Title: AI Boardroom Forecast Audit — MidnightDev
URL: https://midnightdev.dev/work/boardroom
Description: The AI boardroom that killed a $65.9M forecast.
Generated from the live page on 2026-10-22. Do not edit by hand; run `pnpm twins`.

---

case study

# AI Boardroom Forecast Audit

The AI boardroom that killed a $65.9M forecast.

[inspect repository](https://github.com/abouchard11/ai-boardroom-forecast-audit) · github.com/abouchard11/ai-boardroom-forecast-audit

The authority boundary behind all of this work is stated once, in [the model cannot own the money](https://midnightdev.dev/writing/the-model-cannot-own-the-money.md). This page covers what is specific to AI Boardroom Forecast Audit.

![AI Boardroom Forecast Audit diagram showing six synthetic decision lenses reviewing a $65.9M revenue forecast and flagging it as unsupported.](https://midnightdev.dev/screenshots/boardroom-audit.svg)

01 the problem

An AI-built operating plan for a real Houston festival concept produced an approximately $65.9M profit forecast. A six-seat synthetic review cut the headline dramatically, but a later audit found that even the corrected workbook mixed incompatible expense definitions.

02 the solution

Reconstructed the decision as a transparent Python model with 100,000 seeded Monte Carlo trials, standardized regression sensitivity, committed outputs, and invariant tests. Historical claims, reconstructed analysis, and real planning inputs are labeled separately.

03 decisions & tradeoffs

### Competing roles over one agreeable assistant

The review needed distinct incentives and downside ownership, not six paraphrases of the same optimistic answer.

### One canonical assumptions schema

Consensus did not prevent arithmetic drift. Every view must regenerate from the same definitions.

### Reconstruction labeled explicitly

The original run outputs were not fully preserved, so the public model is marked as reconstructed for reproducibility in 2026.

04 tech stack

### Synthetic executive review

Finance, operations, industry, legal, marketing, and investor lenses submitted competing objections.

### Monte Carlo simulation

Seeded trials stress-test sell-through, revenue execution, expense variance, weather, sponsorship, and delays.

### Regression sensitivity

Standardized coefficients expose which uncertain inputs drive the modeled outcome.

### Invariant tests

Machine-checkable reconciliation blocks incompatible definitions from quietly surviving consensus.

05 results

- Reconciled a $3.397M gap between reported and fully loaded base-case expenses
- Median reconstructed profit: $1.22M
- Probability of reaching at least $5M: 10.5%
- Deterministic outputs, tests, evidence ledger, and limitations published publicly

Next step

## Want something like this?

Book a 30-minute call. I will tell you plainly whether I can help and what I would test first.

[Book a call](https://calendly.com/alex-midnightdev/30min?utm_source=midnightdev&utm_medium=site&utm_campaign=case-boardroom) [or email alex@midnightdev.dev](mailto:alex@midnightdev.dev)