What would you hand them?

In July 2026 the Digital Omnibus softened the AI literacy obligation to one of effort. It left Article 26(2) untouched: by 2 December 2027, deployers of high-risk AI must assign oversight to named people with the necessary competence, training and authority.

An attendance list is not evidence.

Most institutions completed AI literacy training in 2025. Very few could name the people overseeing each high-risk system and show they were competent to do it. These are not the same artefact.

What most institutions have

Name
Session attended
Completion rate

What a supervisor asks for

Role and the AI systems it operates
Assessed competence against a defined standard
Date, assessor, re-assessment trigger
Residual gaps, with remediation dates

The AI Competence Evidence Review

Three weeks. Fixed scope, fixed fee. It produces the evidence file — not a training proposal.

Week One

Scope

We map every role that operates an AI system, and every system it operates. Population in scope, high-risk classification, and the roles nobody had counted — usually the largest finding of the three weeks.

Week Two

Assess

We calibrate the competence standard to your institution and assess a defensible sample of each role tier. Self-rating establishes confidence; the assessed component establishes competence. The distance between them is itself a finding.

Week Three

Evidence

Gap register, remediation plan with owners and dates, and a draft attestation formatted for CRO and CCO sign-off. Delivered to your Compliance function, not to L&D.

Five artefacts, all of them auditable.

01

Population-in-scope map

Every role, every AI system it operates, high-risk flag, headcount, named role owner.

02

Competence standard, calibrated

Thirty competencies across six domains, with the required level differentiated by role tier — because oversight competence under Article 26(2) attaches to named individuals and the systems they actually operate, not to a uniform course.

03

Assessed baseline

A defensible sample of each tier, scored against the standard, with the confidence gap surfaced separately.

04

Gap register and remediation plan

Each competency below standard, the population affected, the action, the owner, the committed date.

05

Draft competence attestation

Method statement, coverage, residual gaps, change triggers, and a sign-off block. Ready to take to committee.

Who this is for.

This is for you if

  • You deploy AI in credit, fraud, underwriting, claims or AML
  • You ran AI training in 2025 and cannot evidence the outcome
  • You operate across several jurisdictions and face divergent transpositions
  • Compliance, not L&D, now owns the question

This is not

  • A training course — enablement comes after, once gaps are known
  • A legal opinion — we work alongside your Legal and Compliance functions
  • A platform purchase — the instrument is yours to keep and re-run
  • A twelve-week consulting engagement

Request the review.

We will come back with scope, fee and the earliest start date. If the review is not the right fit, we will say so.

How we handle your data. FIKS d.o.o., Belgrade, is the controller. We use what you submit only to respond to this request and, if you asked for it, to send you the evidence framework. We do not share it with third parties and we delete it on request. Write to raisa.khamidullina@fiksrs.com to access or erase your data.

The Digital Omnibus (Regulation (EU) 2026/1744, in force 27 July 2026) softened Article 4 to an obligation of means and deferred high-risk obligations for Annex III systems to 2 December 2027. The Article 26(2) oversight-competence requirement is unchanged. This page is a professional view, not legal advice.