AI that never leaves your office.

Ownbox gives your team an assistant that answers from your own files. It runs on hardware we install for your firm alone, and nothing is ever sent to an AI company.

Painted after Claude Monet, Cliff Walk at Pourville, 1882

Ask it about your own files

It reads your contracts, policies and spreadsheets, answers in plain English, and tells you which document each answer came from. Choose a question to try it.

Ownbox assistant, Ashford & Lane (example)

Try a question

Speed 0 words/sSent outside nothingRuns on 1 box

Example questions and documents. In a real install, the assistant only reads the folders you connect, and only for the people allowed to open them.

See where every question goes

A cloud AI sends each question, and any file attached to it, to servers run by an AI company. Ownbox answers from a box in your own comms room. Switch between them to compare.

Staff in an office send questions with files. With Ownbox they go to a box inside the office. With a cloud AI they leave the office for an AI company's servers, where copies pile up. Your office Everything inside this line stays with you Ownboxin your comms room AI company Servers in another state or country Copies kept on their servers

Files that left your office0

Every question stays in your office

Nothing sent to an AI companyThe model runs on your own box. No outside company sees your questions, and none of your files are used to train anything.
Access by departmentPeople sign in with their Microsoft work accounts and only get answers from files they're already allowed to open.
Works without the internetIf your connection drops, the assistant keeps working. Our maintenance link is encrypted, and you can switch it off.

Made for offices that keep files confidential

Ownbox is for small and mid-sized firms whose clients expect their documents to stay private. Choose your field to see how teams use it.

A small computer, kept in your building

It sits on a shelf in your comms room, or in a local datacenter if you prefer. We install it, keep it updated and watch it around the clock.

128 GBof memory, enough for large open models
~25regular users per box
1 dayfrom delivery to your team using it
0 bytessent to AI companies
DGX SPARK NVIDIA DGX Spark, about 15 cm wide
Chip
NVIDIA GB10 Grace Blackwell
Sign-in
Microsoft work accounts
Backups
Settings, every night
Updates
Tested, then installed overnight

Models we install

We test models on your own documents and install the one that fits. You can switch later at no cost. Speeds are approximate single-user figures from public benchmarks; we measure on site before quoting.

Choose where it runs

The assistant and the price are the same either way, and the hardware is set up for your firm alone.

Where it runs

Running in a day

  1. 1About an hour, free

    Assessment

    We count who will use it, find where your files live, and give you a fixed quote.

  2. 2Install morning

    Install

    The hardware goes in, and the model is loaded and tested on your documents.

  3. 3Install afternoon

    Connect

    Your folders are read in and access is set department by department.

  4. 4Every day after

    Looked after

    We monitor it, install updates overnight and back up settings nightly.

One flat monthly price

Hardware, setup, updates and monitoring are included, with no usage limits. Move the slider to your headcount.

25

$30

per person, per month

$625

a month for 25 cloud AI seats

None

usage limits

Compare with

Office plan

$750

per month, flat

  • 1 dedicated box
  • Up to 25 people
  • Updates, monitoring and nightly backups
  • $2,500 one-time setup, 36-month term
Book a demo

Office plan: one box for up to 25 people, $750 a month. Firm plan: two boxes for up to 50 people, $1,350 a month. Over 50 people, we quote after an assessment. Cloud prices are typical business-plan list prices. Your final quote comes after the free assessment.

What firms ask us first

Short answers to what partners, practice managers and office managers ask most.

For drafting, summarising and answering from your own documents, it is strong. Open models trail the very best cloud models by a few months on the hardest tasks, so we test it on your files before you commit.

No. The model, the chats and your documents stay on your dedicated hardware. Our maintenance link is encrypted, and you can switch it off at any time.

No. Access is set by department, and people sign in with their existing Microsoft work accounts. The assistant only answers from files the person asking is allowed to open.

On the Firm plan, the second box keeps the office running at reduced speed. Either way we get an alert, and settings are backed up nightly, so a replacement is running within days.

Not during your agreement. The price is fixed for 36 months, whatever happens to cloud AI prices, and it doesn't change with how much your team uses it.

One working day on site after a free one-hour assessment. Your team signs in the same afternoon.

See it running on your own files

Book a free demo and a one-hour assessment. You get a fixed quote before anything is bought.

Prefer email? Write to hello@ownbox.ai. We reply within one working day.