A HIPAA-Compliant Workflow Automation Platform for a US Nonprofit
How developing an automation platform helped a Washington State nonprofit cut FCS claim-to-payment time in half, bring denials below 5%, and open a second funding stream by scaling to peer providers statewide.
Business challenge
In the United States, the amount of time social workers spend on paperwork is as high as 4.3 hours a day*. Our client, Anything Helps, helping people in crisis find housing and work, was no exception.
*According to OPRE
Their case managers were spending the equivalent of two of every five workdays on paperwork for Foundational Community Supports (FCS). The process was entirely manual, making errors difficult to avoid and putting reimbursement at risk.
Buried under more forms than anyone could process carefully, case managers inevitably made mistakes: a wrong code here, a missed field there. But in Medicaid billing, small slips lead to costly consequences downstream:
- Uncompensated work. If a delivered FCS service is recorded incorrectly or not recorded at all, the nonprofit may never be reimbursed for it.
- Cash-flow delays. Faulty claims bounce back for correction and resubmission, stretching the wait between service and payment well past a month and putting pressure on the cash flow predictability a community nonprofit runs on.
The lack of automation in itself posed problems beyond financial ones:
- A ceiling on reach. Administrative work grew with every additional participant served, so helping more people also meant adding more administrative headcount.
- Staff burnout and turnover. Case managers joined the organization to work with people, not spend 10+ hours a week on forms. Over time, that burden risked wearing down experienced staff and driving them away.
Anything Helps concluded that the answer wasn’t adding more hands to the paperwork, but removing as much paperwork as possible. That’s why, they decided to develop a platform to automate the FCS lifecycle from enrollment through payment.
Moreover, knowing the struggles of 250+ similar WA-based FCS non-profits better than anyone, they had an ambitious idea of opening this platform to every FCS provider in Washington who wrestles with the identical bureaucracy: checking eligibility through the same state system, submitting documentation to the same administrator, and moving claims through the same reimbursement process.
A platform built to solve Anything Helps’ own operational bottlenecks could therefore become a product for the wider provider community. Offered on a subscription basis, it could turn the nonprofit’s hard-won FCS expertise into a second source of funding.
To build it right from the start, Anything Helps turned to Instinctools.
Solution
Instinctools built Pathfinders, a HIPAA-compliant platform to handle the full Foundational Community Supports workflow in one place:
- Enrolling applicants
- Checking their Medicaid eligibility
- Securing authorization
- Documenting the services delivered
- Billing Medicaid through to reconciliation
The platform works across web, iOS, and Android from a single Flutter codebase. It was initially developed for Anything Helps’ own team, but after Pathfinders proved successful, the software became available to other FCS nonprofits in Washington state.
Accelerating delivery with a governed AI pod
Anything Helps needed the platform fast and within a fixed budget. To make that possible, Instinctools’ team formed an AI pod – a small, senior-led team accountable for the outcome, working alongside governed AI agents embedded directly in the software development lifecycle.
For this project, having a senior project manager and full-stack engineer was enough to set specifications, make architectural calls, and supervise AI agents handling much of the development. This setup allowed a lean team to cover work that would normally require a much larger delivery unit.
But speed only creates value if quality keeps pace. That is why the AI pod relies on governance rather than trust in generated output. AI-driven delivery here is not a matter of reaching for a few AI tools whenever they seem useful. It’s a controlled framework wired into how the software gets built.
Three safeguards kept the process reliable:
Test and trial builds are designed so that they have no access to the live Medicaid or Wellpoint systems.
Automating the paperwork pipeline end-to-end
The manual version of FCS billing was a relay of manual handoffs, and one dropped baton delay (or derail) reimbursement for services already delivered. We created Pathfinders to turn that series of fragmented tasks into one connected flow, with no need to re-enter data at any stage.
Guided enrollment
Our team replaced scattered intake paperwork with one guided form that walks a case manager through the first meeting:
- Presenting the information-sharing consent agreement
- Gathering the details each program needs, from the applicant’s name and date of birth to a Social Security number or Medicaid ID
Capturing the two digital signatures the program requires (the applicant’s, or their authorized representative’s, and the manager’s)
- Routing the applicant into Supportive Housing or Supported Employment
Two under-the-hood decisions made enrollment easier to manage from the start:
- A record is created early in the process, so an unfinished enrollment can be saved and resumed later instead of started over.
- Age eligibility check is built into the form itself: the system flags anyone under sixteen on the day of registration, removing the need for a manual age check. Even a rejected enrollment remains in the system for a complete audit trail.
Advanced automatic Medicaid eligibility verification
Running an age check was quick. Medicaid eligibility was another story. It used to pull a case manager out of the workflow: sign into the state system, look the applicant up, then copy the answer back by hand. Our team wired that lookup into enrollment, so the platform queries Medicaid electronically through the clearinghouse and returns a status in minutes.
We also made sure the process covered what happens after the first check:
- Adding a daily re-check across every active client to catch changes in coverage
- Designing the flow to keep a “not found” result from blocking enrollment. The applicant stays in the system and is flagged for a case manager to investigate further and follow up.
FCS authorization and the Wellpoint bridge
This part was the trickiest to engineer because of the system on the other side. Every FCS service must first be authorized by Wellpoint, the program administrator. But Wellpoint offers no API for third-party applications, leaving providers with a web form that case managers are expected to fill out by hand.
