State agencies don't get to start over. The mandate keeps arriving, the case files keep landing, and the systems underneath were specified in a different decade. We build the connective tissue and the modern replacements that let health and human services departments meet their obligations without burning out the people who carry the work.
Public-sector operations have a specific texture. A system of record from the early 2000s, still authoritative, still the source of truth for funding and compliance. A second system bolted on for a federal reporting requirement. A third for grants. A fourth that a vendor sunset two years ago but nobody can decommission, because one report still runs on it.
Around all of that, people. Caseworkers, analysts, program managers, contract staff, copying numbers from one screen into a spreadsheet, reconciling identifiers between systems that don't agree, assembling the same monthly report from the same six sources every month. Each handoff is a chance for an error, and every error has to be traceable, because everything here is audited.
The work we take on isn't really the legacy systems themselves. It's the manual layer that has grown up around them to make them usable, and the reporting burden that layer creates downstream.
We meet agencies where they are. The first move is almost never replacement; it's removing the manual grunt work and connecting the systems that already exist, so the people in the building can breathe.
Automate the manual layer. The spreadsheet reconciliations, the recurring exports, the copy-paste pipelines between case management and reporting. We automate them with full audit trails, so what used to be a person's afternoon becomes a logged, repeatable step.
Connect what's already there. Most agencies don't need another system. They need the ones they already have to talk to each other. We build the integrations between case management, finance, grants, and reporting, quietly, in the background, without forcing a migration.
Replace the core when it's time. When a legacy system of record finally has to go, we build the successor for the AI era, with modern data infrastructure and proper auditability built in from day one instead of bolted on.
AI agents with a human expert in the loop. For the work that requires judgment, like eligibility review, anomaly flagging, and narrative summarization of case notes, we build agents tuned to the agency's own policies and procedures, with a human expert in the loop on every consequential decision. The agent is shaped to the rules it operates under and accountable for the work it does inside them.
Co-design with the people who do the work. Caseworkers and program staff know where the friction is. We sit with them, map it, and design the future state together rather than handing them a system from above. See also Think-aloud AI.
In a public health program, a principal investigator pulls numbers out of one system, drops them into Excel, cross-checks against a second system, and rebuilds the same compliance report month after month. Every cycle is a fresh chance to introduce a transcription error into a document that the federal government is going to read.
That was the Nebraska Department of Health, on the PIE Project.
We took the PIE workflow that lived across spreadsheets and disconnected systems and turned it into a single connected pipeline. The source systems feed the workspace directly, the reconciliations run automatically with the audit trail preserved, and the principal investigator now spends time on the judgment calls that require a human expert, instead of on data movement. The legacy systems stayed in place; the manual layer around them did not.
The recipe is the same elsewhere. We start with the grunt work, connect what already exists, and only replace the core once the agency is ready.
Share the spreadsheet. We'll see if it can be a system.