Typical monthly workload
200–1,000
Conditional enrollment applications a month, with higher volumes when programmes reopen.
The Assistance Fund · Nonprofit patient assistance · United States
From handwritten applications to guided review.
The Assistance Fund worked with REDA to configure AI around its enrollment procedures—bringing extracted information, confidence indicators and review guidance into the Salesforce screens its patient advocates already use.
The workload behind the review
Time-sensitive review
30 days
The response window described by TAF—making timely application review important.
Workload and response window reported during TAF discovery in April 2026. These describe the application-review workload, not AI throughput or measured time savings.
The challengeThe information was on the page.
The information was on the page.
The work was getting it into the process.
The Assistance Fund helps people facing high out-of-pocket medical costs access financial assistance for treatment-related expenses. Its patient advocates work through information that matters to each person’s application. About TAF ↗
TAF already had an established enrollment process in Salesforce Health Cloud. An application could arrive as a scanned, photographed or handwritten form. Advocates opened the document and worked through a guided sequence of screens, checking and entering the information.
Each application meant comparing submitted details with existing records and reviewing the relevant programme requirements. Different handwriting, missing information and discrepancies required attention. During discovery, TAF described typical monthly volumes of roughly 200–1,000 applications, rising to a couple of thousand around programme reopening. The team also described seasonal queues that required additional staff effort to clear.
The 30-day response window made the queue more than an administrative burden. TAF explained that late processing could mean an enrollment had to be reactivated, creating further work for the team.
The starting question: how could AI help with that preparation while keeping the advocate involved in reviewing the application and deciding what should happen next?
The approachBuild around the way
Build around the way
TAF already works.
REDA started with one defined workflow: conditional patient enrollment applications. Working with TAF’s business and technical teams, REDA mapped the application, the relevant Salesforce records and the procedures the advocates followed.
The REDA AI platform was configured to read the document, compare information across the related records and apply the agreed review steps. These included household-income criteria, insurance answers, required agreements and the programme’s covered-drug information.
The implementation also extended TAF’s existing document-processing screens. The agent’s reading appears beside the inputs the advocate was already going to complete, bringing assistance into a familiar workflow.
The advocate’s experienceRead it. Check it.
Read it. Check it.
Use it with one click.
AI prepares the reading. The advocate sees the suggested value and its confidence beside the relevant input, checks it and chooses what to use.
- 01 · DOCUMENT ARRIVES
Read the application
An attached application starts processing. AI reads the document and prepares a structured assessment.
- 02 · SUGGESTION APPEARS
See the value and confidence
The existing screens show AI-read values beside the relevant fields, with colour-coded confidence indicators.
- 03 · ADVOCATE REVIEWS
Copy with one click
The advocate checks the suggestion and copies it into the input with a single click—or enters a correction.
- 04 · PERSON DECIDES
Choose the next step
The advocate reviews the assessment and recommendation, then completes the established process.
Assistance at the field level.
The updated screens support demographics, contact details, mailing addresses and alternate-contact information. Suggested values sit where they are needed, ready for the advocate to check and use.
A recommendation for the application.
The assessment also includes an overall reading-confidence indicator, a plain-language summary and a recommended next action.
Ready to ApproveNeeds ReviewCannot Process
Confidence indicators support the advocate’s review; they are not a guarantee of accuracy. “Ready to Approve” is a recommendation—the configured Phase 1 agent does not approve enrollment.
The customer’s perspective
Assistance in a familiar workflow.
“What mattered to us was giving our Patient Advocates useful assistance in a process they already know. With REDA AI, they can review what the AI has read from a document, evaluate the accuracy, and copy appropriate values into their workflow without additional work—all while keeping the decision in their hands.”
John D’AngeloSVP of Business Intelligence · The Assistance Fund
What changes for the teamFrom repeated entry
From repeated entry
to prepared information.
| Manual preparation | AI-assisted review |
|---|---|
| Read the form and re-enter information into the workflow. | Review an AI-read value and its confidence beside the input, then copy it with one click. |
| Compare submitted details with existing Salesforce records. | Start with proposed comparisons and investigate differences. |
| Work through the relevant programme checks. | Review the prepared assessment and its recommended next step. |
| Identify uncertain or missing information while reading. | Use visible indicators to focus attention, while still checking the source. |
The phased adoption journeyWhat’s next:
What’s next:
extend autonomy in stages.
Phase 1 is the implementation described in this story. The three-phase scope also sets out planned extensions: more consolidated review, followed by routine automation. Phases 2 and 3 are not part of the implemented starting point.
- Phase 1 · Assisted processingImplemented · Phase 1
AI prepares.
The advocate decides.AI reads and preparesAdvocate reviews and completesRead the application, prepare an assessment and display suggested values with confidence indicators.
The advocate checks the reading, copies the values they choose and completes the existing process.
- Phase 2 · Semi-automationPlanned · Not yet implemented
AI stages the work.
The advocate approves.AI prepares proposed updatesAdvocate approves or changesPrepare record updates and a recommended decision for qualifying cases, with a more consolidated review experience.
The advocate retains the final decision and can accept or change the proposed outcome.
- Phase 3 · Routine automationPlanned · Not yet implemented
Routine cases progress.
Oversight continues.AI applies configured checksQualifying cases progressExceptions go to peopleComplete qualifying routine enrollments automatically within a defined set of checks.
Exceptions and sampled cases remain available for human oversight, informed by evidence from earlier phases.
Each next phase is evaluated and separately agreed. Human oversight continues as more routine work is delegated.
One implementation. An extensible platform.Start with one process.
Start with one process.
Build from what you learn.
This implementation shows how REDA AI can be shaped around an existing Health Cloud process, including its data, business rules and review experience.
The same platform provides a starting point for considering further workflows, each with its own scope, configuration and testing. The customer’s process comes first. REDA AI is configured to support it.
Explore the REDA AI platform →Your process could be nextWhat is the one problem
What is the one problem
holding your team back?
Tell us where repetitive work is slowing your Salesforce team. We’ll discuss a bounded first-agent proof of concept, agree what success would look like and build it in your sandbox for your team to evaluate.
Tell us your one problem← Explore customer storiesThe Assistance Fund · Customer story