The engineering challenge was compounded by HIPAA-driven strict privacy requirements: real applicant data could not be used in testing, including names, dates of birth, Social Security numbers, or Medicaid IDs.
We therefore engineered Anything Helps’ side of the exchange with Wellpoint in two directions:
- On the way out, Pathfinders compiles the applicant’s assessment and prepares it for submission.
- On the way back, Pathfinders reads Wellpoint’s returned decisions and converts each row into a structured record the rest of the workflow can use.
Our team created a processor that reads Wellpoint’s files and turns every row into a structured record the rest of the workflow can act on, covering the full range of replies Wellpoint sends:
- Authorizations
- Denials
- Rejections
- Waitlist placements
- Exception-to-Rule determinations
- Provider Change Request outcomes
From service note to paid claim
A service can be reimbursed only if the documentation behind it is complete and accurate. Each time a case worker delivers one, they log it with a short note that starts as a draft and is finalized after a review step. A duplicate-detection mechanism flags anything entered twice for the same person, date, and program.
Writing the narrative is one of the most time-consuming parts. That’s why, we used AI to remove much of the manual effort. A case worker dictates the details, and the AI returns a structured first draft in the format FCS requires. Instead of writing the note from scratch, the employee reviews and finalizes an almost-complete record.
Potential denials caught early
Waiting for Medicaid to reject a faulty claim can add weeks to the reimbursement cycle. To prevent avoidable denials from entering that cycle in the first place, we built a validation gate that tests every service against Medicaid’s reimbursement criteria and holds anything likely to be denied.
without the gate
A service that passes all checks is automatically sent to Medicaid in the required electronic format.
Still, there was a question of what happens to the claim after it leaves the platform. Instinctools’ team added a tracking feature, so Pathfinders reads acknowledgments and payment files as they return through the clearinghouse, updates the claim status, and matches each payment to the service it settles. What used to be a month of manual chasing becomes a claim status a case manager can check in seconds.
Built once, running everywhere
Anything Helps needed Pathfinders to work across web, iOS, and Android, as their staff operate both in the field and at a desk. Developing a native app for each platform would have meant three codebases to write, test, and keep in sync. On a fixed budget, that matters even more: every hour spent maintaining parallel versions is an hour not spent on features that enable case managers to help more applicants.
Therefore, we opted for cross-platform development on Flutter:
- One codebase serves all three platforms, directing more of the budget toward product functionality rather than parallel maintenance.
- A consistent experience carries across devices, so workflows behave predictably whether a case manager is working from a phone or an administrator is using the web app.
- Features can be built and tested once rather than three times, which was especially important for an AI pod working against an aggressive delivery timeline.
Designing a user-friendly interface across web, iOS, and Android
Our AI pod used Claude Design for the interface to move in step with the code.
This let us explore screens early, validate them against the actual FCS workflow, and shape the experience around how case managers move from enrollment through authorization, service documentation, and billing. At the same time, working from a shared design direction helped keep the interface consistent across web, iOS, and Android.
AI accelerated the design process, but it did not replace review. Senior experts checked every screen against the data model and the specifications prior to release, catching and reconciling the gaps before they reached production.
Turning Pathfinders into a second funding stream
Years of navigating FCS requirements had given Anything Helps something valuable beyond process knowledge: deep expertise in a workflow that hundreds of peer organizations struggle with as well.
Roughly 300 FCS organizations across Washington operate within much of the same administrative framework.
That meant Pathfinders could be extended to other FCS organizations without rebuilding the core workflow for each one.
Where providers differ (for example, in reimbursement rates), configuration handles the variation without changing the underlying product. Provider-level data isolation also keeps each organization’s records separated from every other tenant.
In the end, a tool built to automate operations for one nonprofit evolved into a subscription product for peer FCS providers, giving Anything Helps a second source of funding and turning its operational know-how into an asset other nonprofits can lean on.
Before / After
The client interview still takes human time, but it’s a single guided in-app form
The platform fires automatically on form submit; the answer comes back in minutes
Pathfinders prepares the assessment and tracks approval; a case worker just clicks ‘Submit’
Staff dictate the note and review an AI draft instead of writing from scratch
The platform checks each claim first and submits only the ones that will pass; the rest are held back for a fix and a later retry.
Pathfinders read payment reconciliations and auto-match them to claims; humans handle only the few flagged denials
Business value
- The entire FCS workflow in one platform, from enrollment and eligibility checks through billing and payment reconciliation
- An additional source of funding through subscriptions from peer FCS providers
- 3 to 6 fewer hours of paperwork per week for each case manager
- 2x faster claim-to-payment processing
- < 5% claim denial rate, compared with 15-20% industry benchmark
Multiplier
effect
What we solved for Anything Helps is a pattern that extends far beyond FCS. Whenever a transaction has to move through several independent systems before it is complete – be it in healthcare billing, insurance claims, logistics, public-sector services, or regulated finance – the same friction tends to appear at the handoffs.
The goal is to connect those handoffs into one governed workflow: carry data forward automatically, validate it before submission, and reconcile responses when they come back. That approach reduces manual work, shortens processing cycles, and catches errors before they turn into costly exceptions.
