Put AI to Work on the Documents You Cannot Send Anywhere
De-Cloud AI drafts, summarizes, and answers questions from the files your organization already holds. The model runs on a server in your building, so nothing your staff types reaches OpenAI, Google, or any outside AI company.
- Drafting, summaries, and answers from your own files
- Runs on a server your organization owns
- No per-query fees, no specialist hire
- The AI model itself
- Every prompt your staff send
- Every answer the AI generates
- Sign-in and account data
- Chat history for your staff
- Encrypted, isolated per organization
40+ trusted community partners
If This Is Your Work, This Was Built for You
Two kinds of organization end up here.
Organizations that cannot
Bound by law, licensing, or a duty of care they owe their clients. These files cannot leave the building.
- Immigration legal services
Asylum declarations and client histories that cannot be re-disclosed.
- Health and social service providers
Intake assessments and treatment notes covered by HIPAA.
- Criminal justice and reentry programs
Sealed records, court filings, and supervision conditions.
- Domestic violence and survivor services
Safety plans and location detail that must never surface.
- Youth and family services
Records on minors, custody filings, and mandated-reporting notes.
Organizations that will not
Nothing stops them legally. They will not route community data through companies whose business model they organize against.
- Environmental and climate justice groups
Campaigning against the data centers this would otherwise run in.
- Data justice and digital rights advocates
Cannot argue for data rights while feeding someone else's scraper.
- Community land trusts and co-ops
Collective ownership of the land, and of the tools.
- Faith-based organizations
Stewardship that extends to what congregants share in confidence.
- Nonprofits serving marginalized communities
Unwilling to fund the companies profiting from those same communities.
Not on this list? If your work involves records you cannot hand to an outside company, the answer is the same. Book a demo.
Your Staff Are Probably Already Using AI
Not because anyone ignored policy. The work is heavy, the tools are free, and a browser tab is always open. It happens on personal accounts and personal devices, where you cannot see it.
Banning it does not stop it. It only makes it invisible, so you cannot say what left or answer for it later.
De-Cloud AI does not ask anyone to stop. It gives them the same help, sanctioned, on hardware you own.
Ask your team, with no consequences attached, whether they have pasted client detail into a chatbot. The answer is your business case.
Personal accounts are not business accounts
This happens on free, personal accounts, where what your staff type is used to help train the AI company's system by default. Turning that off is a setting almost nobody changes.
The question you cannot answer
A funder, an auditor, or a client's attorney asks where that record has been processed. There is no log, no account you control, and no honest answer available.
A promise you cannot verify
You told a client their information stays inside your organization. With a chat log on a personal account, you cannot confirm that it did.
Three requests, three destinations, and no record on your side that any of it happened.
The Documents You Can Finally Use AI On
Drafting and summarizing is not the hard part. Every AI tool does that. The question is which files you are allowed to point one at.
- Redacted summaries, once someone has done the redacting
- Program descriptions already published on your website
- Generic policy questions with every client detail stripped out
- Client case files, unredacted
- Intake forms and assessments
- Behavioral and physical health records
- Immigration status documentation
- Court records, police reports, and legal correspondence
- Personnel files and board minutes
- Donor records and financial detail
What People Actually Ask It To Do
No setup and no training. Staff type what they need, the same way they would ask a colleague down the hall.
Ask your own documents
Answers drawn from your policies, procedures, and prior work, with a pointer back to the document they came from.
“What does our retention policy say about case files after a client exits?”
“Which of our past reports mention youth mental health outcomes?”
Draft from what you already have
First drafts built on your own past submissions and program data, in the voice your organization already uses.
“Draft a two-page progress report for this funder using our last three quarterly updates.”
“Write a reply to this client question using our intake policy.”
Summarize and condense
Long, messy source material turned into something a colleague can read in a minute and act on.
“Summarize these five intake notes into one paragraph a colleague can pick up cold.”
“Turn this sixty-page funder guideline into a plain-language briefing.”
Every AI Option Asks You to Compromise. This One Does Not.
Two things stop mission-driven organizations from using AI on the work that matters most.
You cannot
Client case files, health records, and immigration status cannot go to an outside AI company. The exposure is not worth what the tool gives back.
