Self-Hosted and On-Premise AI
Private AI for business, run where you decide. On-premise AI is hardware we install in your office, and nothing leaves the building. Self-hosted AI can also be a private instance we manage, where the library is yours alone. The product is private AI: ask in plain English, get an answer with a citation back to the file.
The demo line is answering right now. Call the Ava AI employee demo and pretend you were just in an accident.
On premise
Hardware in your office, internet optionalPrivate instance
Yours alone, not pooled, not for trainingCited answers
Back to the document and the pageIn the building, or on an instance that is yours alone
Same answers either way. The difference is where the documents, the index and the model sit.
Nothing leaves the building
On-premise AI is hardware we install in your office. The documents, the index and the model all sit inside your walls, and it works with the internet unplugged.
Yours alone
Self-hosted AI on a private instance we manage. The library is yours, not pooled with anyone else, and not used as training data.
Back to the file and the page
Every answer names the document it came from. If it cannot support an answer from your library, it says so rather than inventing one.
The folder nobody can search
Search finds a filename if you already know the filename. It does not answer a question that spans forty documents.
Contracts, policies, service histories, engineering records, the shared drive that four people understand and everybody else avoids. A deployment indexes the library and answers across all of it at once, in plain English, with a citation back to the document and the page. It reads scans with no text layer. It loads from a network folder or syncs from the systems you already use, and access is granted per person so separate collections stay separate.
This page is the deployment choice. The product, what you ask it and what you get back, is private AI. For a firm, the same build is the AI paralegal, and the wider set is AI for law firms.
Nothing in your library is used to train anyone's model. We run it either way, flat monthly, and you get the founder's cell number rather than a ticket queue. The written terms are the service agreement.
- Answers across every document at once, not only inside one file
- Reads scans with no text layer, and the ordinary office formats
- Loads from a network folder, or syncs from the systems you use
- Access granted per person, separate collections stay separate
- On-premise AI, or self-hosted AI on a private instance
Scoped, built, and then actually maintained
Flat setup fee plus flat monthly management. No per minute billing, and no support queue between you and the person who built it. The monthly work is managed AI services.
A scope call
Thirty minutes on the work you would hand over. We find what is worth doing first and quote it flat. If there is nothing worth doing yet, we say so.
We build it
Trained on your business, your criteria, your calendar and your systems. You approve every word and every permission before it goes live.
We run it
Monitoring, changes, and the fixes for when a vendor moves something. You get the founder's cell number, not a ticket queue.
Where the model and the documents sit
Same product either way. The difference is hardware, whether it works offline, and where it is installed.
Swipe sideways to read every column.
| On premise | Private hosted | |
|---|---|---|
| Hardware | Hardware we specify and install in your office. | No hardware to buy. A private instance we manage. |
| Setup | Hardware plus a flat setup fee, quoted after a scope call. | A flat setup fee, quoted after a scope call, with no hardware to buy. |
| Monthly | Flat monthly management. | Flat monthly management. |
| Offline | It keeps working with the internet unplugged. | A private hosted instance needs its connection. |
| Who maintains it | We run it. | We manage the private instance. |
The questions that come up first
If yours is not here, ask it on the walkthrough. You will get a straight answer, including when the answer is that we are not the right fit.
What is the difference between on-premise AI and a private instance?
On-premise AI is hardware we install in your office. The documents, the index and the model all sit inside your walls, and it works with the internet unplugged. The other way to run self-hosted AI is a private instance we manage: a dedicated instance where the library is yours, not pooled with anyone else, and not used as training data.
Is our data used for training?
No. Nothing in your library is used to train anyone's model, and on premise it does not leave your building at all.
What if the internet goes out?
On premise, it keeps working with the internet unplugged. A private hosted instance needs its connection, the same as any system we manage for you.
What file types can it read?
The ordinary office formats plus scans with no text layer, which is usually the pile that made search useless in the first place.
Has this actually been built?
Yes. The first one runs inside law firms as the AI paralegal, answering across a whole case library. The build underneath is the same for any document heavy business.
How is this different from the private AI page?
Same deployment. Private AI is the product: ask in plain English, get an answer with a citation back to the file. This page is the choice of where it runs, on-premise AI or a private instance we manage. Either way we run it, flat monthly.
What is private (self-hosted) AI?
Private (self-hosted) AI, as Wide Awake AI builds it from San Diego, runs the documents, the index, and the language model on hardware in your office, or on a private instance that is not pooled with anyone else. On premise, nothing leaves the building, and nothing in your library is used to train anyone's model.
What hardware does self-hosted AI need?
On-premise AI runs on hardware Wide Awake AI specifies and installs in your office. This site does not publish a parts list. The other way to run it is a private instance we manage, with no hardware to buy.
Which models does private AI use?
This site does not name a model. It says the language model runs with your documents and the index, on hardware in your office or on a private instance, and that nothing in your library is used to train anyone's model.
Is private AI HIPAA or ABA compliant?
This site does not say the deployment is HIPAA certified or ABA certified. It says an on-premise deployment keeps the documents, the index, and the model in your office, that nothing in your library is used to train a model, and that access is granted per person. Those are confidentiality controls, not a certification.
Where this sits
Start with the umbrella, or go straight to the line that matches the work.
Fifteen minutes, straight with the founder
Bring the thing that annoys you most. If it is worth doing you get a flat number on the call. If it is not, you will hear that too. The demo line, Ava, is (725) 745-7092.
- [email protected]
- Based in
- San Diego, CA