The article presents the results of research conducted during trial evacuations from high-risebuildings. Based on the surveys, areas were identified where the evacuation process requiressupport and an alarm model was proposed to shorten the total evacuation time.Two series of trial evacuations from buildings were carried out, involving 1,333 respondents.The survey was carried out in 28- and 24-storey office buildings, where companies with differentbusiness profiles, e.g. consulting, accounting, media and others, were based. The questionnaireincluded 16 questions and was distributed in printed form to each participant of the evacuationattempt, and the completed questionnaire was collected shortly after returning to the office. Thequestions posed in the survey concerned, among others, the comprehensibility of the alarmmessage, concerns and other feelings during evacuation, the influence of other people at thestart of evacuation, knowledge of people who might need help, knowledge of people responsiblefor evacuation, estimated time devoted to preparation for evacuation, description of additionalactivities performed, preferences for choosing the evacuation method between stairs and elevator,general opinion on the need to carry out exercises, demographic data, etc.Interviewees confirmed that those taking part in the exercise may have doubts about the need tostart an evacuation. Intensifying the frequency of evacuation drills increases employees’ reluctanceto participate in them. It was confirmed that the alerting process is a phenomenon of socialinteraction, and to ensure correct alerting it is necessary to use the direct influence of peopleresponsible for the evacuation on other employees to communicate messages in as much detail aspossible. People who were not familiar with the building were characterized by a higher level ofnegative feelings about the evacuation process and a longer reaction time to the alarm.The use of a voice alarm system in a building guarantees high intelligibility of messages. A modelof evacuation messages was proposed to reduce doubts as to the need of starting an evacuation, dedicated to individual areas of the building, enabling shortening of the total evacuation time. Themodel is clear in its description of the threat and takes into account various possibilities of locatingpeople in relation to this threat.