You will not
Others refuse on principle. They will not feed big tech's data harvesting, the energy its data centers burn, or AI trained on work taken without consent.
Either way, hours that could go to mission work stay locked in manual effort.
So we moved the AI, not your data.
The AI itself runs inside your organization, on hardware you own. Every question and every answer is processed there, never reaching OpenAI, Anthropic, or Google, and never training anyone else's AI.
Your staff still get a familiar chat interface, linked to that server by a private, encrypted connection. Nothing about the day-to-day feels different.
A reliable product, no web servers for you to run, and the substance of the work never leaves your building.
Follow One Question, Start to Finish
Here is where a question goes after someone hits enter, and who can see it at each step.
- 01
A staff member asks a question
A familiar chat interface, hosted like any other secure business application.
Cloud - 02
It travels to your server
Encrypted the whole way, straight to the server in your building.
In transit - 03
Your own AI answers it
The AI reads the question and writes the answer on your hardware. The software is open-source, meaning it is free to use and nobody bills you per question.
Your walls - 04
Only the transcript is kept
The answer returns to the browser. No outside AI provider ever saw either half.
Cloud
Everywhere else, that is exactly what gets sent. The whole case, written out by your own staff, sitting on a server in someone else's building, under someone else's rules.
The controls behind this are written up in full on our security and privacy pages.
From First Call to AI Running in Your Building
Five phases, one on-site visit, and a team that stays after the install.
From the first call to staff using it daily
On site, with an engineer in your building
Per-question fees, no matter the volume
- 01
We scope it with you
Week 1
- 02
Your server is built and shipped
Weeks 2 to 3
- 03
We install it on site
One day, on site
- 04
Your staff are trained on it
Weeks 4 to 5
- 05
We keep it current
Ongoing
One machine, in a room you already have
- A lockable space, power, and a network drop. Nothing exotic.
- On your asset register from the day it arrives.
- Owned outright. No lease, and no contract that takes it away.
- No rack, no server room, and no data center anywhere in the picture.
How This Compares
We serve the organizations commercial providers can't reach and won't serve.
| Your options | Data stays in your building | Staff get real AI help | No specialist staff to hire | Priced for a nonprofit |
|---|---|---|---|---|
Commercial AI ChatGPT, Gemini, and the rest | No | Yes | Yes | Yes |
Custom enterprise AI Built for you by a consultancy | Yes | Yes | No | No |
No AI at all Opting out on values grounds | Yes | No | Yes | Yes |
De-Cloud AI The model on hardware you own | Yes | Yes | Yes | Yes |
Green claims from commercial providers change nothing in the first column. The processing still happens in someone else's data center, on someone else's terms.
To make AI something organizations own, not something they rent.
The organizations doing the hardest work in our communities should not have to choose between modern tools and the trust the people they serve place in them. That trade has been the price of admission to AI, and we think it is the wrong one.
So we build AI that runs on hardware you own and answers to your policies. Nothing leaves the building, no per-query bill arrives, and no outside company learns from the work you do.
What Organizations Ask Us First
The concerns that come up in every evaluation, answered plainly.
Does any of our data reach OpenAI, Google, or another AI company?
No. The AI model runs on hardware your organization owns. Every prompt your staff send and every answer the model generates is processed on that machine and never transmitted to an outside AI provider, and never used to train someone else's model.
Do we need to hire AI engineers to run this?
No. De-Cloud AI is set up for you, start to finish. We handle installation, configuration, updates, and monitoring. Your IT staff need no special AI expertise to keep it running.
What does the hardware cost, and what does it replace?
You invest once in a server rather than paying an outside provider for every question your staff ask. The AI software itself is open-source, which means it is free to use and nobody charges you per question. We size the hardware to your team and walk through the numbers before you commit.
Is free, open-source AI good enough for real work?
For the work most organizations want help with, yes: drafting, summarizing, answering questions against your own documents, and cleaning up writing. We will show you the model handling your actual documents during the demo so you can judge the quality yourself.
See what AI looks like when the data never leaves.
We will walk through your data constraints, what a deployment inside your organization looks like, and whether De-Cloud AI is the right fit.