Private AI · Early access
Use AI on client files. Keep every file in your office.
Your staff already paste client data into public AI tools. Private AI gives them the same kind of assistant, document Q&A and analytics on an on-prem AI server at your site. Nothing goes to a public cloud. First installs begin November 2026.
- Runs fully offline
- Predictable cost, no per-token bills
- Installed and supported by HTL 365
The problem
Your team is already using AI. The question is where your data goes.
A staff accountant pastes a client's trial balance into a free chatbot to save an hour. A paralegal uploads a draft agreement to summarize it. Once it's sent, you can't pull it back, and you may not know it happened.
Public AI also bills by usage, so costs rise as adoption rises. Private AI runs on hardware you own, with the AI platform license included, so you can budget for it.
The solution
AI that runs on a server you own, in your office.
Private AI is a single desk-side server, installed at your site. Your staff get chat, document Q&A and analytics in a browser. The models, the data and the answers stay on that box.
Your data stays put
Runs fully air-gapped, or with outbound-only connections for updates. Files, search indexes and logs are encrypted on the server (AES-256). Nothing goes to a public cloud.
You control who sees what
Role-based access by user, model and data set. An audit log records who did what, to which file, and the result.
A cost you can budget
Hardware you own with the AI platform license included. No per-token bill that climbs every time your team asks a question.
Private AI platform
Five modules. One server in your office.
Every module runs on the same private AI server. Your data, models and answers stay on-site.
Your office · Nothing leaves the building
Installed and managed by HTL 365
Data Hub
Connect databases, files and SFTP. Clean and catalog your data.
Analytics & BI Dashboards
Ask in plain English. Get insights, charts and dashboards.
AutoML & Scheduling
Build and explain prediction models. Run jobs on a schedule.
Knowledge Base (Document Q&A)
Upload files, OCR scans, get answers grounded in your documents.
AI Model Manager
Choose which AI models run, and which model each agent uses.
What your team can do
Six tools, one server.
- 01
Private AI assistant
ChatGPT-style chat on leading open models such as gpt-oss, Llama, Gemma and Qwen.
For a CPA or law firm: Draft a client email or summarize a long memo without it leaving the office.
- 02
Document Q&A
Search your PDFs, Office files and SharePoint, then get answers that cite them. Reads scanned PDFs too.
For a CPA or law firm: "What does our engagement letter say about late fees?" across your own templates and policies.
- 03
Plain-English analytics
Ask questions of your data in plain language and build drag-and-drop dashboards.
For a CPA or law firm: "Which clients had the biggest revenue drop this quarter?"
- 04
No-code predictions
Build and schedule simple prediction models right from the screen.
For a CPA or law firm: Flag invoices likely to go past 60 days.
- 05
Data hub
Connect your ERP, Microsoft 365, databases, files and SFTP drops, then clean the data.
For a CPA or law firm: Combine exports from your practice tools into one view.
- 06
Ready-made AI agents
Install agents like apps from the built-in agent store, or build your own without code.
For a CPA or law firm: Summarize a vendor contract and flag unusual terms for a reviewer.
Examples show how firms can use each tool. We'll confirm fit for your data during the sizing session.
Why HTL 365
We install it, secure it, and retire it.
A private AI server is still a server. It needs a secure network spot, backups, updates and someone watching it. That's what we already do for Houston firms every day.
- 01 · Install
Size the server to your users and models, then set up the network, accounts and access roles on site.
- 02 · Protect
Keep it patched, back it up, watch its health and forward its audit logs to your security monitoring.
- 03 · Retire
At end of life, we wipe or destroy its drives following NIST SP 800-88 Rev. 2 and give you a Certificate of Destruction.
One local partner from first install to final drive wipe.
What's in the box
Starts with one office-size server. Scales up as you grow.
This is our entry-level system. Larger configurations are available for more users or bigger models; we size every install to your firm.
- StartQuiet desk-side system for one team
- GrowRack server for several departments
- ScalePrivate AI data center for the whole organization
- Server
- Quiet desk-side AI workstation; no server room needed
- AI power
- Professional NVIDIA GPU built for running AI models on site
- Software
- Full private AI platform: assistant, document Q&A, analytics, predictions, data hub and agent store
- AI models
- Leading open models, chosen and sized for your firm
- Security
- Encryption at rest, signed models, Secure Boot, audit logging
- Support
- On-site install, updates, security patches and support
- Hardware
- Genuine NVIDIA, Intel and Supermicro parts with full manufacturer warranty
How it works
Three steps to your first install.
- 1
Fit call (20 min)
We learn your users, data sources and compliance needs, and tell you honestly if Private AI is a fit.
- 2
Sizing session
We match the server, models and user capacity to your firm, and list any integration work.
- 3
Pilot proposal
A fixed-scope proposal with hardware, software and install costs. Early-access firms get first November install dates.
Typical time from install to your first working AI agent: 4–6 weeks.
Private AI questions
What is private AI?
Private AI is AI software that runs on an on-prem AI server you own, at your location, instead of a public cloud service. Your prompts, files and answers stay on your server.
Does it need an internet connection?
No. It can run fully air-gapped with offline updates, or with outbound-only connections so nothing can reach in. Your data never leaves the building either way.
Can it read our SharePoint and Office files?
Yes. It brings in PDFs, Office documents and SharePoint content and makes them searchable on your private server, with access controlled by role.
Is our data used to train AI models?
No. The platform runs models to answer your team's questions. Your files are not sent anywhere to train outside models.
How many people can use it at once?
It depends on the models you run and the hardware. We size it for your firm in the sizing session rather than quote a single number.
Does it work with our Microsoft 365 sign-in?
Not yet. Today it uses its own user accounts with role-based access. Microsoft Entra ID sign-in can be scoped as a custom project.
Can it fine-tune a model on our files?
Not on the entry system. It answers from your documents by searching them, which keeps your files as the source of truth. Larger systems add model fine-tuning.
When can we get one?
First installs begin November 2026. Join early access to get on the schedule.
Early access
Get on the November install list.
Tell us a little about your firm. We'll reach out to book a fit call and hold you a spot in the first install window. Early-access firms get first scheduling and a sizing session at no cost.
Prefer to talk now? Call 832-582-1668.