Sample report A KarmaScore report, as it is producedExemple SA does not exist. Its 26 answers are fixed in the code; the report below is recomputed on every view by the same engine that serves real assessments. This is what you get at the end of the questionnaire, with no account and no data sent anywhere. Exemple SA · Industry & Energy · 10–49 employees · assessed on 20 September 2026 Overall score47.9out of 100Needs strengthening 22.1 points from the label (threshold 70) Better than 70% reference SMEs in the Industry & Energy family, indicative comparison: the reference panel currently consists of simulated data, progressively replaced by the real results of pilot companies. Self-declared score: it reflects your answers as of today. Taken again another day, it may move; that is normal. The flash audit stabilises it by verifying the answers against evidence. 1 critical issue: not offset by the scoreA decent overall score can coexist with a serious failing: these indicators touch legal obligations and stand at zero. They come before everything else, whatever your score. - AI Act applicability and classification Without an AI Act applicability review, your exposure is unknown: if the EU market is affected, the transparency obligations of Art. 50 have applied since 2 August 2026.
Your strong points- Hardware life cycle75/100
- Data governance75/100
- Basic IT security75/100
The score is fully decomposable: each of the 26 indicators contributes to the total according to the weight of its dimension. No black box. Action plan: critical risks first, then the quickest winsPoints measure the score, not the risk actually addressed. An action counts as done when its evidence exists. 1. AI Act applicability and classification+1.04 pts Move from “Never examined” to “Aware of the risk categories, applicability not checked”. - Owner
- Management
- Deliverable
- Table of obligations / owners / deadlines
- Effort
- a few hours (estimate)
- Expected evidence
- Every use case on the list has its written classification, and every obligation in the table has an owner and a follow-up date.
2. Frugal AI+1.07 pts Move from “Need questioned case by case, nothing written” to “Written rule: non-AI alternative first, model proportionate to the task”. - Owner
- IT lead, with management for removals
- Deliverable
- Monthly consumption reading per use, converted to energy and location-based CO2e
- Effort
- ongoing effort (estimate)
- Expected evidence
- The latest monthly reading is dated and matched to each use in the table, the default settings have their dated screenshot, and the last annual pass says which uses were cut or kept and why.
3. Responsible IT procurement+1.07 pts Move from “Criterion applied to major purchases” to “Purchasing grid includes the environment”. - Owner
- The person in charge of purchasing + management
- Deliverable
- Purchase scoring grid, with completed copies
- Effort
- 1 to 2 days (estimate)
- Expected evidence
- The latest IT purchase has its grid filled in with the environmental criterion, and the annual table says which supplier commitments were checked.
4. Impact assessments (DPIA)+1.04 pts Move from “Occasional analysis for major projects” to “Documented analysis for every high-risk processing activity”. - Owner
- Management + the project lead concerned
- Deliverable
- Sheet of trigger criteria
- Effort
- 1 to 2 days (estimate)
- Expected evidence
- The most recently launched project has its written risk test, dated before launch, and its impact assessment when the test triggered one.
5. Control over AI vendors+0.89 pts Move from “Informal reputation check” to “Data location verified”. - Owner
- Management + IT or procurement lead
- Deliverable
- Supplier evaluation grid (1 page)
- Effort
- 1 to 2 days (estimate)
- Expected evidence
- The most recently purchased tool has its assessment grid filled in and dated before the purchase, and every critical supplier has its signed data processing agreement on file.
The written summaryDeterministic version of the report, written without artificial intelligence: it is the one served when the writing model is unavailable, or when the guardrail rejects its text. SummaryWith 47.9/100, Exemple SA sits 22.1 points below the threshold of 70 required for the Swiss Ethical AI Inside label (version α). The company does better than 70 % of 10 reference SMEs in the Industry & Energy family: an indicative comparison, as the reference panel is still synthetic. Reading by dimension- T · Technology: 50/100 (weight 25 %), +9.5 points against the reference average of 40.5.
- C · Compliance: 41.7/100 (weight 25 %), -4.5 points against the reference average of 46.2.
- E · Environment: 50/100 (weight 30 %), -5 points against the reference average of 55.
- S · Society: 50/100 (weight 20 %), +7 points against the reference average of 43.
Strengths- Hardware life cycle (E): 75/100.
- Data governance (T): 75/100.
- Basic IT security (T): 75/100.
Plan for the next 90 days- Priority 1, AI Act applicability and classification (C): move from “Never examined” to “Aware of the risk categories, applicability not checked” (+1.04 points estimated).
- Priority 2, Frugal AI (E): move from “Need questioned case by case, nothing written” to “Written rule: non-AI alternative first, model proportionate to the task” (+1.07 points estimated).
- Priority 3, Responsible IT procurement (E): move from “Criterion applied to major purchases” to “Purchasing grid includes the environment” (+1.07 points estimated).
- Priority 4, Impact assessments (DPIA) (C): move from “Occasional analysis for major projects” to “Documented analysis for every high-risk processing activity” (+1.04 points estimated).
- Priority 5, Control over AI vendors (T): move from “Informal reputation check” to “Data location verified” (+0.89 points estimated).
Next stepThe T · Technology dimension is the firmest support (50/100); the C · Compliance dimension (41.7/100) holds the most room for progress. A reassessment is recommended once the plan is implemented. Exemple SA's answersThe 26 indicators, with the level chosen for each. Every level counts according to the weight of its dimension: the sum is the score. | T | Inventory of AI and digital systems | Inventory of the main tools | | T | Explainability of automated decisions | Explanation possible for some tools | | T | Basic IT security | Security policy enforced, backups tested, breaches logged | | T | Data governance | Data owners and managed access rights | | T | Human oversight | Human review of sensitive cases | | T | Control over AI vendors | Informal reputation check | | T | Fallback plan for AI tools | Fallback discussed for a few tools, nothing written | | C | Record of processing activities | Checked whether the obligation applies; main processing activities listed | | C | Informing data subjects | Adapted but incomplete policy | | C | Impact assessments (DPIA) | Occasional analysis for major projects | | C | AI Act applicability and classification | Never examined | | C | Data subject rights | Point of contact identified | | C | Staff training | Written charter distributed | | E | Low-footprint hosting | Energy profile known for the main services | | E | Measuring the digital footprint | First estimate produced | | E | Hardware life cycle | Purchasing policy includes refurbished equipment, storage media wiped before disposal | | E | Digital sobriety | Simple rules (mailboxes, archiving) | | E | Frugal AI | Need questioned case by case, nothing written | | E | Responsible IT procurement | Criterion applied to major purchases | | E | Reduction targets | Target set, tracked irregularly | | S | Digital accessibility | Occasional fixes (contrast, alt text) | | S | Bias prevention | Occasional check on one sensitive tool | | S | Transparency about AI use | Notice on the main channels | | S | Impact on jobs and skills | Occasional training on new tools | | S | Inclusion and diversity | User feedback gathered occasionally | | S | Avenues of recourse | Human contact available on request |
Your report will look like this one, with your answers. It stays in your browser; you can receive it by email, print it, or come back to it later. |