Anyone who’s struggled with frustrating automated phone systems or unhelpful online AI assistants understands that simply adding AI doesn’t automatically make things better. While every organization should strive to make AI a benefit to their customers, implementing AI well is critical for law enforcement agencies.
Traditionally known as careful adopters of technology, many law enforcement agencies are embracing artificial intelligence to try and offset staffing shortages, budget pressure and increasing calls for service from their community. This rush to adopt increases the risk of AI disappointing both the agency and their community. In public safety, every interaction has real-world impacts. Poorly implemented AI slows users down, and may result in consequential decisions being made with insufficient data and user oversight.
After years of development work with the largest team of AI researchers and developers in the industry, Motorola Solutions recently launched Assist, their interactive AI platform, specifically to address the unique needs and challenges of AI in the public safety environment. Thoughtfully integrated at various points from call to case closure, Assist AI offers insights drawn from all of your content — voice, video and data — right in the application where you need them.
Let’s step through a typical public safety workflow and see how Assist can help specific users gain greater speed, efficiency and accuracy in their everyday activities.
AI in 9-1-1 Call Handling — Here to Assist
While we understand the challenges our 9-1-1 telecommunicators face — staffing shortages, training time for new hires, stress and mental health — improvements have been slow to arrive. Assist offers relief through automatic real-time transcription and translation of a 9-1-1 call, with keywords (such as location, names, and the nature of the incident) highlighted off to the side for simple, one-click navigation. This greatly reduces the cognitive load on the telecommunicator. They no longer need to keep all of the details of the conversation in their head; instead they can focus on keeping the caller calm and extracting the necessary details.
An ongoing summary of the call helps supervisors in the PSAP (Public Safety Answering Point) and, eventually, responders out in the field understand at a glance the nature and severity of the incident. With so much supplementary information available, it would be understandable that a telecommunicator might not find all of the key information… or, they may take additional time checking each data source. AI excels at this type of data gathering. Assist checks all available agency data sources, as well as publicly provided information via over-the-top customer data or location providers, to provide the telecommunicator a concise list of options for the location of the incident, nature of the incident, and the caller’s name. The telecommunicator remains in control, while the AI provides all available data in a fraction of the time it would take even the most experienced PSAP professional.
AI in Dispatch — Here to Assist
With the nature and severity of the call established, the appropriate resources must now be dispatched to resolve the incident. For some agencies, these are two different telecommunicators — one to answer 9-1-1 calls, and a second responsible for dispatching. In many agencies, by design or as a result of staffing challenges, it’s a single individual. Dispatching is a unique challenge, as these telecommunicators must split their attention between managing emergency incidents and providing general support and information to responders out in the field.
Assist AI helps dispatchers by simplifying and streamlining some of their most time-consuming activities. License plate lookups are a prime example. Officers in the field often call in for license plate checks when needed. As Assist monitors the transcript of the radio traffic, the mention of a license plate state and alphanumeric will prompt a one-click license plate lookup link for the dispatcher, putting the necessary information at their fingertips. Should the license plate match a stolen vehicle, a similar one-click link to add the plate to the license plate recognition (LPR) hotlist is presented. The dispatcher decides if and when either of these actions takes place.
Self-dispatched incidents — incidents initiated by an officer in the field when they spot something out of the ordinary, as opposed to a call to 9-1-1 — may constitute as much as 40% of an agency’s workload. With Assist monitoring the radio traffic transcription between an officer and the dispatcher, a recognition of sufficient incident elements results in a prompt within the transcript. The dispatcher can create the incident with a single click, with known fields already populated.
By leveraging these simple AI prompts, the telecommunicator can greatly reduce the amount of repetitive typing they do for everyday activities like license plate checks and self-dispatched incidents, reducing their stress level and freeing up time for more pressing emergencies.
AI in Patrol — Here to Assist
Responders in the field have both the most to gain and the most to lose from applying AI in their environment. Using AI to resolve incidents more quickly and spend more time engaging with the community provides obvious benefits in public perception and officer satisfaction. Poorly implemented AI, though, could lead to a cascade of frustration at best, and potential risk to the officer or a community member at worst.
With nearly one hundred years of history providing communication technology to first responders, Motorola Solutions understands the unique challenges field personnel face, and have the human-centered design knowledge to provide maximum benefit with minimal intrusion. APX NEXT and APX N series radios offer ViQi, our public safety voice assistant, to support officers in the field when it’s easier to talk than type. From basic capabilities like changing your radio channel and volume, to location confirmation, to advanced maneuvers like driver’s license, license plate and VIN checks, just ask ViQi. On our newest device, the SVX video remote speaker microphone, officers can also use ViQi to add an incident type to their camera recordings.
Our newest ViQi capability will bring translation capability to the field. Rather than calling in a coworker fluent in a particular language to help, responders interacting with a community member speaking a different language will utilize ViQi for two-way translation capability. The officer’s inquiries are translated into the appropriate language by ViQi, and then she translates the response back into English for the officer. Launching initially with Spanish translation capabilities, this provides another convenient Assist AI benefit on a device the officer always has within reach.
Few items are more closely associated with police officers than their notebooks. However, with Assist, pen and paper are no longer necessary. Using CommandCentral Responder, a mobile application for both Android and iOS devices, officers can record witness statements as well as their own incident narrative, which are transcribed and uploaded to the incident. The days of trying to decipher your own handwriting while hunched over a keyboard in your cruiser — or back at the office instead of out on the street — are over.
