EVORI — Scoring framework for AI use cases
==========================================
Source: https://evori.ai/en/wissen/ki-use-cases-priorisieren-bewertungsraster
As of: September 2026

Score each candidate on four dimensions from 1 to 5 points.
Ideally, two people score independently and then discuss differences.

DIMENSION 1: VALUE (1–5)
  1 = occasional, a few minutes, one person
  3 = noticeable weekly, several people involved, errors create follow-up work
  5 = daily, across teams, a bottleneck affecting customers or closing

DIMENSION 2: FEASIBILITY (1–5)
  1 = paper, isolated systems, no interfaces
  3 = digital data, access can be resolved, systems partly open
  5 = structured data, accessible interfaces, a clear pattern

DIMENSION 3: DATA SENSITIVITY (1–5, higher = less sensitive)
  1 = special categories / sensitive sharing → GDPR review essential
  3 = personal data is involved, ordinary business data
  5 = internal factual/product data without personal information

DIMENSION 4: ERROR TOLERANCE (1–5)
  1 = errors affect others externally / create legal commitments before review
  3 = errors can be corrected internally and are spotted in normal review
  5 = results are suggestions that a person will review and use anyway

VETO RULES
  - Data sensitivity = 1: do not start until data protection is resolved.
  - Error tolerance = 1: do not start without a defined checkpoint
    (approval before an action takes effect, escalation rules).

TABLE (one row per candidate)

Candidate           | Value | Feasibility | Data sensitivity | Error tolerance | Total | Veto?
--------------------+-------+-------------+------------------+-----------------+-------+------
                    |       |             |                  |                 |       |
                    |       |             |                  |                 |       |
                    |       |             |                  |                 |       |

INTERPRETING THE SCORES
  - The total prioritises questions; it does not authorise a project to start.
  - A value score of 1 plus three scores of 5 gives 16 points. This alone
    does not justify custom development to save a few minutes of work.
  - Before implementation: weigh expected workload reduction and quality
    improvements against investment, operating costs and internal effort.
    Human review and introduction are part of that effort.
  - If the business case does not hold or assumptions remain unresolved,
    hold off or investigate further, even with a high total and no data
    protection veto.
  - High value, unclear data access: review again later.
  - This framework does not replace a business case or legal review.