Methodology

Evidence first. Numbers only with evidence.

qasb answers questions about how AI changes work, education, organizations and society. This page explains how we work with evidence, what a result contains and where our statements end.

01 Principles

Four rules for every analysis

qasb makes AI impact traceable: with sources, visible gaps and clear limits.

Supported or open
A statement rests on a source or stays visible as an open gap. Missing data is never filled in.
Origins kept apart
Source evidence, your input and qasb assessments are never silently mixed.
No overall score
qasb does not compress results into a single number and does not create rankings.
People decide
qasb makes no decisions about people. Results support a review and do not preempt it.

02 Evidence

How qasb works with evidence

An answer is only as good as what it rests on. So qasb looks for suitable sources, checks them before use and ties every supported statement to its source.

Suitable sources. qasb looks for studies, official statistics, institutional reports and product documentation that fit your question.
Checked before use. A search hit is not yet evidence. qasb checks whether a source fits the question, where it comes from, whether it may be used and whether it supports the statement.
Cited in the text. Every supported statement carries its reference. Original text appears only to the extent permitted, otherwise as a link.
Gaps stay visible. If suitable evidence is missing, the answer says so. qasb does not fill in a number or invent a source.

How to read a source

Source mark Follows the statement in the text and names the source’s website; a click opens this preview.
Source number A statement points to its source with this number.
Original title As the source gives it, without translation.
Publisher and date Who published the source and when it appeared.
Open original · Show in Verify Leads to the source itself or to its row in “Verify”, with the statements that rest on it.

Example: a source as the dossier shows it when you click its source mark. If a source is graded, the preview also names its evidence grade.

Evidence grade

Where known, the source preview names the evidence grade. It describes a source’s method, not whether a statement is true. qasb counts the grades and never averages them.

A: Systematic synthesis of multiple studies with documented methods.
B: Comparative empirical investigation with a described study design.
C: Descriptive data, preprint or research reporting; stronger methodological assurance is not established.
D: Provider statement, attributed opinion, forecast or secondary reporting.

Without information on its method, a source stays ungraded.

03 Results

What a result contains

qasb saves every analysis as a dossier with a fixed version. It starts with the short answer, then shows what it rests on and what remains open.

Short answer: The key point first, in a few sentences.
What is supported: Statements with their source marks, each one verifiable on its own.
What is open: What the sources do not settle is stated explicitly.
Source list: Every source used, with publisher, date and location, and its evidence grade where known.
Three views: Read explains the dossier as text, Overview shows what matters at a glance, and Verify puts every statement next to its sources.
Fixed version: A dossier never changes silently. Monitoring in Projects reports new sources; the original analysis and its values stay unchanged.

Nine areas, each with its own result structure

Each area has its own questions, sources and results. Personal and organizational questions start in the same analysis.

Work: Examine tasks, workflows and possible AI impacts.
Education: Explore curricula, skills and education pathways.
Product: Review a product’s capabilities, integration and risks.
Organization: Examine processes, value creation and implementation.
Public sector: Examine public services, actors and dependencies.
Security: Review security questions with primary sources and clear limits.
Research: Investigate research questions and primary studies.
Economy: Examine industries, markets and value creation.
Society: Review effects on people, everyday life and communities.

Work

Eight impact dimensions for work

Feasibility
1. Technical feasibility Can AI perform this work step in the planned use?
2. Workflow state distribution Which part of the work stays the same, is assisted or is automated?
Effort & quality
1. Change in human working time How does the time people spend change?
2. Review and control burden How much review and correction does the output need?
3. Quality change Does the output get better or worse?
Impact
1. Augmentation impact What does AI add while people stay involved?
2. Task substitution potential How far could AI take over part of a task, and within which limits?
Risk
1. Residual risk after controls What risk remains after effective controls?

For work, the qasb methodology separates eight impact dimensions. Other areas have their own result structures. qasb attributes only the additional effect of an AI deployment.

04 Statements

What a statement says, and what it does not

Measured is not modeled. With suitable sources, an analysis can show explicitly modeled potential with a scenario range. This value is not empirically calibrated and proves neither a measured effect nor a probability of occurrence.

Evidence classes

What a statement rests on stays visible in the dossier.

Evidence-backedE01: Qualified source bound to a statement and impact dimension
User contextE02: Supplied context, not independent evidence
qasb assessmentE03: Derived assessment rather than independent evidence
Open gapE04: Missing or insufficient evidence remains visible

The terms you will read

For work, qasb describes effects at the level of tasks, never as a judgment about a person.

AI support: AI helps with a task. People still do it and remain responsible.
Partial automation: AI handles individual steps of a task. People direct the work and review the result.
Substitution potential: How far AI could take over part of a task, and within which limits. A potential, not a forecast about a job.
Confidence: How robust an estimate is. Low confidence is a signal to look more closely.

What a statement does not say

A group is not a person.: Studies hold for their population, place and period. What they show for a group does not automatically apply to you.
New sources are not a trend.: Monitoring reports new sources linked to the connected analysis. A change in sources is not a measured trend.
Figures stay with their source.: Every figure keeps its source and its unit. qasb does not merge figures into an overall score.
Reports without customer data.: Monthly reports label every statement as observed, model estimate, scenario or open and contain no customer data. qasb does not currently publish empirical impact values.

05 Limits

What qasb is not

qasb supports reviews and decisions. It does not replace them.

Automated AI estimate: Clearly labeled
Human expert review: Claimed only when evidenced
Legal assessment: Issue screening, not legal advice
AI checks: Multiple models are not independent proof of truth; results may contain errors.
No individual forecasts: qasb gives no individual advice and no forecasts for individual people or jobs.
Unsupported decisions: No assessment, ranking, or decision about individuals in employment, education admission, credit, insurance, benefits, migration, law enforcement, justice, or biometrics. Aggregate analysis and human expert review do not widen this boundary.Technically blocked

Provider and contact: legal notice