:format(webp))
Draft One is a new solution from Axon that force multiplies police officers by automatically drafting report narratives based on body-worn camera audio in seconds, saving an hour or more per shift. Less time spent on reports means more time spent responding to calls for service, reducing crime and supporting communities.
As Axon continues to develop solutions like Draft One that leverage AI, we want to make sure that attorneys are kept in the loop and their questions and concerns are addressed. We know that the reports police write impact the criminal justice process, and so we have worked closely with prosecutors across the United States to ensure we build a tool that both helps police write reports faster, but also improves the quality of those reports.
While Draft One provides several benefits to law enforcement, we also believe it will provide benefits for prosecutors who work with police agencies that leverage Draft One. Feedback from early adopters indicates that prosecutor offices will receive reports that are more consistent, more detailed, easier to read, concise and clear.
In this article, we will explore the most common questions we have received from prosecutors about Draft One. If you would like to learn more, we highly recommend watching our prosecutor webinar about Draft One - AI in Criminal Justice: What Prosecutors Need to Know.
FAQs about Draft One
How are report narratives used within the justice process?
Report narratives are written following calls for service and incidents in the community and are used to summarize these incidents as part of the overall police report. Should an incident lead to a court case, report narratives can be used as an aid to refresh the officer’s memory and contribute to the overview of the case, along with evidence, such as body-worn camera footage.
Does Draft One automatically write and submit report narratives?
No, Axon develops AI technology in a way that ensures human officers are responsible for decisions made and actions taken. Draft One generates report narratives based on body-worn camera audio, but we have designed the software so that these narratives cannot be submitted without officer review and approval. Officers must review the draft, manually fill in missing information and then sign off on the narrative’s accuracy before submission. This workflow saves officers significant time while still ensuring that each report is reviewed for accuracy and key contextual information. Watch this demo video to gain a better understanding of how Draft One works:
Does Draft One add any creative embellishments to the reports?
No, report narratives are drafted from the audio transcript from the body-worn camera recording. The underlying model we used for Draft One transcription is Open AI’s GPT-4 Turbo and we calibrated the model to prevent speculation or embellishments that can be more common and necessary for other AI use cases. Draft One reports stick to the facts, and require officers to review and add any missing pieces of information.
How will Draft One impact report quality?
It’s critical that Draft One helps provide initial report narratives that are high quality, concise and clear. The AI-generated report uses good grammar and spelling, and since it is drafted from the body-worn video, it does not forget details. While AI technology can be helpful in taking on complex and time-consuming tasks, we recognize it isn’t infallible, so we developed Draft One with a range of critical safeguards, and require every report to be reviewed to preserve the crucial role of human decision-making. We ran a study to compare the quality between officer-only reports and reports created by an officer using Draft One. The study was conducted by 24 independent experts, including prosecutors, field operations command staff and inclusion scholars. Results showed that Draft One reports are equal to officer-only reports in terms of comprehensiveness, neutrality and objectivity, and Draft One reports were rated more highly when it came to terminology and coherence. In short, Draft One reports can help agencies improve the quality and consistency of their reports. Review the full study here.
What are some of the benefits that you’ve learned from trial customers?
Draft One reduces the amount of time police need to spend writing reports, which allows them to spend more time in the community or responding to calls for service. Because the reports are drafted from the body-worn camera audio, many officers have shared they speak in a more thorough, interactive and comprehensive way that benefits the report and the community members involved. This includes talking into the body-worn camera on the way to the call-for-service, echoing responses to the victim/witness/suspect during the call, and then debriefing further into their body-worn camera after the interactions. This is great, not only for evidentiary purposes, but also because community members have more information and because improved report quality leads to a more procedural justice process.
Has this been vetted by prosecuting and defense attorneys?
In developing Draft One, we spoke with a number of district and defense attorneys who shared a few perspectives around the benefits they see with Draft One while also agreeing about the essential role officers play in signing off on the report’s accuracy:
Reports are an accelerant to decision making and an aid to refresh the officer’s memory when they’re in court two years after the incident. Reports are not generally submitted as direct evidence, but are often used as a method to refresh the recollection of a testifying police officer.
Agencies and attorneys tell us reports written today are of low quality — poor spelling and grammar, insufficient detail, elements copied over from old reports — leading to inefficient justice for victims. That’s why we conducted a double-blind study with independent experts, including prosecutors among others to measure quality. Results showed that Draft One performed equal to or better than officer-only report narrative drafts across five dimensions, including completeness, neutrality, objectivity, terminology and coherence.
