Personal cloud, termed as the difference in air pollutant concentrations between breathing zone and room sites, represents the bias in approximating personal inhalation exposure that is linked to accuracy of health risk assessment. This study performed a two-week field experiment in a naturally ventilated office during the COVID19 pandemic to assess occupants' exposure to common air pollutants and to determine factors contributing to the personal cloud effect. During occupied periods, indoor average concentrations of endotoxin (0.09 EU/m3), TVOC (231 mu g/m3), CO2 (630 ppm), and PM10 (14 mu g/m3) were below the recommended limits, except for formaldehyde (58 mu g/m3). Personal exposure concentrations, however, were significantly different from, and mostly higher than, concentrations measured at room stationary sampling sites. Although three participants shared the same office, their personal air pollution clouds were mutually distinct. The mean personal cloud magnitude ranged within 0-0.05 EU/m3, 35-192 mu g/m3, 32-120 ppm, and 4-9 mu g/m3 for endotoxin, TVOC, CO2, and PM10, respectively, and was independent from room concentrations. The use of hand sanitizer was strongly associated with an elevated personal cloud of endotoxin and alcohol-based VOCs. Reduced occupancy density in the office resulted in more pronounced personal CO2 clouds. The representativeness of room stationary sampling for capturing dynamic personal exposures was as low as 28% and 5% for CO2 and PM10, respectively. The findings of our study highlight the necessity of considering the personal cloud effect when assessing personal exposure in offices.