Virtual Operator (VO)

SureView’s Virtual Operator uses the latest AI models to automate many of the functions of today's modern Security Operations Center (SOC). The Virtual Operator helps human operators respond faster, identify threats, and deliver better security outcomes.

The goal of the SureView Virtual Operator is to: 

• Take over some of the routine response functions of the operator
• Pickup and determine the nature of an event quicker
• Deliver cost savings for command centers as they grow and scale

Teams are in full control of how they deploy and use Virtual Operators, they can start and stop Virtual Operators at will, and with customized guidelines, they can control their behavior.

Virtual Assist

Executive Summary for every event

The quickest and easiest way to leverage the power of AI in SureView is for the virtual operator to summarize the outcome of events. This immediately automates time-consuming step of every event, writing the short narrative of everything that happened. We call this Virtual Assist. The Virtual Operator provides the full audit trail of the event to AI asking it to summarize the outcome. AI is particularly good at this task, it provides a concise, consistent summary of each event in seconds. If the operator doesn't like the output, they can edit and make it better, they remain in full control. This immediately improves reporting and notifications providing management with clear summaries of every event.

Virtual Operator

Assisting operators with every aspect of event response

The full Virtual Operator uses Generative AI to handle every aspect of responding to an event. As events are received into SureView a Virtual Operator can be configured to: 

• Pick up the event from the queue and immediately start analyzing the alarm details, the location, and the time of the event
• Analyze incoming alarm footage, as well as livestreams detecting people and vehicles while also describing what it saw
• Follow and complete Action Plans - answering questions, adding notes, etc
• Following guidelines to perform specific actions, such as increasing the event priority based on the presence of people in the alarm clip.

The Virtual Operator will only act based on the Guidelines set on the AI Settings page, meaning the AI should never perform a task unless explicitly configured to do so. Whenever the Virtual Operator completes a task, the following information is included in the prompt passed to the Language Model:

From this template, note that content for `Area Summary Text` comes from the value of the Area Summary, configured on the Area Setup page, which can be set here for Suite users (Or here [US/EU], for traditional, Response-only users).

Additionally, the content for `General Guidance Text` comes from the AI Setup page in the Suite, here.

And finally, key frames from each Media source are passed to the model in a format it can understand, such that it can not only understand the text-based content of the Audit Trail, but also what has happened in the Media Matrix. (only key frames from Clips are included by default)

Prior to sending the request to the Large Language Model, all key frames have been sent to an Object Detection model to identify if a Person or Vehicle was observed within them. This not only provides additional context for the VO, but it can directly be leveraged in the Guidelines to enforce the Virtual Operator to respond in a certain way if a person and/or vehicle are detected. This takes advantage of the high reliability of Object Detection models, while also making use of the advanced visual reasoning capabilities of modern Vision Language Models.