AI-Powered Geotechnical Compliance Platform

Every ground investigation produces borehole logs containing detailed records of the soil and rock layers beneath a site. Before those records can be trusted to support foundation design, slope stability assessments, or other engineering decisions, every geological description must be reviewed against multiple overlapping standards, including BS 5930, ISO 14689, TP208, CIRIA C570, internal procedures, and Chalk-specific guidance. 

A single borehole log can contain 20 to 50 stratum descriptions, each requiring manual validation of geological terminology, colour, consistency, spelling, material descriptions, and cross-consistency with supporting test results such as SPT N-values. 

This process created several challenges: 

• Manual reviews slowed project delivery and created engineering bottlenecks. 

• Different reviewers could reach different conclusions on the same borehole log. 

• Compliance reviews depended on the limited availability of experienced engineers, making the process difficult to scale. 

Although an earlier proof of concept demonstrated the potential of AI-assisted reviews, it relied on US-hosted infrastructure, third-party AI providers, and generic prompting that could not consistently deliver accurate, reliable, or auditable results while meeting strict UK data residency and IT compliance requirements. 

The challenge was to build a platform that combined AI with engineering best practices to deliver secure, explainable, and production-ready compliance reviews. 

Methodology

Rather than relying solely on generative AI, the solution was developed using a hybrid engineering methodology that combines deterministic validation with AI-assisted reasoning. The objective was to ensure compliance reviews remained accurate, explainable, consistent, and suitable for production use in a regulated engineering environment. 

An Agile development approach was adopted to design, develop, test, and refine the platform through four focused sprints over eight weeks. Throughout development, the platform was continuously validated against recognised geotechnical standards while security, governance, and UK data residency were built into the solution from the outset. 

This methodology ensured that objective compliance checks were handled through deterministic rules, while AI reasoning was applied only where engineering judgement was genuinely required, creating a reliable and scalable review process.  

The Solution 

The completed platform transforms manual borehole log reviews into a secure, AI-assisted compliance process that delivers faster, more consistent, and standards-driven assessments. 

The solution follows a structured four-stage review process: 

Intelligent Document Extraction 

Azure AI Document Intelligence extracts borehole log information while preserving the relationship between depths, geological descriptions, and associated test results. If information is unclear or ambiguous, it is flagged for review rather than inferred. 

Automated Rule Validation 

Deterministic validation identifies spelling mistakes, missing mandatory information, geological shorthand, and inconsistencies between geological descriptions and SPT N-values before AI analysis begins. 

AI-Powered Standards Compliance 

Using Retrieval-Augmented Generation (RAG), the platform retrieves the relevant clauses from BS 5930, ISO 14689, TP208, CIRIA C570, Chalk-specific guidance, and client-specific standards before Azure OpenAI performs the compliance assessment. Every finding references the applicable clause, ensuring recommendations are grounded in recognised standards. 

Structured Reporting 

Each finding is classified as Critical, Moderate, or Minor using a client-defined assessment framework. The platform then generates downloadable PDF and CSV reports detailing findings together with the standards and versions applied during the review.  

Built for Enterprise Trust 

For geotechnical engineering, compliance reviews must be accurate, transparent, and defensible. The platform was therefore designed with enterprise governance and long-term operational trust at its core. 

Key capabilities include: 

• UK data residency, with all processing and AI services operating within Microsoft Azure UK South inside the client’s own Azure tenant. 

• Grounded, explainable AI, where every finding references the relevant clause from the applicable engineering standard. 

• Immutable seven-year audit logging for uploads, reviews, and administrative changes, ensuring every action remains fully traceable. 

• Client-controlled standards management, allowing administrators to upload, update, retire, and prioritise standards without requiring software releases. 

• Enterprise-grade security, including Microsoft Entra ID integration, single sign-on (SSO), role-based access control (RBAC), multi-factor authentication (MFA), encryption in transit and at rest, and private networking across Azure services. 

Results 

• Reviews completed in under 90 seconds per borehole log, replacing manual reviews that could occupy senior engineers for significantly longer. 

• Over 95% finding-level accuracy, with a target of zero critical compliance issues missed, representing a significant improvement over the original proof of concept, which achieved approximately 80% accuracy. 

• Consistent, repeatable assessments, ensuring the same borehole log produces the same result regardless of the reviewer. 

• Management dashboards providing visibility into recurring issues, engineering trends, and training opportunities across teams. 

• A fully operational, security-hardened platform ready for enterprise deployment.