# Case Study: An Insurer's First 12 Months — AI Strategy, Ethics and Governance

Source: https://www.geekswithgeeks.com/en/ai-strategy-governance/wrap-bank-case

> Plan an AI programme for a mid-sized insurer, from use-case selection to governance and reporting.

## A phased plan

**Months 1-2 (strategy)**: leaders choose goals (faster claims, lower support cost). A scoring model ranks ideas: claims-document summarisation and support reply drafting rank high; automatic claim rejection ranks low because it is high-impact on customers. **Months 2-4 (foundations)**: AI policy, risk tiers, inventory, a small review board, approved tools, staff training. **Months 4-8 (pilots)**: two assisted (human-reviewed) pilots with baselines, evaluation sets including Hindi and English slices, PII masking, impact assessments and vendor questionnaires. **Months 8-12 (scale and assure)**: dashboards, quarterly reviews, an incident drill, a regulation tracker, and decisions on build versus buy using three-year TCO. Automatic rejection is deliberately left as "not now" until fairness, appeal routes and regulation are clear.

## From strategy to accountable practice

A complete programme links value, ethics, risk, governance and monitoring into one routine.

![Four parts of an AI programme.](assets/figures/ai-strategy-governance/section-8-map.svg) — Figure 8.1 — Strategy, ethics, governance and monitoring.

## The programme on one page

Each line maps to a section of this course.

```text
Strategy      goals -> use cases -> scoring -> readiness              (Sec 1)
Business case ROI/NPV, build vs buy TCO, roles, training               (Sec 2)
Ethics        principles as questions, stakeholders, fairness slices   (Sec 3)
Risk          register, tiers, impact assessments                      (Sec 4)
Governance    policy, board, inventory + gates, vendors, evidence      (Sec 5)
Regulation    tracker, NIST AI RMF mapping, control matrix             (Sec 6)
Operations    incident runbook + drill, balanced KPIs, quarterly review (Sec 7)
```

## Say "not yet" with reasons

Good governance includes the discipline to defer a risky idea and record why, what would need to change, and when to revisit. That is a decision, not a failure.

**Quiz:** Why was automatic claim rejection ranked low in the case study?

- [ ] It is too cheap
- [ ] Insurance cannot use AI
- [x] It has high impact on customers and needs fairness and appeal safeguards first
- [ ] Scoring models ignore risk

*Answer:* It has high impact on customers and needs fairness and appeal safeguards first. High-impact decisions about people need stronger safeguards before automation.