AI in Reporting — Here to Assist
Capturing information at the scene is only half the battle, as any law enforcement personnel can attest. Writing a clean, concise report that captures all of the incident information and passes supervisor review on first submission may be more difficult than the field work. Here again, Assist can help. Using the officer’s recorded narrative, Assist populates the fields in the incident report to give you a head start in completing the paperwork. But that’s just the beginning…
Leveraging all of the information available on the incident — 9-1-1 call transcripts, radio traffic transcripts, body camera and in-car camera recordings and transcripts, recorded witness statements, incident scene photographs, publicly provided information and more — Assist AI builds a timeline of the incident, with links to the pertinent content. It can even use the content to help the officer generate a draft of their narrative. Users can also proofread and polish their narrative, catching any grammar, spelling or transcription issues.
Perhaps most impressively, Assist also verifies the content of the report, identifying any mismatches between what the narrative states and what the incident media — 9-1-1 recordings, radio traffic transcripts, in-car and body-worn camera recordings, witness statements, photos from the scene, and more — actually shows. Inconsistencies are presented to the user, inviting the officer to correct the issue, add additional media or context or disregard, as appropriate. Of course, the original version of all content and subsequent versions are preserved for audit purposes.
AI in Police Investigations — Here to Assist
Resolving open cases — again, facing the universal agency challenge of not enough staff and too much work — often revolves around license plates. License plates locate vehicles of interest, which usually leads to a person of interest, and therefore, the potential of moving an investigation forward. However, license plates can be elusive, even in a camera-dense city environment. Without a license plate, many investigators are forced to fall back to canvassing nearby neighborhoods or hoping someone submits a tip.
Artificial intelligence reinvents the world of investigation. For Assist, a license plate recognition camera doesn’t just capture a license plate. It’s a picture of a vehicle, with identifying characteristics that the AI can categorize and retain. Even without a license plate, if the investigator can provide a make, model, and color, Assist can find it. Bumper stickers? Aftermarket accessories? Promotional writing? Damage? Assist can find that. The vehicle’s environment — on a highway, in a parking lot, near a forest preserve — Assist can find that.
Natural language search capabilities within Assist mean that investigators can simply describe the vehicle of interest, without a license plate, and find probable matches — with license plates — for additional investigation. By inverting the license plate recognition concept, Assist provides enormous benefits for frustrated investigators.
AI in Evidence Management — Here to Assist
When it comes to frustration, evidence managers rival investigators on a daily basis. With body cameras adopted by agencies large and small across the nation, an explosion in Freedom of Information Act (FOIA) requests has followed, flooding the inboxes of evidence staffers. In order to satisfy these FOIA requests, evidence technicians must redact any sensitive information before releasing it to the public.
Once a painstaking frame-by-frame process, AI revolutionizes the redaction experience with persistent object recognition. This means that an evidence technician identifies the item to be redacted – a face, a person, a vehicle, a screen, a license plate or a document – and the AI tracks and redacts that object, even if it goes offscreen for an extended period and then reappears.
With Assist, the transcription of the audio track from a video provides a simple way to navigate to key points in a video. Simply search for a phrase of interest in the transcript, click on that phrase, and the video will jump to that point for easy redaction. AI also recognizes over 50 forms of common, sensitive spoken information, including names, addresses, SSNs, health information, and more. Users can easily redact a single instance or all instances of these terms in the audio track.
With over 1.1 million FOIA requests in 2024 (up 30% since 2022, according to foia.gov), your evidence technicians need all the support they can get. Fortunately, AI from Motorola Solutions is here to Assist.
Our foundational beliefs about artificial intelligence
At Motorola Solutions, we believe purposefully applying AI can complement human decision-making and cultivate public trust while preserving individual rights.The most direct way to accomplish this is to always put the user of the technology and the community members they support at the center of each innovation. When building new technologies and innovating with AI, we hold ourselves to the highest ethical standards.
All of our AI technology is created to be equitable, transparent, understandable, accountable, secure and reliable. Additionally, we adopt four fundamental tenets to guide the development and use of AI:
- AI should be used to assist and accelerate human decision making, not replace it. Our AI systems are designed to be advisory in nature and to not take consequential actions on their own. Using AI-generated guidance, we can help customers make better decisions faster.
- We are applying proven AI technology in a focused and collaborative way to narrowly-defined and well-understood use cases. This enables users to accomplish their tasks faster and more successfully.
- We apply proven AI and machine learning technologies that are carefully trained, thoroughly characterized and extensively tested. This results in AI-powered solutions that are far less likely to behave unexpectedly in the field.
- The Motorola Solutions Technology Advisory Committee (MTAC) serves as a ‘technical conscience’ and advises the company on ethics, limitations and implications of specific product technologies. Additionally, we consult with objective third parties to provide an outside-in point of view to guide our decisions.
For more details on our commitment to responsibly applying AI at every step of our product delivery — from ideation to customer deployment — please review our position paper on The Responsible Use of Artificial Intelligence.
Here to Assist
If you’ve spent any time with commercial LLMs (large language models) like ChatGPT or Gemini, you’re aware of the risk of the AI generating false content in response to a question, commonly known as a hallucination. Annoying for a general user and problematic for a college student, hallucinations pose a much greater risk for public safety professionals.
AI models are prone to hallucinate when they lack sufficient data to draw a conclusion. In the absence of information, AI will create its own. The power of the Assist platform stems from the width and breadth of the Motorola Solutions portfolio. The amount of information available not only supports the user in terms of output, but also supports the AI in providing high quality insights with less risk.
Thoughtful execution is critical when deploying artificial intelligence in an industry, and particularly so for law enforcement. There is a world of difference between adding artificial intelligence to a product, and intelligently adding artificial intelligence to a product. We’re confident that we’ve accomplished this with Assist, and look forward to your feedback as you experience the capabilities of this new platform. We are all here to Assist.
