Skip to content
AA Consulting

Case studies

Outcomes, fully anonymized.

Regulated work carries a confidentiality obligation, so these are reported as sector and outcome only, never a client name or any identifying internal detail. The metric is the point; the internals are a liability we will not publish.

Engagements

  • Multinational insurer

    Three cloud guardrail systems consolidated behind one gateway policy.

    A regulated insurer ran separate content-safety controls per cloud, with no shared definition of what the AI should refuse. We consolidated them behind a single governed gateway policy with PII anonymization and national-ID regex screening, one place to decide, one place to enforce.

    3 → 1guardrail systems behind one policy
    100%LLM traffic screened for PII
    Guardrails, content safety & PIIAI Gateway architecture & build
  • Tier-one bank

    From zero cost visibility to full per-team attribution.

    Agent traffic was billed as one opaque line item across many teams. A governed gateway with mandatory cost-tracking headers and semantic caching gave finance per-team, per-app attribution and cut redundant model spend, without teams changing how they build.

    0 → 100%per-team cost attribution
    ~90%redundant endpoint cost removed
    AI Gateway architecture & buildSecure agent CI/CD & enablement
  • Healthcare provider

    EU-only residency held while putting clinical experts in the eval loop.

    A provider needed agent behavior graded by clinical domain owners, not engineers, while keeping all data in-region. Framework-agnostic observability plus annotation queues wired into CI let non-technical experts score traces, with residency preserved across the stack.

    EU-onlyresidency across the agent stack
    100%traces graded in CI
    Agent observabilityAgent evaluation

Illustrative engagements pending real anonymized data, figures are representative, not yet client-sourced.

Want the same, framed around your constraints?

Discuss an engagement