Speaking at Fortune’s Brainstorm AI conference in San Francisco on Tuesday, Saptharishi described AI capabilities like real time language translation and image recognition that can change the way first responders react, overcoming things like language barriers on 911 calls. Saptharishi estimated that saving 60 seconds from the average 911 response time could save 10,000 lives per year.  

Saptharishi spoke alongside Accenture AI head Lan Guan on a panel about the challenges facing organizations as they seek to expand AI efforts from small test projects to widespread deployment. Those include everything from concerns about spiraling costs to worries about the accuracy of the technology. 

As companies consider scaling AI, Guan discussed the value of measuring the performance of a human doing the job with the assistance of AI, and then comparing it to the performance of the human doing the job on their own, and against the AI doing the job on its own. 

Guan said it’s also crucial to measure quantitative wins—like time saved by 911 operators and a decrease in steps in a workflow. 

Motorola’s vision for AI in public safety goes beyond time saved on calls and into the tricky work of public surveillance. 

In Saptharishi’s example, a parent gives a 911 operator a description of what their child was wearing when they went missing. That description could be immediately “pushed to every single camera in that locality” and “automatically distributed to law enforcement.” Motorola supplies police departments, schools, and more with video technologies.

“Everything the cameras do is built in,” Saptharishi added, meaning that cameras with those data analysis and detection abilities don’t cost much more than a more basic or “dumb camera.”

It’s worth remembering that surveillance technologies, especially surveillance systems that could profile passersby based on physical descriptions, are controversial. Billionaire Larry Ellison caught criticism for predicting the rise of an AI-powered surveillance state. Earlier this year, Motorola and Massachusetts police were hit with a lawsuit alleging residents were unlawfully secretly recorded.

Saptharishi noted that descriptions from 911 calls such as those he described would be shared in a way that protects an individual’s privacy. While AI capabilities can be easily integrated into the cameras, the data is selectively processed, he said.

In his view, the future of “sensory infrastructure,” is a combination of generative AI capabilities plus more traditional technologies. And privacy is just one of many societal issues that become more critical as AI technology and sensory infrastructure is deployed at scale.

Guan noted that companies adopting new AI tech should consider the qualitative impact, like whether workers are satisfied and if the public trusts the new tools.

Read more coverage from Brainstorm AI:

Amazon’s top AI exec says industry concerns that LLMs hit a ‘wall’ are overblown, says Jeff Bezos ‘very involved’ in AI efforts

Pitching investors is like the NFL draft says Colin Kaepernick — ‘it only takes one’

Microsoft’s biz dev boss on the big bets that could define the future of AI