Draft One includes safeguards that require every report to be reviewed and approved by a human officer, ensuring accuracy and accountability of the information. There are multiple ‘humans-in-the-loop’ to ensure the information in the report is accurate and the right criminal justice outcome happens — the original officer, their supervisor, the detective working the case, a detective supervisor, the prosecutor, and the defense attorney.
Are other prosecutors and defense attorneys accepting Draft One reports?
While it’s true that every jurisdiction needs to define their own policies when it comes to the use of any public safety technology, there is some related legal precedent here. Today, some officers call into phone lines to complete reports, then sign off on the accuracy of those reports before submitting. Draft One is simply the next step in the evolution of accelerating report writing to give officers more time back to serve their communities. We’ve met with a number of attorneys and their feedback has helped guide our safeguards and ensure the final report is both accurate and high quality. We recommend you check out our Draft One webinar for prosecutors to learn how Draft One can impact the criminal justice process, and of course we recommend you communicate with your local attorneys to discuss Draft One. We have found that transparency with attorneys and other key stakeholders is key to successful deployments of Draft One, and even provide default footer message templates to make this easy.
How will Axon keep tabs on cases to know that issues didn’t occur later down the line as part of the justice process?
There is an initial feedback tool when the report is submitted so officers can give us feedback on the draft. Longer term, we’ll continue to work with agencies using Draft One to get feedback directly and conduct additional studies to measure the impact of Draft One on report quality and the rest of the criminal justice process. We recognize the immense promise of ethical AI innovation and are committed to transparency around the development of these tools as we innovate responsibly.
How does Axon ensure AI solutions are developed and used in a responsible, ethical way?
Axon follows our Responsible Innovation framework whenever we develop new technology. We recognize the immense promise of ethical AI innovation. Our mission is to harness cutting-edge AI technology to revolutionize community safety, all the while prioritizing the rigorous mitigation of biases and other potential risks. We build our AI-enabled technology with controls so that it never removes human decision-making in critical moments. Axon also rigorously tests our AI-enabled products to reduce inherent bias in the models before products are released. We are continually monitoring and improving any unforeseen instances of bias that may arise in the models and continue to seek external feedback from our Ethics and Equity Advisory Council and others when implementing AI.
Will officers be transparent about how the report was drafted?
We have encouraged all of our law enforcement customers to be transparent from the start with their attorney’s offices. Based off some feedback we’ve received from law enforcement officers and attorneys, we are implementing a customizable disclaimer that can automatically appear at the end of each report draft. This will help ensure officers are always up front about how a draft was created, and reduce the risk of prosecutors being surprised in court.
How does it handle multiple speakers and voices?
We leverage speaker labeling, which means our transcribing technology is capable of identifying different speakers and labeling them. When officers use names more consistently, it makes it easier for the AI technology to apply specific names to unique speakers. When names are not used, generic labels like “speaker 1” or “complainant 2” will be used.
Do you store the original draft and highlight the officer's edits?
No. Historically, edits — whether it be the officer’s or the officer’s back-and-forth with a supervisor — are stored through the editing process until the report is finalized. With Draft One, the officer has to review and edit the report before verifying that every word is accurate in their final report, which is then the artifact made available to the public. Just like with word processing documents, edits, additions, deletions, etc. are not generally saved; only the final approved narrative is saved and added to the incident report. The final draft of the report is not saved by Axon Evidence, but rather the officer saves it inside the agency's record management system (Axon Records or other third party RMS).
Does Axon plan on creating AI tools for prosecutors?
We believe there are many opportunities for developing AI tools for Prosecutors. We believe that AI should never remove the human element, but make it much easier for humans to focus on the evidence that matters so they can effectively make decisions. We also always start with the problem and look for a variety of solutions, and then we turn to AI if it is the only real way to solve a problem (learn more about how we develop AI solutions in a responsible way).
We are in the early stages of exploring AI tools for prosecutors. Some of the problems we’re looking to solve include tools that can help review evidence and prepare for cross-examination. We also know prosecutors are “drowning in evidence,” and AI is really good at analyzing, summarizing, and highlighting clips that may require focus. We welcome your feedback if you have ideas - please contact us!
Next steps
You can learn more about Draft One by visiting our product page or watching our Draft One Webinar for prosecutors. We also encourage you to reach out to your local police departments and check to see if they’re considering Draft One. We believe that transparency about this solution is essential to it’s success, and love it when different agencies discuss the possibilities. Lastly, please contact us if you have questions, concerns or ideas about the future of AI in criminal justice.