Private AI for Minnesota Counties, Cities & Agencies

That's precisely the problem private AI solves. Under the Minnesota Government Data Practices Act, government data is classified—public, private, confidential, nonpublic—and not-public data can't be disclosed to unauthorized parties, which is what sending it through a consumer AI service risks. In a private AI environment, the model runs on infrastructure dedicated to your entity, no outside AI provider processes or retains the data, and your existing data-practices controls keep applying. Your counsel and responsible authority make the determination; we provide the written technical facts they need.

Three options: (1) managed hosting on single-tenant hardware we provision and operate in Minnesota; (2) hardware installed inside your own facility, behind your network controls; (3) your entity's own government Azure tenant. In all three, documents stay in your dedicated environment, nothing trains any outside model, and storage location, access, and retention are documented in writing for your records.

We scope engagements to fit small-purchase authority rather than tell you where your threshold sits—that determination belongs to your purchasing staff and counsel, and the limits differ by entity and by whether you are buying a service or equipment. What we control is the shape of the purchase: managed deployments start at $1,500/month, and we structure pilots as fixed-price purchases with defined deliverables and a stated term, so your purchasing staff can evaluate them against whatever authority applies. We are also a Minnesota-Certified Veteran-Owned Small Business (Service-Disabled), which carries a preference on State of Minnesota solicitations, and CERT-certified for Hennepin County, Ramsey County, and City of Saint Paul supplier programs.

Those categories carry additional requirements (CJIS, HIPAA, and federal CUI frameworks) that we scope honestly rather than hand-wave: some fit a private deployment inside your own facility or government tenant with your controls applied, and some require infrastructure commitments we'd discuss openly before proposing anything. We'd rather tell you a category is out of scope than sell past a compliance boundary.

They don't take them on faith—that's the design. Document Q&A returns answers with citations to the source documents so staff verify before acting, and drafting tools produce drafts for human approval, never auto-sent output. This is the same human-in-the-loop discipline our founders run in their own production AI platform and the accountability structure we bring to our Department of Defense SBIR work.

With a short discovery call and a fixed-price pilot: we stand up a private environment, load a sample of representative (public or test) documents, and your staff evaluate real output on real work before any monthly commitment. Call (612) 208-3232 or email ryan@newbloomai.com—you'll reach a founder, not a sales team.