Case StudyAffordable Housing Owner-OperatorSigned CA
How Sunrise Affordable Housing cleared its data-privacy bar before deploying an AI employee
The question that decided this engagement was not what the AI could do. It was whether firm information would ever be used for anyone else. The answer was no, in writing, and the build went live in June.
Isolated
0
Under NDA
The Challenge
What was actually in the way.
Sunrise Affordable Housing is an owner-operator that would not put firm information into a shared AI tool. Affordable housing carries tenant and financing detail that does not belong in a system the operator cannot see the boundaries of.
Before scope, before capability, the firm asked the question directly: would its information be used for other clients of ours? Most AI conversations skip past that. This one did not proceed until it was answered.
The Solution
What PSV built.
PSV built a custom AI employee on a separate hosted instance under NDA. Isolation is a property of the deployment rather than a setting inside a shared product, so the guarantee does not depend on anyone remembering to toggle it.
Nothing is copied out of the instance, and no client data is used to train a model or to improve any other client's system.
With the privacy question settled first, the build itself was straightforward. The engagement went live on 3 June 2026 and was handed to the client's team the same month.
01
Custom AI employee
Built for this firm's work rather than configured out of a general-purpose tool, so it fits an affordable housing owner-operator's actual process.
02
Per-client data isolation
One client, one instance. Nothing crosses between engagements, and no client data trains anything.
03
Private hosting
The instance is hosted separately and under agreement, so the privacy commitment is architectural rather than contractual only.
The Impact
Zero client data used to train anything. That was the whole question.
The firm got an AI employee it could actually use, because the deployment matched the standard its data already had to meet.
Live since June and running in the team's own hands rather than as a PSV-operated service.
The full scope stays under NDA at the client's request. We publish what we were cleared to publish and leave the rest alone.
Isolated
Separate hosted instance
The build runs on its own instance rather than a shared tool, so there is no path by which one client's material reaches another.
0
Client data used for training
Nothing the firm puts in trains a model, and nothing is copied out to improve anyone else's system.
Under NDA
Full scope protected
The engagement runs under a signed agreement. What the AI employee does day to day stays with the client.
The system behind it
AI Employee, built for one firm at a time.
Nothing here is a product with a login. Every build starts from the firm’s own models, templates, and criteria, which is why the output lands in a format the team already trusts.
See how we build firm systemsMore case studies
Keep reading.
AI Underwriting Analyst
How Westlake Realty Group cut per-deal underwriting from 6-8 hours to 20 minutes
18-24x Faster per-deal underwriting
Four AI Employees
How Accomplish Collective put four AI employees across its deal cycle
4 AI employees across the deal cycle
Deal Origination & Market Intelligence
How a team at Cushman & Wakefield expanded origination with zero added hires
0 Hires added to expand origination
AI for CRE · Weekly newsletter
The PSV newsletter, read by CRE operators.
Workflow teardowns, vendor takes, and field notes from custom AI builds. Free. Drop your email and the next issue lands in your inbox.
No spam. Unsubscribe in one click.
Let's get to work.
Thirty minutes. We'll tell you whether PSV is the right partner for what you're trying to build.
Or reach us directly at zed@pacificsoftwareventures.com