AI-Powered Risk-Based Enforcement

Identify the most dangerous non-compliance. Target it first.

AI-driven tools for governments and businesses that need to find the highest-risk products, transactions, and supply chains — and act on them proportionately, with limited resources.

P×C Risk Map — live example

Each point is a product. Position = non-compliance risk. Size = market exposure.

P × C
Probability × Consequence
  • PProbability that a product, shipment, or transaction is non-compliant — based on origin, history, market structure, and intelligence signals
  • CConsequence of non-compliance — severity of harm, exposure breadth, reversibility, and regulatory impact
  • RResidual risk after controls — the level remaining after enforcement action, benchmarked against a tolerable threshold

The methodology behind every tool

Every ResidualRisk.ai tool is built on a single framework: P×C risk scoring. The goal is never to inspect everything — it is to ensure that the most dangerous non-compliance is always found first, and that limited enforcement resources are concentrated where they reduce risk most effectively.

This methodology was developed within the UNECE Group of Experts on Risk Management in Regulatory Systems over fifteen years of intergovernmental work, and is codified in UNECE Recommendations R, S, and V and in the ITC/UNECE Guide for Border Regulators. It has been applied across food safety, consumer product safety, customs, plant protection, and financial compliance.

The AI layer does not replace the methodology — it operationalises it at a scale and speed that was previously impossible for most regulatory authorities.

UNECE Recommendation R UNECE Recommendation S UNECE Recommendation V ITC/UNECE Guide 2022 ISO 31000 aligned

Five Tools. One Framework.

AI-powered enforcement across the product lifecycle

Each tool addresses a specific enforcement challenge. All share the same P×C risk logic, ensuring that outputs are consistent, traceable, and defensible — not black-box predictions.

Tool 01

AI for Establishing a Risk-Based Surveillance Framework

Most enforcement authorities operate without a systematic framework for deciding what to check, when, and how. Inspections are driven by habit, political visibility, or available capacity — not by risk. This tool uses LLM-based analysis to design the processes, criteria, and data flows of a risk-based surveillance system tailored to the authority's mandate and product scope.

  • Maps regulatory objectives to enforcement processes
  • Designs risk-scoring criteria specific to product categories and market context
  • Operationalizes complaint handling, document inspection, and sampling workflows
  • Produces a documented, auditable framework ready for institutional adoption
Regulatory grounding: Implements UNECE Recommendation R — Managing Risk in Regulatory Frameworks
Discuss this tool →

Tool 02

AI for Targeting Dangerous Non-Compliance

When an authority cannot inspect everything — and no authority can — it needs a principled method for deciding which products to check. This tool runs a P×C risk scoring pipeline across a product universe, generating a ranked inspection list where the most dangerous non-compliance surfaces first. Every score is accompanied by a traceable, evidence-based justification.

  • LLM evaluates consequence factors: hazard severity, exposure breadth, product use context
  • LLM evaluates probability factors: origin risk, market structure, compliance history
  • Generates continuous P×C scores and a Pareto-prioritised inspection list
  • Each output traceable — evidence cited, reasoning visible
Regulatory grounding: Operationalizes UNECE Recommendation S — Predictive Risk Management Tools for Targeted Market Surveillance
Discuss this tool →

Tool 03

AI for Prioritising How to Test a Suspicious Product

Once a product has been sampled, the enforcement authority faces a second resource constraint: laboratory testing is expensive, time-consuming, and often cannot cover every applicable regulatory requirement. This tool selects the optimal set of laboratory tests for a given product, maximising the probability of detecting non-compliance within the available testing budget.

  • Identifies the full set of applicable tests from regulatory requirements and product profile
  • Ranks tests by expected risk-reduction value using P×C logic
  • Selects the highest-value subset within capacity and cost constraints
  • Produces a documented, justifiable test plan for the inspector and the laboratory
Regulatory grounding: Extends P×C methodology to the enforcement decision layer; supports Recommendation S implementation
Discuss this tool →

Tool 04

AI for Integrated Risk Management in Border Control

Border control agencies operate in silos. Customs clears a shipment. Plant protection checks phytosanitary certificates. Food safety reviews labelling. Each agency applies its own risk criteria, often to the same consignment, without coordination. This tool implements a multi-agency P×C pipeline: each agency scores the shipment independently, and a coordinated inspection plan resolves overlapping priorities into a single, efficient enforcement action.

