Entity type

Nonprofit, VA-accredited veterans advocacy (initial claims through appeals)

What changed

Structured medical-record AI (Superinsight) plus internal tools and LLM-assisted narrative drafting

Stated headline outcome

Leadership attributed a large annual labor-cost reallocation to the adoption of Superinsight

Credibility bar

Self-reported; read the caveats before you model your budget on theirs

Executive summary

This organization operates at a scale more often associated with commercial claims shops than with volunteer-heavy models. Leadership described a multi-track docket: initial claims, higher-level review, supplementals, and parallel pathways that can all be live for the same veteran. That structure creates relentless demand for accurate medical chronology and for writing that ties facts to legal standards.

The nonprofit’s answer was not a single tool but a stack: Superinsight on the record side, internal systems for service and occupational context, and consumer-style LLMs for certain drafting tasks where counsel still controls inputs and edits outputs. The controversial headline in this account was economic: leadership said that adopting Superinsight was associated with roughly $300,000 per year that no longer had to fund the same clinical review staffing model.

This case study explains how they described using the stack, why that cost story might be true for them, and why your firm should still run its own math.

Operating model: intake, records, and narrative

Why records sit at the center

In federal veterans adjudication, the file is the spine of the argument. Leadership emphasized a line that general audiences miss: if it is not written in the file, it does not exist for legal purposes. That standard pushes organizations to invest in anything that makes writing grounded faster, not writing that merely sounds fluent.

How Superinsight fit the stack

Superinsight reports and summaries are combined with service history, military occupational specialty context, and veteran statements to build a snapshot of medical and service history. That snapshot becomes the shared object around which advocates, veterans, and downstream drafting tools align.

Why narrative still matters

Leadership described adjudicators as responsive to coherent stories supported by evidence, not to exhibits alone. That is why the organization did not treat the LLM layer as a replacement for Superinsight. The record layer and the language layer solve different problems.

Challenge in depth

• Labor intensity: High-accuracy medical review historically required clinical and mid-level professional time that does not scale cheaply.
• Throughput pressure: Multiple simultaneous claim tracks per veteran increase the risk of missed exhibits or inconsistent theory across filings.
• Ethical and reputational risk: At nonprofit scale, word of mouth dominates growth. A pattern of shallow reviews would compound faster than at a small firm.

Numbers and quantities in this profile

Topic	As described	Limitation
Annual cost figure	Leadership said Superinsight saved about $300,000 a year and that they replaced three RNs and two nurse practitioners in that shift	Self-reported; no audited financials supplied here
Veterans represented	Leadership said they represent over 6,000 veterans and, at any given time, about 12,000 to 15,000 open “cases” because one veteran can have multiple parallel claim tracks	Dynamic counts; not an audited census
Outcomes (context)	Public materials also reference very large cumulative benefits secured for veterans (order of billions over the life of the organization)	Treat as organizational storytelling, not your forecast
Workflow	Superinsight combined with service history, MOS context, and veteran statements for snapshots	Process description as reported by leadership

How to interpret the $300K claim responsibly

Even if the number is directionally right for that organization, your shop should decompose it before budgeting:

• Loaded vs. unloaded wages: Did the estimate include benefits, overtime, supervision, or contractor markup?
• Avoided hires vs. reduced hours: Did the organization stop growing clinical headcount, or shrink existing roles?
• Substitution risk: If AI output is wrong once at the wrong time, reputational cost can swamp spreadsheet savings.

The defensible learning for peers is simpler: they believed the record layer was worth buying rather than staffing indefinitely at the margin, and they said it loudly enough that others should ask the same question with their own numbers.

Takeaways: how Superinsight helped (per their account)

• Cost structure: A stated shift of roughly $300,000 per year away from prior clinical review spend toward software-assisted workflows.
• Time to structured evidence: Faster path from raw agency and private records to hearing-ready snapshots that anchor narrative work.
• Insight extraction: Better odds that the right facts land in the written product that adjudicators actually read.