Research

One method, applied to questions that matter.

Published with its limits stated. We are not a consultancy with a research budget, and the way to tell the difference is that we publish things that are inconvenient for us.

One report published · One Note · Monthly is the goal

Two series, labeled before you read a word.

900 Notes 900 Reports
What it is What we are thinking about, and why. Fast. What we found, and how confident we are.
Cadence Monthly is the goal Two or three a year
Preregistered No Yes, on OSF
Independently verified No Yes, and the error rate is published
Systematic search No, and it says so Yes
Can it settle a question? No Yes

A Note may raise a question. It may not settle one.

If a Note reaches a conclusion that would change a client’s decision, an investment, or a public policy position, it stops being a Note. The practical test: if the piece would make someone spend money differently, it is a Report.

Every other firm publishes everything at one apparent level of confidence. A point of view and a global survey look identical on the page, and the reader cannot tell which one was checked.

The label is not a disclaimer. It is the product.

Read the Notes Read the report
900 Report · March 2026 Ungated PDF

The AI Execution Gap

Why about 95% of enterprise AI fails or stalls, and what the top 5% do differently to reach measurable ROI

The number everyone quotes is not one number. “Failure” means something different in each survey it comes from, which is why the reported percentages look inconsistent even when they describe the same system. One measures organizations with no measurable profit‑and‑loss return. Another measures proof‑of‑concepts scrapped before production. Another measures projects abandoned after the pilot. They are not the same thing and they do not have the same fix.

The report reconciles the definitions, then asks the question that follows: why is spending accelerating while outcomes stay this uneven. Its answer is that the binding constraints are organizational rather than algorithmic.

36 sources Cross‑model adversarial review with primary‑source verification Pipeline v1.3

What this report is not

Read before downloading

It was produced before our current methodology and it does not carry the machinery that methodology requires. It was not preregistered. It was not checked by an independent verifier. It carries no published error rate. Our next report will carry all three.

Download the report (PDF)

2.3 MB. Your email is not required.

What is in the pipeline

A subject enters our pipeline only when it has a named owner who can carry it. Ideas without an owner are not listed here, because a list of directions nobody is working on is a prospectus rather than a pipeline.

Work Class Where it stands
The Agentic Enterprise Registered review Protocol approved. Public registration on hold pending a methodology revision. No sources collected yet.
Neurobehavioural decision dynamics Original study Led by Dr. Pontus Inerup, under academic or clinical governance. Ethics approval and a named principal investigator precede any work with human participants.
Evidence synthesis in treatment‑resistant depression Original study Dr. Inerup is principal investigator. Grant application live.

Other directions are under consideration. They are not listed until somebody owns them.

Corrections

Errors are corrected in the published artifact, in place, dated, with the change visible. Silent edits are prohibited. An error material enough to change a headline conclusion causes the piece to be retracted publicly, with reasons.

Corrections to date: none.

This list is not empty because we are careful. It is empty because we have published very little so far. It will not stay empty, and when it fills, it fills here.

If you have found something wrong in anything we have published, we want it: hello@900labs.com

Get them when they publish.

Monthly Notes is the aim. Reports two or three times a year. Nothing else.

900 Labs · Page revised 2026‑09‑12 · Corrections: none