  • Builds agency-specific risk scoring pipelines (food safety, plant protection, customs, SPS)
  • IRM Coordinator layer routes shipments to relevant agencies automatically
  • Pareto-based prioritisation within each agency queue
  • Produces a coordinated inspection plan — who does what, concurrently
Regulatory grounding: Operationalizes UNECE Recommendation V — Addressing Product Non-Compliance Risks in International Trade; Single Window compatible
Discuss this tool →

Tool 05

AI for Regulating Online Markets

Online marketplaces have transformed product distribution — and the enforcement challenge that comes with it. Unsafe products can be listed, sold, and replaced faster than traditional surveillance can respond. This tool deploys agentic AI to continuously monitor online market activity: scraping platforms, building non-compliance profiles for specific products, and feeding intelligence back into the risk targeting pipeline.

  • Agentic AI autonomously searches and profiles product listings across platforms
  • Builds evidence-based profiles of online non-compliance patterns per product
  • Outputs feed directly into the P×C targeting pipeline as probability signals
  • Supports both market surveillance authorities and platform compliance functions
Regulatory grounding: Supports UNECE online marketplaces project; compatible with EU Product Safety Pledge monitoring requirements
Discuss this tool →

Delivered in the field

Projects across 20+ countries

The methodology behind ResidualRisk.ai has been implemented in government systems, tested at live borders, and validated through international peer review — not just demonstrated in slides.

🇧🇹 Bhutan

ITC / STDF

AI-assisted integrated border risk management pilot for agri-food products — multi-agency pipeline covering BFDA, Customs, and Border Police at Phuentsholing crossing.

🇰🇪 Kenya

ITC / EU-EAC MARKUP II

Multi-agency IRM pilot design for JKIA: KEPHIS, Port Health, KRA Customs, and KenTrade — covering fresh horticultural exports.

🇺🇬 Uganda · 🇷🇼 Rwanda · 🇸🇸 South Sudan · 🇧🇮 Burundi

ITC / EU-EAC MARKUP II

Border risk management capacity assessments and agency roadmaps across four EAC countries, covering customs, food safety, plant protection, and standards bodies.

🇺🇿 Uzbekistan

ITC

Integrated risk management assessment and implementation roadmap for border control agencies.

🇮🇱 Israel

Ministry of Economy · Ministry of Agriculture · Food Safety Authority

Risk-based import compliance and market surveillance frameworks across consumer products, fresh produce, and food. Plant protection methodology recognised as international best practice by NAPPO.

🇬🇷 Greece · 🇱🇻 Latvia

OECD / EU-funded

Market surveillance capacity building — training materials, inspector training, and comparative EU framework analysis.

🌍 UNECE / Global

UNECE WP.6 RAMS

Vice Chair since 2010. Led development of five intergovernmental Recommendations. Demonstrated AI-powered enforcement tools at the RAMS Annual Forum 2026.

🇲🇾 Malaysia · CEFTA countries

ITC

Risk management training courses for regulatory authorities — technical regulation and integrated risk-based import compliance frameworks.

Valentin Nikonov

PhD · Vice Chair, UNECE RAMS · Lead Expert, ITC

20+
countries delivered
15
years at UNECE WP.6
5
UNECE Recommendations led

I design and implement AI-driven, risk-based regulatory compliance and enforcement frameworks — helping governments and businesses build systems that identify the most dangerous non-compliance, whether in products, transactions, or supply chains, target limited resources effectively, and reduce non-compliance risk to a tolerable level without creating unnecessary costs.

Since 2010, I have developed intergovernmental risk management methodologies within the UNECE Group of Experts on Risk Management in Regulatory Systems, including methodologies for risk-based targeting and integrated risk management in border control. I currently serve as Vice Chair of the Group of Experts on Risk Management and Market Surveillance (RAMS), having led the development of foundational UNECE Recommendations and publications in this field.

As an international expert at ITC, OECD, UNIDO, and UNECE, and as an independent consultant, I have delivered projects in more than 20 countries covering food safety, consumer product safety, plant protection, border control, autonomous vehicle safety, and financial compliance, including fraud and AML.

Get in touch

Discuss a pilot or project

Whether you are scoping a technical assistance programme, designing a compliance system, or looking to pilot AI-powered risk targeting in your agency — start with a conversation.

Responses within 2 business days. No sales calls.