Search

See how BD Performing Arts saved $180K with AI. Read the case study.

A California cemetery cleared a 15-year, $21.9M reconciliation backlog with auditable agentic AI

How a 110-year-old institution resolved deferred revenue across approximately 9,600 contracts in weeks, not months, with every figure traceable to its source.

Circle_Operations_Cost_icon

Fraction

of the budget the client had planned for

Reorganized

<1

month to resolve a 15-year backlog

Circle_Compliance_icon

100%

of financial figures traceable across 200k-300k source records

The challenge

A California cemetery has served families in the region since the early 1900s, performing compassionate stewardship services in a sensitive setting where privacy, expediency, and trust are the foundation of relationships. But its century of empathetic guidance does not make this organization immune from business challenges.

Like countless enduring businesses, the combination of decades of operation and growing regulatory requirements has made solving business challenges complicated. The organization found itself facing an all-too-common financial quandary: compliance flags due to aging deferred revenue. Specifically, it had more than 100,000 data points spread across about 9,600 contracts, totaling roughly $21.9 million in deferred revenue. Because the company’s source records dating back to 1911 were never fully reconciled, the organization faced a non-trivial revenue recognition challenge.

Revenue reconciliation is never just simple bookkeeping. It’s judgment encoded as arithmetic. Deciding how much prepaid revenue has actually been earned requires untangling overlapping service types, contracts amended and re-amended over decades, and myriad records, in the case of the cemetery some more than a century old, that don’t always tell the same story. The rules that separate “earned” revenue from “still owed” revenue are subtle, and they’re exactly the kind of judgment that a spreadsheet alone can’t hold, and that a person can’t apply consistently across tens of thousands of data points.

For 15 years, every attempt that the organization made to reconcile its revenue by hand ran into the same wall: the effort was enormous, and the final result was still hard to trust. “This had been hanging over us for years,” explains Executive Director at the cemetery and mortuary. “For the first time, I felt confident we had a process we could trust, not just for this audit, but going forward.” The problem of revenue reconciliation also resisted the two obvious modern approaches: a rigid script was too brittle to survive so much real-world variation, and a naive AI approach, however impressive in a demo, couldn’t be trusted with numbers that land in a financial statement. As the Executive Director explains, “We needed an answer we could defend, not just a shiny output. Prowess gave us both, and did it in a fraction of the time and cost we’d been bracing for.”

“For the first time, I could see exactly how the AI was making decisions. It wasn’t a black box; it reflected the way I already understood the business, which made it something I could trust.”

– Executive Director, California cemetery

The solution

To help the organization implement effective revenue reconciliation, Prowess Consulting built a solution around a deliberate division of labor: an AI agent does the reading and reasoning while the system is being built; a deterministic engine does the deciding when the system runs. The quality and speed of the results were nothing short of astonishing.

With this solution, all the interpretation is worked out up front, pressure-tested with the client, and written down as fixed rules. At runtime, there is no model in the path and no guessing; the same inputs always produce the same outputs, and every number traces straight back to its source. That’s what makes an AI-assisted result auditable rather than something taken on faith.

A durable, reusable solution

Two design choices make the solution durable, not just correct:

  1. A natural-language rulebook: The engine’s logic is captured in a plain English specification precise enough that the system could be rebuilt from the document alone. It serves as a standard operating procedure (SOP) that any qualified person or auditor can read, verify, and hand forward.
  2. Resilience to changing inputs: Unlike traditional automation, which is tightly tied to specific file formats and hard-coded rules, AI understands the meaning of the data. It can interpret information in context, determine whether a business rule still applies, and adapt when reports or inputs change. Instead of breaking whenever a spreadsheet or report is reformatted, the AI maps new inputs to the intended business logic, keeping the solution accurate, flexible, and durable without requiring a costly rebuild.

 

The human in the loop

The solution was built around the person who knew the domain best, the client stakeholder. Rather than presenting a finished answer, Prowess Consulting surfaced findings and open questions as the rules took shape, and the client made the calls that only a domain expert should. Where records were ambiguous, we developed the AI solution not to resolve the ambiguity on its own, but instead to make the ambiguity visible so a human could decide, and to then apply that decision consistently across every contract. With this structure, the expert stays in control, while the system does the tireless, uniform work.

The result

The full working dataset was resolved into a clear, per-contract determination that the client can trust and defend. The deliverable is a self-validating workbook that re-derives and re-confirms the headline totals, so a reviewer can watch it check itself. Contracts that look unusual are quarantined for human review rather than guessed at, which is precisely why the trusted result stays trustworthy.

The solution passed two independent trust tests:

  1. A second, separate implementation of the solution built purely from the natural-language-written specification reproduced the engine’s result exactly.
  2. Those results were independently re-derived on a separate path and matched.

Just as important as the outcome, the solution belongs to the client, not the people who built it. We delivered the natural-language ruleset, a methodology overview, a data findings report, and an operator’s runbook, giving the organization everything needed to operate, audit, and evolve the solution independently. While the immediate project resolved a century-old challenge in just weeks, it was intentionally designed to be reused for future reconciliations, providing a sustainable path to reduce effort, improve workflows, and prevent the organization from facing the same challenge again.

 

What changed

  • A 15-year, $21.9M reconciliation backlog was resolved in weeks, not months, at a fraction of the expected cost.
  • Every figure in the final determination is traceable to the source record that produced it.
  • A plain-English rulebook enables any auditor or successor to read, verify, and maintain the solution.
  • Human judgment is preserved where it matters, with uniform application everywhere else.
  • A reusable pattern provides build-time reasoning, runtime determinism, and human-in-the-loop review, all in a form that the organization can apply to future reconciliation and audit problems.

A concise solution for “records don’t agree” problems

With the deferred-revenue determination in hand and defensible, the organization now has both the answer and the operating model behind it. The same build-time-reasoning and runtime determinism solution developed by Prowess Consulting is a fit for any records-don’t-agree type problem where the answers demand real domain judgment, yet still have to be auditable, repeatable, and durable as the data changes.

All regulated entities confront this shape of problem at some time, and many live with the problem far longer than they might want to admit. The difference now is that the path to resolution for such problems can be moved out of the hands of a costly army of practitioners and into a hyper-efficient, automated system with fast, durable, audit-ready results.

If you’d like to learn more or connect with Julian Lancaster, CISO at Prowess Consulting, fill out our Contact Us form or schedule a time to meet with him directly.

 

About Prowess Consulting

Prowess Consulting helps organizations adopt agentic AI that works with existing systems, not against them. With over 15 years of experience in AI, automation, and data—and deep roots in financial services and enterprise—we reduce time-to-value from months to weeks and empower teams to lead the change.

Learn how Prowess can drive measurable impact in your business.

Learn more about Prowess' agentic AI offering

To protect client confidentiality, certain names, details, and identifying information have been modified or omitted. The results and outcomes presented are based on a real client engagement.

Prowess and the Prowess logo are trademarks of Prowess Consulting, LLC.

Copyright © 2026 Prowess Consulting, LLC. All rights reserved.

Other trademarks are the property of their respective owners.

Related Posts

Are you ready to expand your reach?

Get in touch today to discuss how Prowess can make a real difference for your enterprise.