Introducing AI Summaries - Get Up To Speed Quickly, Safely
A colleague is out sick and their caseload was just transferred to you. A client you haven't seen in six months just called you. You have a home visit in an hour with a family you've never met.
You know what comes next. Before you can help anyone, you're digging through the referral, case notes, the care plan, and months of contacts written by people who are not available to help. Things get missed. Too often, the client ends up telling their story all over again.
I've seen this up close. Years ago I built a case management system for the public health nurses at the County of Ventura, and catching up on a case was always part of their day.
So today I'm excited to introduce AI Summaries in UnaCasePro.
What it does
Open a client or a case and click AI Summary. UnaCasePro reads the record and gives you the key points: what happened recently, what's still open, what's coming due, and anything you should know before you talk to them.
Every point quotes the note it came from, word for word, with the date and a link. If something looks off, you're one click from the original. The summary is labeled as AI-generated and never gets saved into the record. Think of it as a really good handoff from a coworker.
Why we didn't just plug in ChatGPT
The easy way to add AI is to ship your notes off to one of the big AI companies. Plenty of software does exactly that. We didn't, because of who you are. Your teams hold some of the most sensitive information there is, and the minute case notes go to an outside AI company, your IT and compliance folks have a long list of questions.
I wanted to answer all of them with one sentence. Your case notes never leave our server.
UnaCasePro already runs on Microsoft Azure Government, and the AI runs there too, on our own server. That server has no public internet address, and a firewall rule blocks it from sending anything out to the internet. Even if it wanted to reach outside servers, it couldn't. The model is Qwen3.5, an open model we downloaded and run ourselves. It doesn't learn from your notes or remember anything between requests.
You also decide what it reads. Each agency picks which programs the AI can see, and sensitive programs like TB, HIV and substance use treatment can stay off the list entirely. Social Security and record numbers get stripped out before the AI sees anything, even the ones somebody typed into a note.
How we keep it honest
My biggest worry with AI summaries has always been that they'd be confidently wrong. A summary that invents a missed appointment or creates a false detail is worse than no summary at all.
So we made it show its work. The AI has to back every point with an exact quote from the notes. If the quote isn't really there, UnaCasePro throws the point out before you see it. A second check then removes anything the notes don't support.
We tested it on made-up clients built to trip it up: Social Security numbers buried in notes, TB notes in excluded programs, contradicting income figures, notes in Spanish. In our final round, it made zero wrong statements in 120 runs.
Is it perfect? No AI is. That's why every point carries its source. The notes are still the record.
And just so it's clear: it doesn't write notes, make decisions or run in the background. It runs when someone clicks the button, and it's off until your agency turns it on.
Come see it
I'd love to show you AI Summaries on a demo client. If your IT or compliance team wants the details, ask for our AI security brief. It covers where the AI runs, how the server is locked down, and what it can and can't see.
AI Summaries keeps UnaCasePro focused on what it does best, so your case managers can do what they do best: care for their people.