Secure AI · the pillar

Private, on-premise AI for regulated firms — your data stays on hardware you own.

Built, installed, and maintained in your office. Owned by you. No per-seat fees, no cloud.

On-premiseYou own itThe 14-day guarantee

The problem

Staff using ChatGPT with client data is a compliance event.

It doesn't announce itself. A note pasted into a browser tab is a disclosure to a third party under HIPAA, a confidentiality question under Rule 1.6, and an unlogged transfer of non-public information under GLBA. The firm usually finds out during a risk assessment, an audit, or an exam.

The question	Consumer cloud AI	BAA / enterprise cloud	Handistack on-prem
Where client data goes	Bad: The vendor's cloud	Qualified: The vendor's cloud, contract-bound	Good: Stays on hardware you own
Trained on your content	Bad: Yes, unless you opt out	Qualified: No — by contract	Good: Never. No upstream to train
Retention of prompts & files	Bad: Vendor policy, human review possible	Qualified: Configurable, vendor-controlled	Good: Your policy, your disks
BAA / DPA in place	Bad: Not available	Good: Available	Good: Not needed — no third party
Audit trail you control	Bad: None	Qualified: Vendor console	Good: Your log store and SIEM
Cost shape	Qualified: $20–30 per seat / mo	Bad: $60+ per seat / mo	Good: One-time build, one flat retainer
Works with the internet down	Bad: No	Bad: No	Good: Yes

The solution

A private AI trained on your files, used through a simple chat window.

Your staff don't need to learn a platform. They ask a question; the answer comes back with the document it came from. The difference is where the thinking happens.

Ask your own files

Every document the firm already holds, searchable in plain language, with the source named in the answer.

Draft from your own work

Letters, notes, memos and summaries drafted from your material, in your voice — not the open internet's.

Summarize what nobody has time to read

Intake packets, transcripts, credit files, 200-message email chains — reduced to a page a human can act on.

Who it's for

Four regulatory conversations, one appliance.

Law firmsYou’re the firm where privilege is the product — and ABA 512 is on your desk.See the law firms page →HealthcareYou hold PHI, and there is no BAA standing behind a consumer chatbot.See the healthcare page →Financial & AccountingYou hold client NPI, and GLBA makes where it goes your responsibility.See the financial & accounting page →Community BanksYour examiner reads your controls, because the rulebook isn’t written yet.See the community banks page →

How you own it

One price to own it. One retainer to keep it running.

The one-time price covers the hardware and the install. The system is yours from day one — if you ever part ways with us, it keeps running.

One room: $12,000 – $24,000; + $1,000–$2,500/mo
Whole floor: $32,000 – $56,000; + $2,500–$5,000/mo
Every floor: Custom; monthly quoted with the build

See the full pricing →

days, money back

Every dollar — hardware included. Signed into the build agreement, not a footnote.

Compliance controls

What ships with every system — and what it doesn't cover.

Encryption at rest

Full-disk encryption on every drive in the appliance. Keys held by you.

TLS in transit

TLS 1.3 between the chat client and the appliance. Nothing crosses your perimeter.

SSO + MFA

Binds to the identity provider you already run — Entra, Okta or Google Workspace.

Audit logging

Every prompt, document read and answer written to your log store. Retention is your policy.

Data never leaves

No outbound inference calls. Prompts, files, embeddings and outputs stay on the box you own.

Documented for your file

Control narrative, network diagram and configuration handed over as a PDF for your compliance binder.

On-premise is not, by itself, a compliance programKeeping inference in the building removes third-party-disclosure risk — the largest single item. You still need written policy, staff training and a current risk assessment. We hand over the technical control documentation; your counsel or compliance officer owns the program.

FAQ

The questions firms actually ask.

Answered here and in Answers, at length.

What is on-premise AI?

A model that runs on a server you own, inside your building. Prompts, documents and answers stay on that machine — there is no vendor cloud in the path, and no account that can be terminated.

Is a private model as good as ChatGPT?

For general knowledge, frontier cloud models are still ahead. For answering questions about your own documents — which is what regulated firms actually need — an open-weight model with a well-built retrieval pipeline is at parity, because the value sits in the files, not the weights.

What hardware does this take?

One appliance, sized to the work you want it doing. It needs a rack space or a closet shelf, a 20-amp circuit and a network port — nothing structural, and we bring everything else.

Does it work if the internet goes down?

Yes. Inference is local, so the AI keeps working when your connection doesn't — which is not true of any cloud tool.

What happens when better models come out?

We swap them in under the maintenance plan. The hardware you own runs whatever the best open-weight model is that year.

Is on-premise AI enough for compliance?

It removes the third-party-disclosure risk, which is the single largest item. It is not by itself a complete compliance program — you still need policy, training and a current risk assessment. We document the technical controls; your counsel or compliance officer owns the program.

Serving the Suncoast

Sarasota, Bradenton, Lakewood Ranch, Venice & the Suncoast.

Local means the engineer who built your system is the one who shows up. Book the free analysis and we'll tell you what a private AI is worth in your firm — or that it isn't, yet.

Book your free 30-minute analysis