The product
KLARMASS examines what the statements in a professional work product rest on — before you hand your own work over, and before you rely on work you received. However the work product was created: with AI, without AI, or through a mixed process.
Methodologically this creates an evidence layer over the document: for each identified claim it records what the claim rests on, what is missing, and how the open point was handled.
What the product contains
Claim discovery
Identification of statements that can be checked against something outside the document: quantities, rates, legal provisions, policy targets, attributions and forecasts.
Evidence inspection
For each claim: what the document offers as support, how adequate that support is for the claim actually made, and what is missing.
External verification
Where useful, KLARMASS discovers candidate sources, retrieves them, and records the passages that speak to the claim — supporting, contradicting or inconclusive.
Findings
Source scope, source access and claim support findings, each with the passage it refers to and a recommended next step.
Resolution
Your decision per finding, with author, timestamp and rationale. Nothing is silently repaired.
Review
Maker–checker separation: the person who prepared a work product cannot approve it.
Decision brief
A traceable projection of the records that already exist: what still requires attention, what has been handled, and where a human decision remains open. No additional assessment.
Evidence-bound feedback
Feedback to the creator in three sections. Positive items are derived only from observable records; the absence of a finding is never evidence of a strength.
What the product deliberately does not contain
- No truth score, confidence score, trust score or probability of correctness.
- No statement that a claim is true or false.
- No automatic rewriting of your document.
- No autonomous decision-making, approval or sign-off.
- No claim to have checked every statement in a document.
- No employee assessment, no qualification statement, no ranking of people.
- No percentage time-saving claim.
System outputs may be incomplete or inaccurate and may require professional interpretation. The responsible person remains responsible for final review, judgement, approval and use.