Halton Police Are Using AI to Hire and Train 911 Call Takers

The person who picks up when a Burlington resident calls 911 trained for close to a year before taking a live call. Halton police are now using artificial intelligence to work out much earlier who will make it through that year.

The service is running software from GovWorx, an American company, that builds simulated emergency calls for applicants to work through. The simulations test decision making, communication and how someone holds up in a crisis. Halton police say the AI screening cuts candidate evaluation time in half, and that they are the first Canadian agency to use the technology.

Behind it is a staffing problem. Brian Dodd, who manages the Halton police 911 communications bureau, says not every candidate who gets hired makes it through training. "The training takes quite a long time, takes close to a year," he said. The same software is used during training and will be used to evaluate communicators once they are working. Halton has 75 dispatchers and call takers now and wants to add between four and eight this year.

A second system is already answering phones. SARA AI takes certain non-emergency calls, and police say it has handled 95,367 calls so far this year, spent 1,195 hours talking to residents, and sent more than 2,470 text messages pointing people to the right place to report an incident, such as the collision reporting centre.

Moving those calls off the main line cut almost two seconds from the average 911 wait. The North American standard is 90 per cent of 911 calls answered within 15 seconds, a mark Halton says it already meets.