintained, and operated it under signed sla. they never argued price. what they needed was a way to verify who they could trust with the substrate. it was a trust gap, not a money gap.
ai sits in the same shape today. the high-trust operators i work with don't need persuading on the technology — they need to know who they can trust to put it inside their building. so i built it for my own family first. now i run the same playbook for a small number of clients who are early to the same curve.
what the pilot delivers
step 1 — discovery
we map your real workflows
remote-secure access via twingate. agent-assisted scan of your systems. we identify the two or three highest-leverage automations — ranked by time saved, weighted by sensitivity of the data involved.
step 2 — build
hardware: assess, then upgrade or procure → model selection → integration
first we look at what you already have. if your existing hardware fits the workload, we use it. if it's close, we recommend the upgrades or additions to bridge the gap — added ram, a gpu card, a second drive, whatever fits. if you're starting fresh, we spec the right gear (mac mini, mini-pc, or right-sized server) and help you procure it. budget and capacity drive the conversation, not a pre-built spec sheet. then we pick the local models that match what you'll run and wire the integrations to your existing tools. no cloud apis in the data path. no vendor lock-in.
step 3 — prove
install, validate, hand off
remote-first deployment, alongside your local it team for hardware handling. on-site engagements available for the right fit. we validate the stack against the workflows we mapped in step 1 and hand off ownership with a runbook your team can operate. you keep the hardware, the configs, the playbook.
step 4 — support
90 days of async support included
email or direct message. typically responded to within 2 business days. covers clarification, tuning, and troubleshooting. optional managed-services retainer available after the 90-day window if you want us to keep watching it.
how we scope it
every pilot is scoped to your systems, your workload, and your existing infrastructure. pricing reflects engagement complexity — not a per-seat formula and not an off-the-shelf tier. you tell us what you're running and what you want automated. we come back with a written proposal: scope, hardware path, timeline, and a single price you can read, share, and sit with.
hardware is a separate line item — see step 2 for the three paths. 50% at signing, 50% on delivery.
optional managed-services retainer
monthly retainer for ongoing tuning, model upgrades, and second-opinion review. scoped per engagement after the pilot lands. you don't need it on day one.
how we engage
we don't run sales calls or demo gauntlets. you send a message. we ask a few questions to see if it's a fit. you get a written proposal you can read, share, and sit with. if it's not a fit, we say so plainly.
who this is for
family offices that need to stop pasting client material into cloud chatbots. wealth managers whose compliance officers don't want to hear another "ai-assisted" story. law firms with privileged data that can't leave the building. medical practices, small manufacturers, founder-led operators who handle sensitive data and are ready to own the ai stack instead of renting someone else's.
we take 2-3 pilots at a time. limited capacity by design.
what you walk away with
hardware in your building running a working sovereign ai stack, tuned to two or three automations that actually move time off your team's plate. a runbook your operators can follow. a 90-day support window so nothing stalls in the first quarter. and a clean answer when someone asks where your data goes: it doesn't.
submit for pilot scope
tell us what you're running. we'll come back with a scoped proposal in 2 business days. if it's not a fit, we'll say so plainly.
no calls. no demos. just a conversation.