Nordglade
Est. 2026

AI work should be inspectable.

You can see what it used, what it checked, and where it stopped.

This matters when software answers an operational question, prepares a report, or recommends what to do next.

A public laboratory by Imre Bartis

01

The laboratory

Nordglade records experiments with AI used for business decisions. Each entry shows what information the software used, what it checked, and when it did not have enough evidence to continue.

For an SME leader, that matters more than a polished answer. You need to know whether the answer can be trusted.

02

Portfolio

Two entries

Two entries · open the portfolio

§

The evidence

Entry 02
The walkthrough at 64 seconds: three opportunities with their readiness percentages, and a panel headed “greenlighting this does not make it possible”.
The walkthrough shows predictive maintenance scoring well as an opportunity but poorly on readiness.

The missing maintenance records changed the recommendation from a priority project to “not now.” See why in the full entry.

03

Method

What every entry shows

01

What it knew

The company information available to the software.

02

What it checked

The checks it performed and what they found.

03

Where it stopped

What was missing when it could not support an answer.

04

What the experiment proves

What the result demonstrates and what it does not.

04

About these experiments

Fictional scenarios

The companies and their data are invented unless an entry clearly states otherwise.

Working prototypes

These experiments test how software handles a business question. They are not finished products or client results.