We wrote what we call our AI Constitution before we shipped a single AI feature or deployed a single internal AI workflow. It isn't a governance policy buried in a legal document. It's the framework every AI decision at Statewise runs through, whether that's a product feature, an internal operation, how we handle data, or how we develop software.
The Stakes Are Higher in Healthcare
Healthcare technology companies face a unique challenge in 2026. URAC's new AI accreditation standards, CMS's evolving guidance on AI in Medicare systems, and emerging state-level AI regulations are reshaping how we think about artificial intelligence in care delivery. The regulatory landscape moves fast, but the core question doesn't change: how do you harness AI's capabilities without compromising the trust that healthcare relationships depend on?
That question became urgent for us faster than we expected. Within six weeks of giving our engineering team access to AI-assisted coding tools, every developer was doing AI pair programming. Velocity went up significantly. But watching that happen inside our own walls, building software that manages clinical documentation and Medicaid billing for vulnerable populations, made it clear that speed without a framework is a liability. That's what pushed us to write the Constitution before we shipped a single AI feature.
Our Governance Framework
Our Constitution has six principles. They aren't aspirational values and they aren't limited to product features. Every AI decision we make, whether it's a new feature, an internal workflow, how we handle data, or how we develop software, runs through this framework.
The first principle is that AI drafts, flags, surfaces, and suggests, but it doesn't get the final say on anything that matters. Closely tied to that is the idea that AI takes load off our team rather than replacing people with hard-won experience. AI only works well when guided by people who know when it's right and when it's wrong. Those two principles shape every conversation we have, from how our support team drafts responses to how our engineers use AI-assisted coding tools.
The third principle is one we don't hear other vendors talk about much: AI runs on a short leash by default. If it needs more access, it asks and explains why. We decide what it can do before it runs, not after something breaks. That applies internally just as much as it does in the product.
Explainability is non-negotiable. If we can't describe how something works in plain English, it isn't ready, whether that's a shipped feature or an internal automation we're running ourselves.
PHI gets its own rule because it deserves one. Patient data stays in our infrastructure. No PHI trains outside models. If a client asks where their data goes, we can give them a direct answer, not a terms of service reference.
Finally, we don't outsource AI governance to our vendors. Our controls live in our code, our policies, and our audits. When someone asks how we govern AI, we show them our framework.
The most critical application of these principles is found in our first pillar: AI will not run billing rules or make claims decisions.
This isn't a temporary limitation. It's a permanent principle. Medicaid billing requires 100% deterministic accuracy. A single error can trigger audits, recoupments, or compliance violations that threaten an agency's ability to serve clients. The reliability simply isn't there yet for AI to make these decisions autonomously.
Instead, we use AI to help the people who build and maintain our rules engine work faster and make fewer mistakes. AI assists our engineers in writing cleaner code and supports our product team in understanding complex regulatory changes. But when it comes to determining whether a service qualifies for reimbursement, our deterministic rules engine makes those calls. Anyone claiming their AI can autonomously handle Medicaid billing is getting ahead of the evidence. We've chosen honesty over hype.
Rollout With Purpose
Rolling out AI tools and telling people to figure it out is how you get inconsistent usage, bad habits, and eventually a real problem. We took a different approach.
We designated AI Champions in each department: engineering, QA, support, product, and customer success. Each champion is trained on our AI Constitution and is responsible for understanding how AI applies to their team's workflows, where it's appropriate, and where it isn't. They're the ones who catch it when someone is about to use AI in a way that cuts against our framework.
The result is that AI decisions don't get made in isolation. Our support team uses AI to draft responses to complex technical questions, but every one gets reviewed and personalized before it goes out. Our engineers use AI-assisted coding tools, but they own the output. The Constitution gives every champion a consistent standard to hold, regardless of department.
A People Business First
Our CEO, Cole Ballweg, captured our philosophy perfectly: "This is a people business. AI should take great people and make them even greater."
When I kicked off our company-wide AI training, the framing I used was this: AI won't replace your job, but the person who knows how to use it will. That's not a threat, it's the opportunity. We want every person at Statewise to have a new skill in their toolkit, one that takes the repetitive and time-consuming work off their plate so their judgment, experience, and relationships can go where they matter most.
That's the version of AI adoption we're building toward. Not fewer people, better equipped people.
The Trust Advantage
When agencies ask how we handle AI, we don't point them to a vendor's white paper. We walk them through our Constitution. That's a different kind of conversation than most software vendors are having right now.
What we've found is that sharing how we think about AI has prompted our clients to ask the same questions about their own organizations. Many of them are running their own AI initiatives, and our platform is a core part of their operation. Being transparent about our framework gives them something concrete to react to and build from. We're not requiring them to adopt our Constitution or follow our rules. But we are in a position to support them as they develop their own.
The Constitution has slowed some feature releases. That's a real cost we've accepted. But it's opened client conversations we wouldn't have had otherwise, and it's given us something most vendors can't offer: a clear, specific answer to how we govern AI across our entire operation.
One thing we'd say to any healthcare organization thinking about AI: policies matter, but only if they're real. Don't bury them in a legal document nobody reads. Write them in plain language, pull them out, hang them on the wall, and make them part of every AI decision and idea your team has. A policy that lives in a drawer doesn't protect anyone.
