Managed IT · Private AI

AI that never leaves the building.

SNSYS builds a private AI system for the business and installs it inside the company's own walls. One machine, loaded with a leading open model and a library of the company's documents. The team asks questions in plain English and gets answers drawn from their own files, with the source cited. Questions and answers live and die on that machine.

From $9,900 installed + $995 / month · no-upfront option available

The Internet · out of the loop
Inside the building
The Library

Company files

Contracts, drawings, SOPs, job history

The Machine

Private AI

Reads the library, answers with sources

The Team

Any browser

Plain-English questions from any office desk

Documents · questions · answers stay here Out: hardware pulse only
How It Works

Built at our bench, delivered to your rack.

The same operations center that builds and runs our managed networks builds every AI machine, burns it in, and installs it on site.

01

We build it at our bench

The machine is assembled and burned in at our operations center. The model goes on, the security policy loads, and the unit learns exactly which shelves of the company's files it may read.

02

It goes in the rack

The crew that installs our networks mounts it: one power cord, one network cable. It appears on the office network the way a printer does. Its only connection out is to our bench, for signed updates and the hardware pulse.

03

The team just asks

A page opens in any office browser. Ask what the Hartford bid closed at in 2023, and the answer starts in about a second, with the file it came from. New documents join the library on their own.

Data Security

The machine exists to protect sensitive documents.

Drawings, contracts, client records, and trade secrets carry a company's worth. Every AI question about them normally travels to a cloud service, lands on someone else's servers, and lives under someone else's terms. This machine removes the trip. The model, the documents, and every question and answer stay on hardware inside the building, encrypted at rest, under keys the company holds.

Each document library is scoped to the people cleared for it, and the owner can read the full record of who asked what. Free chatbot tabs are open in every office in America today, and confidential files drift into them one paste at a time. The machine gives those questions a private place to land.

Who Sees What

We see the machine's pulse and nothing else.

Stays in the building
  • Every document the machine reads The library lives on its drives, encrypted at rest
  • Every question and every answer Nothing typed by the team ever crosses the property line
  • The tuned model itself Built from the company's files, it belongs to this deployment
  • Who asked what Usage stays a company record, readable by the owner alone
Reaches SNSYS
  • Hardware pulse Temperature, fan speed, drive health, uptime
  • Update status Model and security updates, staged and signed at our bench
  • Capacity signals A heads-up that the office is outgrowing the machine
  • Nothing else File contents, questions, and answers are unreadable to us by design
The Model

A model tuned to the business, owned like the building.

The Source

Company documents

Quotes, drawings, SOPs, job history. Two paths, both chosen by the company.

The Library

Indexed as files change

New and edited documents join the index on their own

Daily Result

Answers that cite the file

Every answer names the document it came from, current as of today

The Training

The model studies the corpus

An approved set of documents teaches it the shop's language and habits

Base model · stays intact

gpt-oss, by OpenAI

+ Add-on · 1% size

Training builds a small add-on that snaps onto the base model

Ships back home

The company's model

Loaded onto the machine in the building. Nobody else ever runs it.

Training corpus: approved by the company · deleted after training Base model: gpt-oss, published by OpenAI · Apache 2.0 · full paper trail
Who It Fits

Built for the files that can never leave.

Defense Suppliers

CUI meets AI, legally

Drawings marked ITAR and files marked CUI stay inside the facility while the whole archive becomes searchable. The Complete plan includes the data-flow documentation a CMMC assessment asks for.

Law Firms

Privilege stays privileged

Ask across every matter the firm has ever closed, in plain English. Client material stays in the office, every word of it. The confidentiality answer for clients becomes one sentence long.

Accounting Firms

Client books stay home

Years of returns, ledgers, and workpapers become one searchable memory while every record stays under the firm's roof. Busy-season questions come back in seconds, with the workpaper named.

Pricing

One machine, one monthly price.

Essentials
$995
per month + $9,900 installed
  • The AI machine, built, installed, and replaced free if it fails
  • Private chat for the whole office
  • Document library with cited answers
  • Quarterly model upgrades at our bench
  • Business-hours support
Most Companies
Managed
$1,245
per month + $9,900 installed
  • Everything in Essentials
  • Microsoft 365 and Google Workspace connectors
  • Single sign-on with the company directory
  • Monthly usage review for leadership
  • Priority support queue
Complete
$1,495
per month + $9,900 installed
  • Everything in Managed
  • A model tuned on the company's own corpus
  • Compliance documentation pack for CMMC and insurers
  • 24/7 coverage with 15-minute critical response
  • Annual tuning refresh

No-upfront option: add $500 per month to any plan on a 36-month term and installation is $0. Larger offices size up to a dual-engine machine, priced at the walkthrough. Download the one-page overview (PDF) to pass around the leadership table.

Questions

Asked on almost every call.

Does our data train anyone else's AI?

No. The model runs on the machine in the building and sends nothing back to any AI company. Documents, questions, and answers stay on that machine. The tuned model on the Complete plan is built from the company's own files and belongs to that deployment alone.

Which AI is inside the machine?

gpt-oss, the open model OpenAI published for exactly this kind of deployment, under a license that lets it live permanently on hardware the business controls. American lab, public license, full paper trail. A security reviewer can trace every part of the stack.

What happens when the internet goes out?

The AI keeps answering. It reads the model and the documents from its own drives, so the outside connection was never part of a question. Updates and our health checks resume when the line comes back.

What can SNSYS see?

Hardware pulse only: temperature, disk health, uptime, update status. File contents, questions, and answers stay on the machine, unreadable to us by design.

We already use ChatGPT or Copilot. Where does this fit?

Many clients keep both. The cloud accounts stay for public questions. The machine takes the files that can never leave: contracts, drawings, client records, anything under CUI, privilege, or client confidentiality.

How many people can it serve?

A whole office at once. We size the machine to headcount at the walkthrough, and a bigger unit swaps in as the company grows.

See the data path on a 15-minute call.

Bring ten documents to the walkthrough. We load them live, show where they sit and who can reach them, and answer the security questions on the spot.

Book a 15-Minute Walkthrough Download the One-Pager