FICTIONAL / SYNTHETIC DEPLOYMENT MODEL. "Forge Property Management" is an invented customer. No real customer data was used and no real production results are claimed. All figures come from a deterministic synthetic corpus and stated illustrative assumptions.
FForge Outcome Delivery
100,000 Work Order Simulation

Overview

← Back to case study

Forge Property Management — 100,000 Work Order Automation Impact Study

FICTIONAL / SYNTHETIC DEPLOYMENT MODEL. "Forge Property Management" is an
invented enterprise customer. No real customer, property, tenant, vendor, or
work-order data was used. Every figure here comes from a deterministic synthetic
corpus and stated illustrative assumptions — not from a real production system.
This is a synthetic deployment model, not a real customer deployment, and no real
production results are claimed.

What this is

A scaled enterprise simulation of the Work Order Agent Ecosystem for a fictional 42,000-unit property manager handling 100,000 work orders a year. It runs the real, unmodified ecosystem — four agents (classify → route → validate → action) over a real PostgreSQL engine (PGlite) and a real gRPC dispatch service — against 100,000 synthetic, labeled work orders, then measures operational performance and models the financial impact versus a fully-manual baseline.

Why it exists

To answer one question honestly: *if* an enterprise the size of the fictional Forge PM routed its entire annual work-order volume through this ecosystem, how much of it could be actioned automatically, how safely, and what would the labor economics look like — with every number reproducible by a stranger.

Headline results (synthetic, illustrative)

Run on the full 100,000-order synthetic corpus over live infrastructure (exact numbers are emitted to verification-report.json / verification-report.md):

  • More than half of all work orders are actioned automatically, with the remainder

routed to humans as exceptions.

  • Zero false auto-actions on the synthetic answer key (the safety-critical metric).
  • 100% of work orders carry a complete, append-only audit trail.
  • An illustrative annual labor saving in the mid-six figures, with a sub-12-month

modeled payback, under the stated assumptions in assumptions.json.

Quick start

# from this folder (runtime deps are resolved from ../work-order-agents)
node verify.mjs                 # full 100,000-order simulation + checks (~15-20 min)
WO_SIM_N=2500 node verify.mjs   # fast development run

node publish.mjs                # build the public case-study page

verify.mjs regenerates the synthetic dataset, runs the ecosystem end-to-end, writes verification-report.json/.md, and emits machine-readable evidence to evidence/ (simulation-results.json, roi.json, audit-trace-sample.json).

Contents

PathWhat
src/enterprise-synth.mjsDeterministic 100k synthetic work-order generator
src/roi.mjsTransparent manual-vs-agent ROI model
assumptions.jsonAll stated operational + financial assumptions
verify.mjsEnterprise simulation runner + MUST_PASS verification
build-deliverables.mjsGenerates the executive report + certification report from results
report/impact-study.mdExecutive impact study (PDF-ready)
datasets/Dataset card, sample rows, and dataset manifest
social/Social media package (clearly labeled fictional)
proof/Full IRS_AUDITOR proof package
publish.mjsStatic public case-study page generator

Honesty & certification

The Proof Layer (IRS_AUDITOR) assigns the authoritative certification state independently. Because the customer and data are fictional — a disclosed outcome seam — the state is intentionally not PRODUCTION_VALIDATED. The simulation-tier label is FICTIONAL_DEPLOYMENT_MODEL_CERTIFIED, meaning "the simulation itself is validated and reproducible," which is explicitly distinct from any production certification. See proof/LIMITATIONS.md.