Winnipeg examining using AI for non-emergency police calls, 311

In the coming months, when you call the Winnipeg Police Service's non-emergency line, the voice that answers the call may not be human.

That's because artificial intelligence is coming to the City of Winnipeg.

"We're excited to release six pilot projects using artificial intelligence to improve customer service for the citizens in Winnipeg and help our staff become more efficient," said Tyler Gooch, the city's director of innovation and technology.

The city will be testing the use of AI to evaluate building permit applications, monitor potholes and road conditions using cameras mounted on buses, and automate invoice processing.

The city also wants to start using AI to answer non-urgent calls to police.

"The non-emergency line would be answered by an automated voice agent using generative AI to produce results similar to what a human agent would," Gooch said. "These are going to be the types of calls that could be administrative in nature, or something that a human could follow up on later."

There's only five to 12 people fielding both 911 and non-emergency calls for the Winnipeg police. Insp. Gord Spado with the WPS communications division said there's more than 300,000 non-emergency calls received on average each year (though about half are redirected to an automated response).

Spado said the use of AI could free up these 911 operators to deal with emergencies.

"We already sort of put some of the non-emergency things on hold, even if you're mid-call, so that we can answer those 911 calls where there are life and death issues," he said.

"We're aware that there's other agencies across Canada and the US that are currently exploring the use of AI for non-emergency calls, and it's something that we're excited to explore ourselves."

The program could then be expanded to other city call centres like 311 and Transit. The city also plans to test an AI chatbot to field 311 inquiries.

David Gerhard, the head of the University of Manitoba's computer science department, said he's glad to hear the city is starting to use artificial intelligence, but warns there are risks.

"It's appropriate to be concerned. These models have come onto the scene very quickly and have been deployed widely without a lot of consultation, and are making mistakes," he said.

He pointed to various mishaps when corporations have used AI to interact with the public.

"Because these new AI models can hallucinate or make things up or lie, the city needs to understand that they're going to be responsible for what this chatbot says."

But on the other hand, Gerhard said AI is incredibly powerful and can be a useful tool.

"So it would be irresponsible not to explore the use of these models in appropriate contexts."

He said it’s important humans are still a part of the process.

"In the end, the responsibility of any of these decisions should still be on a human," he said. "As long as the city is planning to implement it in that way, then we can be confident that the decisions that are made are still reliable and accurate."

When it comes to job security, Gooch said the city has no intention of replacing the humans answering the phones.

"We're learning what the technology can do and thinking about it as a force multiplier to help people become more effective and efficient."

Gooch said details will be finalized later this year, including launch dates. The city noted projects may be changed or cancelled as it learns more about the benefits and risks. Each project comes with a $75,000 budget and will go through planning, procurement, analysis before being implemented and deployed.