ThinkNEOThinkNEO
Case study

SHK Import & Export Runs Governed Multi-Provider AI Across Global Trade Operations

SHK Import & Export is a Hong Kong-based international trade operator active across Asia, Europe, and Latin America. Its operations team runs AI-assisted workflows for trade documentation, counterparty screening, and logistics intelligence. ThinkNEO is the control plane between those workflows and the underlying LLM providers.

Book DemoBook a Demo
  • Governed multi-provider routing with deterministic budget ceilings
  • Compliance-grade audit trails across every AI request
  • Sovereign GPU fallback to preserve continuity during provider incidents

Customer Profile

SHK Import & Export is a Hong Kong trade operator coordinating cross-border flows across multiple jurisdictions.

Headquarters
Tsim Sha Tsui, Kowloon, Hong Kong
Industry
International trade and cross-border commerce
Operating footprint
Asia, Europe, Latin America
AI workloads in scope
Trade documentation, counterparty screening, logistics intelligence, market research
Status
Active on ThinkNEO

Operating Signals

Qualitative indicators of how ThinkNEO changed the operating model. We do not publish absolute transaction volumes or raw cost figures.

Provider independence
Multi
Requests route across OpenAI, Anthropic, Google, and sovereign models without application changes.
Budget ceiling
Hard cap
Monthly AI spend is enforced at the gateway. Overruns are structurally prevented, not just alerted.
Audit coverage
100%
Every request is logged with hash-chained, append-only audit events preserved indefinitely.
Fallback tiers
3
Primary provider, secondary provider, and sovereign GPU fallback keep workflows alive during incidents.

Why SHK Needed A Control Plane

Before ThinkNEO, SHK's AI-assisted workflows had the same structural problems that affect most international operators when they put LLMs into production.

  • Multiple LLM providers integrated ad hoc, with no shared policy layer or unified cost view.
  • No way to enforce budget ceilings at request time. Overruns only became visible on the next invoice.
  • Audit evidence lived in scattered application logs that were mutable and incomplete for compliance.
  • Provider incidents or rate limits caused operational workflows to stop with no graceful degradation.
  • Data residency obligations across jurisdictions required routing logic that application code should not own.

How ThinkNEO Is Deployed

SHK runs all AI traffic through a single ThinkNEO endpoint, using the OpenAI-compatible URI as a drop-in gateway. ThinkNEO handles routing, enforcement, logging, and fallback. The application layer stays unchanged.

1. OpenAI-compatible gateway
SHK applications point at the ThinkNEO endpoint instead of the provider directly. No SDK rewrite, no custom routing logic in application code.
2. Runtime policy enforcement
Policies for model selection, data residency, and budget ceilings are applied at request time. Decisions are deterministic and logged.
3. Multi-provider routing
Requests route across OpenAI, Anthropic, Google, and Mistral based on policy, availability, and cost. Routing is transparent to callers.
4. Sovereign GPU fallback
If commercial providers fail or budget caps are reached, ThinkNEO falls back to self-hosted models running on sovereign GPU infrastructure. Workflows continue.
5. Immutable audit trail
Every request generates an audit event preserved with hash-chained integrity. Update and delete are blocked at the database level.

Operating Model Changes

These are the operating outcomes SHK reports from running the control plane. Absolute figures are private.

Predictable AI spend
Monthly AI budget is a hard ceiling, not a target. Finance no longer learns about overruns weeks later.
Continuity during provider incidents
Provider outages and rate-limit events no longer cascade into workflow downtime. Fallback is automatic and observable.
Compliance-ready audit posture
Regulators and internal audit can inspect an append-only, hash-chained record of every AI interaction across providers and jurisdictions.
Policy ownership outside application code
Model selection, data residency, and guardrails are managed at the control plane. Application teams focus on product, not on policy plumbing.
“ThinkNEO gives us full control over every AI call across our international trade operations — cost, compliance and continuity in one gateway.”
David Chen, Commercial Manager, SHK Import & Export — Hong Kong

Related Material

Continue exploring how ThinkNEO is positioned for regulated, multi-jurisdiction AI operations.

Resilience: how fallback and hard caps work
The architecture behind three-tier fallback and deterministic budget enforcement.
Open
Compare ThinkNEO with alternatives
Capability matrix against Portkey, LangSmith, and Helicone for side-by-side evaluation.
Open
Enterprise trust posture
Security, compliance, and audit posture for enterprise procurement.
Open

Evaluate ThinkNEO For Your Operations

Book a neutral technical review and map ThinkNEO's control plane against your provider stack, compliance obligations, and cost governance needs.

Book DemoBook a Demo