AI-Enabled Indoor Air Quality Monitoring for One of the Biggest Banks in Africa

Introduction

Indoor air quality moved from a facilities concern to a public health priority after the COVID-19 pandemic. For financial institutions with dense, air-conditioned open-plan floors, poor indoor air carries measurable cost: impaired cognitive performance, absenteeism, and growing disclosure obligations under ESG frameworks.

This trial set out to establish whether continuous, AI-interpreted environmental sensing could identify specific airborne hazards inside a working bank facility, not merely record that conditions were 'acceptable', and translate those findings into interventions the estate team could act on.

Methodology

The trial followed a four-pronged approach: data collection from deployed sensors, pattern recognition by the AI engine, near real-time reporting through dashboards, and targeted mitigation recommendations.

Non-intrusive, event-driven sensors measured temperature, humidity, particulate matter (PM1, PM2.5, PM4, PM10), gaseous pollutants (CO, CO₂, VOCs) and spectral signatures. Spectral analysis is what allows individual compounds to be named rather than reported as an undifferentiated VOC index.

• Kitchen. 333,974 observations
• Office floor. 340,336 observations
• Fresh air shaft. 316,958 observations
• Total. 991,268 observations over three phases spanning 2.5 years

Trial structure

• Phase 1, telemetry collection to establish baseline metrics for each zone
• Phase 2, risk assessment development, deriving Pathogen Risk Index and IAQ Index
• Phase 3, forensic analysis to attribute contamination to sources

Results: carbon dioxide

Average concentration was reported at 2.25% of the measurement scale, approximately 755 ppm, classified as normal. Levels were consistent across observation hours with no significant peaks.

While within normal bands, this sits close to the range at which measurable declines in decision-making performance begin to appear, which is why the study still recommends ventilation improvement rather than treating CO₂ as resolved.

Results: particulate matter

The pattern matters more than the headline: coarse particle control was working while the finest, most health-relevant fraction was passing through. A filtration specification written against PM10 alone would have looked compliant.

• PM1, high peaks indicating fine particle accumulation
• PM2.5, elevated levels posing respiratory risks
• PM4 and PM10, normal levels, suggesting existing filtration handles coarse particles effectively

Results: hazardous compounds identified

Spectral signature analysis identified specific compounds and graded them by concentration. The following were the most significant.

Environmental KPIs

• IAQ Index, varied by location; the top 20 observation hours showed stable conditions
• Pathogen Risk Index, normal for CO₂, PM4 and PM10; elevated for PM2.5 and VOCs

Discussion

Elevated PM2.5 and VOC levels, particularly formaldehyde and benzene, indicate significant health risks and necessitate immediate mitigation. CO₂ remained normal but approaches thresholds that may impair cognitive function, suggesting ventilation improvements are warranted.

The ability to identify these risks in real time, through spectral signature analysis, offers a competitive advantage over traditional monitoring systems that report only aggregate indices.

Limitations

• Single location, limiting how far findings represent facility-wide IAQ
• Specific VOC compounds not fully detailed, restricting source attribution
• Effectiveness of proposed mitigations not quantified within the trial
• Seasonal variation and occupancy patterns not analysed

Conclusions and recommendations

The trial establishes that continuous AI-interpreted sensing surfaces hazards that periodic spot-check audits do not. For the bank, acting on the findings supports ESG compliance and occupant well-being, reduces absenteeism, and positions the institution ahead of tightening disclosure expectations.

• Expand sensor deployment across facilities
• Implement air purifiers, enhanced ventilation and low-VOC materials
• Integrate real-time dashboards for continuous oversight
• Conduct longitudinal studies evaluating mitigation effectiveness