Meridian Financial Services Ltd
Reporting period April 2026 – June 2026 · Generated 30 Jun 2026 by Alice Okonjo, Chief Risk Officer
Overall governance score
Up 6 points since the previous report
Mature governance in operation
7
AI initiatives
1
High-risk initiatives
78%
Incident readiness
Section 1
Overall AI governance score is 71%. That is up 6 points against the previous quarter, driven by completed model documentation and the new supplier due-diligence pack.
7 AI initiatives are on the register: 3 approved, 1 approved with conditions, 2 in review, 1 draft. Two initiatives moved into production monitoring this quarter.
1 initiative is high risk — the credit decision support model. It is approved with conditions, has documented human oversight, and is reviewed monthly by the Model Risk Committee.
The top open priorities are: complete fair-lending disparate-impact testing on the credit model; finalise the EU AI Act technical documentation pack; close the vendor transparency gap on the CV screening tool.
Incident readiness stands at 78%. One corrective action was raised and closed this period, with no AI incidents affecting customers.
Section 2
| Initiative | Owner | Risk tier | Status | Last review |
|---|---|---|---|---|
| Credit decision support model | Priya Raman | High | Approved with conditions | 12 Jun 2026 |
| Customer service chatbot | Daniel Whyte | Limited | Approved | 28 May 2026 |
| Document AI for claims intake | Sofia Marchetti | Limited | Approved | 04 Jun 2026 |
| CV screening assistant | Helen Brooks | Limited | In review | 19 Jun 2026 |
| Fraud anomaly detection | Priya Raman | Limited | Approved | 21 Apr 2026 |
| Marketing copy generation | Tom Reilly | Minimal | In review | 09 Jun 2026 |
| Internal knowledge assistant | Daniel Whyte | Minimal | Draft | 24 Jun 2026 |
1
initiative
4
initiatives
2
initiatives
Section 3
74%
37 of 50 items complete
68%
30 of 44 items complete
77%
27 of 35 items complete
Section 3
81%
13 of 16 items complete
60%
6 of 10 items complete
Section 4
| Initiative | Decision | Decided by | Date | Conditions |
|---|---|---|---|---|
| Credit decision support model | Approved with conditions | Alice Okonjo | 12 Jun 2026 | Monthly disparate-impact testing, mandatory human review of all declines, and re-approval in six months. |
| Document AI for claims intake | Approved | James Fielding | 04 Jun 2026 | None recorded |
ISO/IEC 42001 Clause 6.1.4 and Annex A.5 · EU AI Act Article 27
Credit-risk scoring model · High risk
Residual impact: Medium · updated 18 Jun 2026
Affected: Approximately 4,200 consumer credit applicants a month, including first-time borrowers and applicants with thin credit files.
Potential impacts: Incorrect or biased scores could deny credit unfairly, with disproportionate effect on younger applicants and applicants in lower-income postcodes.
Mitigations: Documented model validation, quarterly disparate-impact testing, mandatory human review of all declines, and a published appeal route with a five-day response target.
CV screening assistant · High risk
Residual impact: Medium · updated 02 Jun 2026
Affected: All external candidates applying to advertised roles, around 900 applications a quarter.
Potential impacts: Ranking errors could exclude qualified candidates and introduce indirect discrimination on age, disability or ethnicity.
Mitigations: Recruiters review every shortlist and rejection, protected characteristics are excluded from features, and quarterly fairness testing is recorded with the equality impact assessment.
Section 5
Incident readiness
78%
| Action | Context | Owner | Due |
|---|---|---|---|
| CA-2026-03 Chatbot disclosed an outdated fee schedule to 14 customers | Customer service chatbot — closed 27 May 2026 | Daniel Whyte | 27 May 2026 |
Section 6
FIN-2 Run fair-lending disparate-impact testing on the credit model
Industry-specific AI requirements · Owner: Priya Raman · Due 31 Jul 2026
EUAI-4.2 Complete the EU AI Act technical documentation pack
EU AI Act · Owner: Sofia Marchetti · Due 15 Aug 2026
UK-5 Complete the equality impact assessment for CV screening
National & regional requirements · Owner: Helen Brooks · Due 22 Aug 2026
ISO-9.2 Schedule and run the first internal AIMS audit
ISO/IEC 42001 · Owner: James Fielding · Due 30 Sep 2026
NIST-MEASURE Establish ongoing performance drift monitoring
NIST AI RMF · Owner: Priya Raman · Due 30 Sep 2026
Appendix
Each activated framework produces a checklist scoped to your organisation profile — jurisdictions, industry, company size, AI maturity — and to the risk tier of each registered AI initiative. Items that do not apply to your profile are never counted.
A framework's percentage is the share of its visible items marked Done or Not applicable. The overall governance score is the unweighted mean of the active framework percentages, so no single framework can dominate the headline number.
Crosswalk references mean one piece of evidence can satisfy several frameworks at once: completing an item ticks every framework item that shares a reference code, which is why percentages can move together.
Incident readiness is measured separately as completion of the post-market monitoring and continual improvement sections. Decisions and approvals are drawn from the immutable decision log, and every generation of this report is written to the audit log.
This report is generated from the organisation's own governance records and is guidance to organise compliance work, not legal advice. Verify obligations against the official framework texts.