How is sensitive data protected?

All documents are encrypted with AES-256-GCM at rest and TLS 1.2+ in transit. A GDPR anonymization layer removes PII before any AI processing. Hosted on Vercel + Supabase (both SOC 2 Type II); Decision Intel’s own product-level audit is targeted for Q4 2026.

How long does integration take?

Less than 30 minutes. Upload documents directly, connect via OAuth for Slack, or use our REST API for bulk processing.

How does outcome tracking work?

Outcomes are detected automatically from follow-up documents, Slack messages, and web intelligence. You confirm with one click. Each reported outcome makes your future analyses more accurate.

How is Decision Intel different from ChatGPT or a general AI assistant?

ChatGPT gives one opinion from one model: ungoverned, untraceable, unaudited. Decision Intel is the reasoning audit platform. Most tools audit your data; we audit your reasoning — and catch the fatal blind spots in strategic memos before the committee does. It measures the noise in your reasoning the same way Kahneman did in the insurance underwriter study, simulates an AI boardroom of CEO, CFO, and board personas, runs what-if interventions against a 143-case public reference library, and compounds every confirmed outcome back into a calibrated Decision Quality Index your audit committee can defend. Not a chatbot; a reasoning audit, checkable from memo to outcome.

Prospective track record

Reasoning audits, locked in public, before the outcome is known.

Every call below names a specific reasoning-risk in a public decision, a falsifiable test, and a due date, all published in advance. We score whether the risk we flagged materialised, not whether we predicted the price, and we publish the calls we get wrong. The decisions are public and checkable; the reasoning-risk each one names is the shape that recurs in the memos and deals you actually work on.

The method

Score the flag, not the forecast.

Why this exists: you should not have to take our word that the audit catches what a committee misses. So the call is made in public, in advance, with a falsifiable test and a due date, and the result is published either way. Check the hit rate. That is the only honest proof a reasoning audit can offer before it is in your workflow.

A forecaster predicts the outcome. Decision Intel is a reasoning auditor: it names the specific reasoning-risk a committee should pressure-test, and the public test is simply whether that risk materialised. The share price can run while the risk we flagged still bites, and the call would still be validated, because the call was never about the price.

Decision Intel identifies reasoning-risk indicators correlated with poor outcomes — it does not claim to establish causation. That honesty is the procurement weapon: a CSO can defend a score that names the risk indicators a committee should pressure-test, not one that pretends to prove the cause.

Calibration across many calls beats being right once. We publish the false positives, because a record you can only see the wins of is not a record.

Locked

Published in advance, the falsifiable proxy set. The due date has not arrived.

Tracking

A proxy window has opened. The signal is being watched against the flag.

Flag confirmed

The reasoning-risk we named materialised, a true positive. The share price is irrelevant to this result.

False positive

The flagged risk did not bite. Published anyway, because calibration needs the losses.

Mixed

The flag partially materialised. Scored on its merits, never rounded up to a win.

On the record

One call locked so far. The record is public from the first.

New calls land as they are made, on public decisions only. Each is dated, falsifiable, and left on the record whichever way it resolves.

SpaceX (SPCX) S-1 IPO thesisLockedLocked Jun 21, 2026 · Due Dec 31, 2026

FlagValuation gated on a Starship timeline the reference class (5/25 flights in 2025) says is optimistic.

Falsifiable proxyFirst commercial Starship payload flies by Dec 31 2026 (the S-1’s own "H2 2026" milestone).

Why this is your risk too

You will never audit a $1.77T IPO. But a GP backing a late-stage round, and a CSO defending a market-entry memo, face the same shape of risk: a coherent narrative gated on a timeline the reference class says is optimistic. This is a public, checkable instance of the exact reasoning-risk your own decisions carry.

Monitoring ladder

3-month · ~Sep 2026Does the first post-IPO reporting reaffirm or quietly soften the "H2 2026 commercial payload" milestone, and is the test-flight cadence an H2 commercial debut would require actually happening?

6-month · Dec 31 2026Did the first commercial Starship payload fly? (primary) Secondary: did Starlink net-adds decelerate or ARPU keep falling?

12-month · Jun 2027Is the gap between the 2030 milestones priced in today and what has actually shipped narrowing or widening?

How this is scored

We are not predicting the share price. The stock could rise on Starlink alone while the Starship timeline slips, and the flag would still be validated, because the flag was about the reasoning-risk, not the price. The scored unit is simple: did the flagged risk materialise?

The fine print

What this record is, and what it is not.

What it is

Public, dated, and falsifiable. Each call is published before its due date, on a decision anyone can check.
Scored on the flag: whether the reasoning-risk we named materialised, win or lose.
Inclusive of the losses. False positives stay on the record, because calibration needs them.

What it is not

A forecast of prices. We do not predict the stock; we test whether a flagged reasoning-risk bites.
A causal claim. We name risk indicators correlated with poor outcomes, never a proof that a bias caused one.
A verdict on the companies audited. We audit the reasoning, not the company.

Run it on your own decision

Paste a memo, closed or upcoming, and see the same audit on your actual text. No signup, no card.

See the same engine in hindsight

The identical methodology applied to decisions whose outcomes are already known, with no hindsight used.