One private AI brain for your whole company. On a machine you own.

Your company's AI brain: a machine in your office running local AI, connected to your files, email, CRM, accounting and phones. Everyone gets an account. At implementation it reads every file once, so it knows what each file is, whose it is and which version is current. Ask it about a client, a job, an invoice or a contract and it answers from all of those systems at once, with a link to the source. One source of truth, and nothing is sent to an outside AI company.

Every answer comes with a link to the file or record it came from, so anyone can check it, and every link is checked against your real files, so it cannot send you to a file that does not exist. It is strongest on totals and counts from your job records, looking up a customer, facts from a document with the source named, summaries of a job, and finding photos. Finding one specific file among thousands is harder: expect the right file, or a short list with the right one in it. Like every AI, it is not right every time.

Those run in the vendor's cloud and charge per person per month, and their terms and the law can change what happens to your data. This runs on hardware you own, with no per-seat fee. It costs more than a per-seat plan: you buy the machine, and we implement it and manage it every month. And the big cloud models are still smarter on long, complex tasks. Private AI is about control, not about being cheaper or cleverer.

For 2 to 10 people, the Team level: a Mac Studio M5 Max with 128 GB, running Qwen3.6-35B. For 10 to 40 people or long documents, the Office level: a Mac Studio M5 Ultra with 256 GB, running GLM-5.3-Flash, a model about nine times the size. For 40 or more people, the Company level: the 512 GB Mac Studio Ultra, running the same model at full precision plus a second model. Apple ships it in late October 2026 and has not announced the price. The levels page explains what each step unlocks.

It is the quietest way to get a lot of fast memory in one box. The M5 Ultra moves memory at 1.2 terabytes a second, more than four times a DGX Spark, and comes with 256 or 512 GB, enough for a 320B-class model. A graphics-card tower with that much memory is a loud, multi-card server. The Mac sits silently on a shelf, and your IT provider already knows how to look after it. We still build towers for one person; see custom builds.

No product does. Keeping client data on a machine you control removes the outside AI vendor from the picture, which is the part those rules worry about most, but your own policies, access controls and agreements still apply. Ask your own adviser; we will happily walk them through how it works.

The implementation is $5,000 to $7,500 at every level, based on how much data you have and how many software platforms we connect, and the management plan is $1,500 a month. Team on a 128 GB Mac Studio M5 Max: $4,799 hardware, the implementation and twelve months of the plan come to $27,799 to $30,299 in year one and $18,000 a year after, $63,799 to $66,299 over three years. Office on a 256 GB Mac Studio M5 Ultra: $9,499 hardware, so $32,499 to $34,999 in year one and $68,499 to $70,999 over three years. Company is the same implementation and plan plus the 512 GB Mac, once Apple prices it. Hardware prices are Apple's and move when theirs do.

If it isn't answering questions from your own documents 30 days after implementation, the implementation fee is refunded. You keep the machine.

Yes. That is the point of it. We connect the machine to the systems your business runs on, with the permissions you set. Your files stay where they are: nothing is moved or renamed, and what the AI knows about them stays on the machine. Name the systems you use on the consultation and we will tell you exactly which connect; the fixed quote lists every one.

Three sister brands. Spectre implements a private company brain: local AI on a machine you own, connected to all your systems, with every model running inside your building. Fountainhead builds the same brain in your own cloud and AI accounts, asked from Teams, Google Chat, ChatGPT or Claude, and manages it month to month, with no machine to buy. The management plan is the same price at both; our implementation costs a little more because of the hardware. If your data has to stay in the building, that's us. Greatvine is an AI receptionist for business phone lines; it answers ours. You can buy any one of them on its own.

On site within about two hours of Nashville, Tennessee. Everywhere else in the United States we configure the machine here, ship it, and finish the implementation with you over a video call. Questions first? Call (615) 861-8883; an assistant answers around the clock and can book a consultation.