1,913 products · 66 categories · 6 AI judges each
Business software, judged by AI, out loud.
Every product scored by a bench of AI judges against a published rubric — and when they disagree, you read both sides.
- Cited, not claimed. Every score links to the vendor page it was read from.
- Disagreement printed. Six judges, six weightings, no averaging away.
- Provenance proven. Where a tool is made and hosted, sourced from its own pages.
Hosting options & portability: 0 out of 10, and 8 out of 10. Same evidence.
- Operations Analyst 0
- Integration Engineer 7
- Process Owner 7
- Skeptic 7
- Data Protection Officer 8
- Head of IT 8
The Operations Analyst A full self-hosted edition deploys with Docker, air-gapped if needed, with the source on GitHub under a fair-code licence and a...
The Head of IT This is the exit story I look for: a self-hosted edition deployable with Docker, air-gapped installation offered, source code on...
Every evaluation here is AI‑generated. On purpose. Out loud.
Incentivized reviews, gated results and pay-for-placement rankings leave buyers with stars nobody can trace to a fact. Stars you can’t check don’t tell you much. A machine judgement can be audited: its rubric is published, its evidence is timestamped, its judges are versioned, and when they disagree, the disagreement is printed — never averaged away.
Stars are the panel’s opinion on a published 0–5 scale. Each one resolves to a score against written anchors, and each reason cites the source it was read from. Where the data is thin, the page says so — and anyone can report an error. How to read the stars
The bench
6 judges, built for Workflow Management
-
The Process Owner
Runs purchase approvals and invoice release for a 400-person Mittelstand company. Wants appr...
-
The Integration Engineer
Wires the ERP to the shop, the shop to DATEV and the CRM to everything else. Wants connector...
-
The Operations Analyst
Is asked every Monday why invoices took eleven days. Wants a searchable execution history, c...
-
The Head of IT
Answers for every automation a department builds without asking. Wants SSO, separate test an...
-
The Data Protection Officer
Reads every workflow as a chain of transfers. Asks where the execution log keeps its copy of...
-
The Skeptic
Knows that "7,000 integrations" counts connectors nobody maintains, that "unlimited workflow...
Every judge is AI, versioned, and scores independently against the same published anchors. Nobody sees anyone else’s answer, which is why the numbers come back apart.
Categories
Each with its own bench of six judges
Across categories
How European is your software stack? Check it tool by tool →
How a verdict is made
Step 1
Capture
Public vendor pages are harvested, hashed and timestamped into an evidence registry. No claim exists without a screenshot behind it.
Step 2
Judge
A bench built for the category scores independently against published 0–10 anchors. Different priorities, same evidence, no peeking at each other.
Step 3
Publish
Verdicts ship with citations, capture dates and every disagreement intact — plus the ways they might be wrong, and a standing invitation to correct them.
Latest verdicts
Each dot is one judge, on the 0–10 scale behind the rating
Vendors: claim your listing, correct any fact, add screenshots and evidence, fill in what our checks left out. No score or placement is for sale.
Is this your product?