Real-World Assurance for Real-World AI
Our audit and assurance services have helped organisations leverage the power of AI safely and responsibly, across 15+ industries across the world
Healthcare
When AI supports diagnosis, triage or care decisions, patient safety and clinical effectiveness are paramount — which is why healthcare AI attracts high-risk obligations and requires validation on real clinical data with clear human oversight.
Applications addressed: Diagnostics support; Medical devices; Companion tools.
Systems addressed: Predictive ML; Computer vision; GenAI for summarisation/UX.
Services provided: Audit; Bias testing; Impact assessment; Monitoring.
Recruitment
Hiring AI can produce unfair outcomes — tools that seem neutral may screen out some groups more often. That’s why bias checks and clear documentation are needed to meet rules like NYC Local Law 144 and other anti-discrimination laws.
Applications addressed: Target discovery; Trial design optimisation; Evidence generation.
Systems addressed: Predictive ML; Anomaly detection; NLP for claims.
Services provided: Audit; Explainability documentation; Governance playbooks.
Insurance
Pricing, underwriting and claims AI must balance fairness, explainability and consumer protection; supervisors emphasise governance, data quality and non-discrimination, so firms need auditable models and controls that withstand regulatory scrutiny.
Applications addressed: Underwriting; Pricing; Fraud detection; Claims triage
Systems addressed: predictive ML, computer vision
Services provided: audit; impact & risk assessment; bias testing; monitoring playbooks
Government
Public-service AI affects real lives—who gets benefits, housing, healthcare or support. Decisions on rights need transparency and recourse, with records that let people understand and challenge outcomes.
Applications addressed: Resource allocation; Risk scoring; Emergency services; Research.
Systems addressed: Predictive ML; Rules + ML hybrids.
Services provided: Audit; Community-led audits; Risk assessment; Governance playbooks.
Customer Operations
AI in customer care promises speed and scale, but unmanaged hallucinations, inconsistent answers, and privacy issues can erode trust and create regulatory exposure
Applications addressed: Chatbots; Agent assist; Knowledge retrieval; Call summarisation/routing.
Systems addressed: GenAI (LLMs); Retrieval-augmented generation; Intent classifiers.
Services provided: Audit; Assurance documentation; Prompt evaluation frameworks; Monitoring.
Culture & Media
GenAI is transforming creation but introduces copyright, provenance, and editorial consistency risks; teams need guardrails that protect IP, maintain standards, and mitigate misuse at scale.
Applications addressed: Content generation; Editorial assistance.
Systems addressed: GenAI (LLMs; Image models).
Services provided: Red-teaming; Copyright compliance assessment; Quality audits; Provenance testing.
Technology
Data lakes, MLOps and platform tools underpin AI at scale; but gaps in lineage, quality, security and governance flow into models. Documenting provenance and controls is key to reliable training, deployment and monitoring.
Applications addressed: Data lake build; Feature store; MLOps pipelines; Data quality monitoring.
Systems addressed: Data platforms; Model registries; Monitoring tooling.
Services provided: Data provenance documentation; Control design; Monitoring & evidence packs.
Emergency Services
Safety, triage and situational-awareness models must manage false negatives and false positives under time pressure; assurance focuses on calibrated thresholds, human-in-the-loop decision-making, and accountability for interventions.
Applications addressed: Incident detection; Dispatch prioritisation; Situational awareness; Resource allocation.
Systems addressed: Predictive ML; Computer vision; NLP; Decision-support.
Services provided: Audit; Threshold calibration; Human-in-the-loop design; Monitoring.
Life Sciences
If AI steers trial design or evidence on shaky data or methods, it can misguide research and future treatments. Teams must show results are reliable and fair, protect sensitive data, and meet tough medical regulations—or risk delays, recalls, or loss of approval.
Applications addressed: Target discovery; Trial design optimisation; Pharmacovigilance; Evidence generation.
Systems addressed: Predictive ML; Rules+ML hybrids.
Services provided: Audit; Community-led audits; Risk assessment; Governance playbooks.
Agriculture
In our African projects, crop outcomes affect food security and livelihoods; AI errors or poor generalisation can mean lost yield and income.
Applications addressed: disease detection, micro-climate forecasting, decision-support for agricultural inputs.
Systems addressed: predictive ML, computer vision
Services provided: audit; impact & risk assessment; bias testing; monitoring playbooks
Automotive & mobility
AI that assists or informs driving must cope with edge cases and human factors because misjudgement or over-trust can cause physical harm; assurance therefore aligns to ISO 26262 (functional safety) and SOTIF (hazards without a system failure).
Systems addressed: Predictive ML; Hybrid decision-support.
Services provided: Impact assessment & risk mitigation; Audit.
Climate & energy
Resilience planning depends on trustworthy forecasts; as climate baselines shift, AI models face non stationarity and distribution change, so organisations need traceable inputs and robust validation.
Systems addressed: Predictive ML
Services provided: Audit; Assurance documentation; Monitoring setup
Social Services
Here the stakes are resource allocation and harm prevention: false positives can stigmatise or misdirect scarce support, while false negatives miss genuine need; systems therefore require measurable harms frameworks, appeal routes, and ongoing monitoring to ensure equitable outcomes.
Applications addressed: Harm detection; Social care triage.
Systems addressed: Predictive ML; Moderation classifiers.
Services provided: Community-led audits; Governance & recourse design.
Transportation
From network optimization to demand forecasting, errors propagate into service reliability and safety; assurance addresses model drift, scenario coverage and human-centred operating procedures for real-world conditions.
Applications addressed: Demand forecasting; Route optimization; Scheduling; Load balancing.
Systems addressed: Predictive ML; Optimization/operations research; Simulation.
Services provided:Audit; Scenario coverage assessment; Monitoring.