What if the person who answers your town’s non-emergency line at 2 a.m. isn’t a person at all? In Penobscot County, Maine, an AI 911 dispatch assistant built on Amazon Web Services technology is already doing exactly that. It screens routine calls so human dispatchers can stay focused on real emergencies.
The tool never touches an actual 911 call. It only lives on the non-emergency line, offering menu options, routing to a live dispatcher when needed, or texting a link to file a report online.
How This AI 911 Dispatch Tool Actually Works
That distinction matters more than it sounds. Only six of Maine’s 22 dispatch centers were fully staffed as of a 2024 study. Penobscot County alone fields more than 60,000 real 911 calls a year, plus over 400,000 non-emergency ones. Every extra call ties up a dispatcher who could be answering someone in real trouble instead.
Brad Fitzgerald, who directs the Penobscot County center in Bangor, calls it “a phone tree to give people some options and prompts.” He is careful about what it isn’t: “We still have to treat every call as if it could be an emergency.” It just clears out calls that were never emergencies, so a dispatcher has time to focus when it counts.
Picture a 4 a.m. Call That Actually Gets Answered
Picture Denise, a home health aide in a rural stretch of the county. Her elderly client falls out of bed at 4 a.m., and Denise dials 911. She needs a human on the line immediately, not a hold message.
Now picture Marcus down the road, who just wants to report a downed tree branch. His call matters, but it isn’t urgent, and routing it through the AI is exactly what frees the dispatcher answering Denise’s call to give her client full attention.
The Numbers Behind the Faster Answer
Penobscot County is new to this, and Fitzgerald is honest that his own drop in call volume “hasn’t been astronomical.” Other agencies running similar systems for longer have clearer results to show for it.
| Agency | What Changed |
|---|---|
| Jeffcom 911 | 33% fewer administrative calls; 5%+ more 911 calls answered within 15 seconds |
| Monterey County, CA | 36% average drop in non-emergency call volume |
| Penobscot County, ME | New rollout; smaller but measurable relief on dispatcher workload |
Five percentage points sounds small until you remember what a 911 call actually is. During a heart attack or a stroke, every extra second on hold works against the caller. A dispatch center that answers more calls inside that critical first 15 seconds isn’t offering a convenience upgrade. It can be the difference between reaching help in time and not.
There’s a budget angle too. Hiring and training a new telecommunicator isn’t cheap, and small-town budgets rarely have room to spare. A tool that clears 30 to 36 percent of the routine call load buys back staffing capacity for free.
Why Nobody Is Handing This Off Entirely
Not everyone is comfortable moving fast here, and that caution is healthy. The National Telecommunications and Information Administration wants national standards on transparency, cybersecurity, and human oversight before AI spreads further into emergency dispatch.
“I don’t see us ever getting away from a human in the loop,” said Jeremy Hill, the NTIA’s public safety adviser.
A related industry briefing from Mission Critical Partners makes a similar point: AI can plow through routine, data-heavy work fast, but the last stretch of judgment calls still needs an experienced person. That’s exactly how Penobscot County is using it: as a filter, not a replacement.
Maine isn’t alone in modernizing how emergency calls get handled. Kentucky’s Bath County built a text-to-911 system so callers who can’t safely speak out loud can still reach a dispatcher, and Wyoming’s Teton County uses AI wildfire cameras that spot fires before a single 911 call comes in. None of these tools work in isolation. Together, they describe a 911 system quietly rebuilding itself around the calls that matter most.
If your dispatch center is short-staffed, and plenty are, ask your local government whether it’s exploring anything like this. The question isn’t whether AI belongs near 911. For actual emergencies, it clearly doesn’t. It’s whether AI belongs on the line right next to it, clearing space so a human can answer faster when it counts most.