Transforming BIM engineering and code compliance with generative AI

In this article, Oh Qi Qi, Exyte Subject Matter Expert for Artificial Intelligence (AI) and regulation, examines compliance challenges in pharma facility design, outlines how the ReguBIM AI pilot explores possible solutions, and discusses what early results suggest for future engineering workflows.

The challenge of compliance in pharma facility design

Designing facilities for pharma and biotech companies requires adherence to strict regulatory frameworks, commonly referred to as code compliance. These regulations aim to protect building users and safeguard manufacturing environments that directly affect patient safety. They address fire safety, airflow systems, cleanroom spacing, and access routes such as exits and service roads.

Today, compliance verification is largely manual. Engineers cross-check 3D Building Information Modeling (BIM) data against extensive, text-based regulations. BIM models, often created in Revit or similar tools, function as detailed digital representations of a facility, while regulations are published as long technical documents. Compliance reviews involve checking hundreds of requirements, from corridor widths to fire exit spacing, which requires time, experience, and careful interpretation.

This manual approach is widely seen as demanding in terms of effort, susceptible to human error, and expensive when issues are identified late in a project lifecycle.

Introducing ReguBIM AI as a prototype

ReguBIM AI is developed by Exyte as a prototype to explore whether generative AI can assist with compliance verification. Rather than replacing existing processes, the pilot acts as a digital assistant that is being tested for its ability to interpret regulatory text and relate it to 3D BIM data.

Within the testing environment, ReguBIM AI can analyse selected regulations, identify potentially relevant clauses, and flag areas in a BIM model that may require closer human review. Outputs are intended to support engineers during early design phases and internal checks, not to provide final compliance approval.

The prototype runs on Databricks infrastructure and uses a medallion architecture to structure data processing. ETL pipelines convert BIM files and regulatory documents into formats suitable for analysis. The application is deployed as a Databricks App on serverless infrastructure to allow controlled experimentation and iteration as part of the pilot setup.

ReguBIM AI remains in an early stage of development and is used internally to evaluate technical feasibility, data quality requirements, and integration effort.

Core capabilities under testing

As part of the ongoing pilot, ReguBIM AI focuses on several experimental capabilities:

• Connecting selected BIM elements with relevant regulatory text to support assisted compliance checks
• Using an AI-based architecture built on Databricks, MLflow, and orchestration components for modular experimentation
• Performing spatial analysis such as distance checks and route evaluation within defined test scenarios
• Applying retrieval-augmented generation to surface regulatory clauses that may relate to user queries

• Producing draft compliance summaries intended for expert review rather than formal submission

   

These capabilities are subject to change as the prototype evolves and are not validated for production use.
 

Observations from early pilot work

Initial testing indicates that AI-assisted compliance analysis could reduce the effort required to identify potential issues in early design stages. However, results also show that expert judgment remains essential, particularly when interpreting ambiguous regulations or project-specific requirements.

From a business perspective, the pilot helps Exyte assess where automation may support engineers, where data preparation remains a constraint, and how such tools could fit into established governance and approval processes.

Digital innovation at Exyte

Exyte is a global Engineering, Procurement, and Construction firm focused on complex high-tech facilities. Alongside its core engineering activities, the company evaluates digital tools that may support quality, safety, and efficiency. ReguBIM AI is one such proof of concept, developed to test how generative AI might contribute to regulatory analysis in the future.

Exyte views this prototype as a collaborative support tool rather than a decision-making system. The intent is to help engineers spend less time on repetitive checks and more time applying expertise where it matters most.

Smarter compliance, explored through testing

Through ReguBIM AI, Exyte is examining how BIM, generative AI, and regulatory data could work together in a controlled pilot setting. While still experimental, the initiative provides practical insights into how compliance processes might evolve and what safeguards are required before any broader application.

For pharma and biotech projects facing complex regulatory environments, such exploratory work helps clarify both the opportunities and the limits of AI-supported engineering.