AIA and Building Safety 24-04-2026

AIA AND BUILDING SAFETY 24-04-2026

Meeting Summary

Workshop: AI and Building Safety (Workshop No. 5)
Date: 24 April 2026
Duration: 2 Hours 46 Minutes

Workshop Objective

The workshop explored how Artificial Intelligence (AI) can support building safety, fire safety, facilities management, and competency management while emphasizing that AI must complement—not replace—professional expertise. The session brought together professionals from universities, healthcare, construction, and building safety to discuss practical applications of AI, current limitations, and the importance of human oversight.

Session 1 – AI and Building Safety

Speaker: Mark O'Reilly

Purpose of the Presentation

Mark explained that AI is rapidly becoming part of construction, facilities management, competency assessment, and building safety. His goal was to help professionals understand:

  • what AI can realistically do,
  • where it should be used,
  • where it should not be trusted without verification,
  • and why professional judgement remains essential.
  1. AI is a Productivity Tool, Not a Replacement for Professionals

Mark described AI as a technology that dramatically improves productivity rather than replacing experienced professionals.

He explained that tasks which previously required weeks or months can often be completed within hours by combining AI with professional expertise.

Examples included:

  • competency framework creation,
  • report generation,
  • document drafting,
  • structured data analysis,
  • project planning,
  • inspection templates.

However, every AI-generated output must still be reviewed by a competent person before being relied upon.

  1. Types of AI

The presentation distinguished between two broad categories.

Predictive AI

Used for:

  • analysing historical data,
  • identifying patterns,
  • predicting risks,
  • supporting building management systems,
  • maintenance planning,
  • CAFM systems.

Generative AI

Used for:

  • writing reports,
  • drafting documents,
  • producing inspection reports,
  • creating presentations,
  • summarising information,
  • assisting with project documentation.

Most professionals are currently interacting with Generative AI rather than building their own AI models.

  1. Popular AI Platforms

The workshop compared several commonly used tools.

Microsoft Copilot

  • tightly integrated with Microsoft 365,
  • suitable for organisations,
  • works inside corporate SharePoint environments,
  • offers stronger data security.

ChatGPT

  • versatile,
  • excellent reasoning capability,
  • useful for drafting content,
  • limited native document generation.

Claude

Mark stated that Claude had become his primary AI tool because it supports:

  • long-term projects,
  • document management,
  • automated workflows,
  • Word and Excel generation.

Gemini

Google's equivalent to Copilot for organisations using Google Workspace.

Perplexity

Particularly valuable for research because it cites information sources, making verification easier.

  1. Human-in-the-Loop (HITL)

One of the strongest themes throughout the presentation was Human-in-the-Loop (HITL).

Mark repeatedly stressed that AI must never operate completely independently in building safety.

Instead:

  1. AI produces draft work.
  2. A qualified professional reviews it.
  3. Corrections are made.
  4. Only then should work continue.

He described this as gated workflows, where every important stage requires human approval before progressing to the next stage.

  1. AI Hallucinations

The presentation warned about AI "hallucinations."

These occur when AI confidently produces incorrect information that appears convincing.

This is particularly dangerous for:

  • fire safety,
  • risk assessments,
  • compliance reports,
  • regulatory documentation.

Mark noted that AI may produce outputs that are "99% correct," but the remaining 1% could involve a serious safety or legal issue.

Because of this, every AI-generated document requires professional validation.

  1. AI in Building Safety Today

Several existing AI-enabled tools were discussed, including systems supporting:

  • CCTV-based safety monitoring,
  • passive fire compliance,
  • building inspections,
  • competency management,
  • BIM data analysis,
  • digital twins,
  • Golden Thread information.

The speaker explained that AI is already embedded within many commercial software products, even if users do not always recognise it.

  1. AI and BIM

The workshop discussed the growing relationship between AI and Building Information Modelling (BIM).

Key points included:

  • AI can analyse BIM models.
  • AI can identify competency requirements based on building systems.
  • AI supports digital twins.
  • Well-structured BIM information produces better AI outputs.
  • Poor-quality data leads to poor AI performance.

The discussion reinforced that structured information is essential for reliable AI applications.

  1. Context Engineering

Mark introduced the concept of Context Engineering, explaining that it is more advanced than simply writing better prompts.

Instead of asking isolated questions, professionals should create AI environments containing:

  • project documents,
  • standards,
  • objectives,
  • workflow rules,
  • sequential review processes,
  • role definitions.

Providing AI with richer context produces significantly more reliable outputs.

  1. Competency and AI

The presentation linked AI to competency requirements under the Building Safety Act.

Future building professionals will increasingly need competency not only in construction but also in:

  • understanding AI,
  • reviewing AI outputs,
  • validating AI-generated information,
  • managing AI-assisted workflows.

AI itself cannot be considered competent.

Only qualified professionals remain accountable for decisions made using AI.

Discussion Following the Presentation

Anthony Taylor emphasised that professionals must actively challenge AI-generated documentation rather than accepting it at face value.

George Stevenson shared experiences of using AI to audit O&M and handover information. He agreed that breaking complex processes into smaller AI-assisted tasks, combined with human verification, is essential for reliable outcomes.

Participants also discussed competency frameworks, knowledge management, and the need for AI literacy throughout the construction and building safety sectors.

Session 2 – Fire Door Compliance

Speaker: Alan Oliver

Alan Oliver focused on improving fire door compliance within healthcare buildings, drawing on lessons learned since the Grenfell Tower fire.

  1. Fire Door Inspection Framework

Alan described the three inspection categories developed through NAHFO guidance.

Type 1 Inspection

Inspection of newly installed fire doors to verify compliance with fire test evidence.

Type 2 Inspection

Assessment of existing fire doors where historic documentation may be unavailable, determining whether the door remains suitable for its location.

Type 3 Inspection

Routine periodic inspections following initial assessment to ensure continued performance.

  1. Reference Documents

Alan explained that several guidance documents have been produced covering:

  • healthcare,
  • housing,
  • universities,
  • fire dampers,
  • fire stopping,
  • cavity barriers.

These documents are available free of charge to encourage wider adoption across the sector.

  1. Continuing Compliance Challenges

Despite the availability of guidance, Alan expressed concern that many organisations still fail to install fire doors in full accordance with fire test evidence.

Issues identified included:

  • inconsistent installation quality,
  • inadequate inspections,
  • weaknesses in third-party certification,
  • organisations obtaining certification without consistently applying required standards.
  1. Hospital Fire Door Compliance Pledge

Alan introduced a proposed pledge intended for launch ahead of the tenth anniversary of the Grenfell Tower fire.

The pledge would commit organisations to:

  • use certified fire door sets,
  • employ accredited installers,
  • undertake robust handover inspections,
  • ensure competent sign-off,
  • publish annual compliance audits,
  • strengthen the Golden Thread of fire door information.

The pledge aims to improve accountability and transparency across NHS estates.

Key Messages from the Workshop

  • AI is becoming an important tool for building safety and facilities management.
  • AI should improve productivity but must never replace competent professionals.
  • Human review remains mandatory for all safety-critical outputs.
  • Well-structured data and BIM models significantly improve AI performance.
  • Competency in AI will become an increasingly important professional skill.
  • Fire door compliance continues to present major industry challenges despite existing guidance.
  • Stronger inspection processes, accredited installers, and transparent compliance reporting are needed to improve building safety.