Axon Enterprise and Machine Learning

Axon, known for its tasers and body cameras, is now using enhanced video, image, and audio processing to make communities safer

Introduction: Axon Enterprise (formerly known as TASER International)

Axon Enterprise’s (“Axon”) mission is to protect life through innovative technologies that make communities safer. The company sells stun guns (tasers), body cameras, in-car cameras, and sensors to law enforcement. They aim to have every public safety office carry a taser, deploy an Axon camera, and be connected to the company’s proprietary network. The company generated $344 million in revenue in 2017 and is headquartered in Scottsdale, Arizona.1The company has been innovating in the law enforcement industry since inception. In 1991, founders Rick and Tom Smith developed a non-lethal electric weapon that used compressed nitrogen instead of gunpowder. Later, the company introduced other hardware products such as body cameras, and has more recently moved into software with a cloud-based digital evidence system.

Why is Machine Learning important to Axon?

Machine learning and artificial intelligence have the potential to dramatically impact the world of law enforcement. Each year, it is estimated that over $100 billion is spent on law enforcement in the United States.2 On the job, law enforcement officers collect countless hours of footage (data) through their body cameras. While controversial, the cameras offer several advantages including increased officer and public safety, accountability, continuous learning, and the potential for improved processes. When a law enforcement officer wears an Axon body camera, on average complaints decrease by 93%, use of force decreases by 70%, and guilty pleas increase 20%.3 As of the company’s most recent Analyst Day, Axon has more than 15 petabytes of customer data on their network, increasing at 1.5 petabytes every month.3

The use of artificial intelligence could help officers reduce the time they spend watching hours of inconsequential video, which will enable them to invest more time and resources into actual field work. Machine learning algorithms can be programmed to differentiate between people and objects, recognize a variety of different events, and identify perpetrators caught on tape through facial recognition software. In addition, as discussed in a Harvard Business Review article, machine learning can aid human decision making by discovering patterns in the data that would otherwise not be recognized.4 Some examples of this include: are officers more likely to use a lethal weapon earlier or later in their shift, typical perpetrator reactions in car chases, and many more.

Globally, law enforcement will benefit from machine learning, which in turn should create shareholder value for Axon given the size of the market and demand.

What is Axon doing to address the issue in the short- and long-term?

Over the past few years, Axon has taken several steps that will enable the company to provide this technology to its customers. In early 2017, the company acquired two companies that specialize in machine learning, Dextro and Misfit. Post-acquisitions, Axon merged the two companies and created an artificial intelligence group called Axon AI. The goal of the combined group, which consists of 20+ engineers and researchers, is to introduce artificial intelligence-powered capabilities for the public good (law enforcement, public safety, etc.).

Dextro offers the first computer-vision and deep learning system to make the visual contents in video searchable in real time. Law enforcement will have the ability to isolate and analyze the most important parts of footage from large amounts of video data. Misfit is focused on improving the accuracy, efficiency, and speed of processing images and video.

Initially, the company is focusing on eliminating paperwork from the field and triple the amount of time officers can spend serving their communities. The areas of investment include automated redaction, automated transcription, and automated reporting.5

My recommendation

While Axon’s management team has recognized the importance of adopting and utilizing machine learning in its solutions for law enforcement, there is still work to be done. I believe there are three things the company should focus on going forward to improve its value proposition:

  1. Privacy: body cameras have raised several privacy concerns both for the public and officers. Axon needs to work with regulators, law enforcement agencies, and the public to create clear guidelines for what will be recorded, how the algorithms work, and so forth.
  2. Adoption: ~80% of police officers are in the field without a body camera.6
  3. Continuing to innovate and come up with new applications: there are several other potential applications in predictive policing. Perry et al. (2013) distinguish three main objectives of predictive analysis in criminological applications: (1) predicting perpetrators, (2) predicting victims, and (3) predicting when and where there is a higher risk of new crime events.7

Open question

Given how nascent the technology is and the limited amount of information on the benefits and costs, how should Axon ensure communities employ good policy frameworks that benefit both the public and law enforcement?

Word count: 786


  1. Axon Enterprise, Inc. Form 10-K. Securities & Exchange Commission (For the fiscal year ended December 31, 2017).
  2. Justice Policy Institute. United States Continuing to Overspend on Police, Despite Decreasing Crime Rates. Justice Policy Institute. (May 22, 2012).
  3. Axon Enterprise, Inc. Investor & Analyst Day Presentation. Axon Enterprise, Inc. (November 16, 2017).–Analyst-Day-Presentation/default.aspx
  4. Wilson, S. Sachdev, and A. Alter. How companies are using machine learning to get faster and more efficient. Harvard Business Review Digital Articles (May 3, 2016).
  5. Rick Smith. Axon’s AI Work What’s Ahead. Axon Enterprise, Inc. (May 5, 2017).
  6. David Gershgorn. America’s largest body-camera supplier is giving police free AI to analyze crime footage. Quartz. (April 5, 2017).
  7. Perry, W. L., McInnis, B., Price, C. C., Smith, S. C., & Hollywood, J. S. Predictive policing: The role of crime forecasting in law enforcement operations. RAND Corporation. (2013).


Alibaba: Application of Machine Learning in Logistics


BMW & Additive Manufacturing: Using New Technologies to Drive Down Production Costs & Timelines

Leave a comment