Evercam Copilot (Beta)

This is on limited availability. To know more, please contact your Success Manager.

Evercam Copilot is an AI-powered construction intelligence chatbot tool integrated into the Evercam platform. It uses machine learning, natural language processing (NLP), object detection, and image analysis to help users track project progress, enhance safety, and drive productivity insights based on real-time construction site footage.

Copilot generates AI-powered summaries, visual reports, and observations, providing actionable insights directly from camera feed and project data. It is currently in beta and will continue to evolve based on user feedback and ongoing development.

Vimeo

While Evercam Copilot shows great potential, it is important to acknowledge that many features are still in early development. We're actively working to improve its accuracy and usefulness.

What’s new with Copilot (v3.0)? 🤖

  • Dedicated Copilot Interface: You can now access Copilot through the dedicated user interface in the upper right-hand toolbar labeled “Copilot” (beside the Help button).

  • Custom Prompts for Images/Video: Copilot now supports custom queries about specific images or video clips, allowing direct analysis of site activity based on footage. This expands its capabilities beyond querying pre-collected data from Evercam's database.

  • Enhanced Feedback System: User feedback is integral to improving Copilot. You can now provide thumbs-up/down responses on answers, which helps refine future AI outputs

Current Capabilities

Copilot can assist in the following areas, though accuracy may vary as the tool is still evolving:

  • Safety hazards and compliance 🦺

  • Project progress tracking and milestone reporting 📝

  • Equipment and manpower productivity 🚚

  • Site conditions monitoring 📺

  • Compliance with regulations and site-specific requirements 📖

  • Assistance with Evercam features like integrating Procore, creating time-lapses, and more

Please note that as Copilot is still in beta, some results may not always be perfect. We're actively working to improve its accuracy and capabilities, and user feedback is critical in this process.

Expectations for Users

Accuracy of AI-Generated Responses

Copilot assists in analyzing data, but its insights should not replace human oversight. Responses are a starting point for decision-making and should be validated manually. For example, always cross-check a Copilot-generated safety report with on-site observations or official safety guidelines.

Ongoing Feedback and Improvement

User feedback directly impacts Copilot's performance. Each time you ask a question or rate a response, the system learns and improves. Use the thumbs-up or thumbs-down buttons at the bottom of each response to indicate accuracy or relevance. 👍👎

How to Use the Feedback Button

  1. Click thumbs up if the response is accurate or relevant, or thumbs down if incorrect or irrelevant.

  2. Be specific in your questions and provide feedback whenever possible to improve the system’s accuracy.

Glitches and Bugs:

As Copilot is in beta, occasional issues such as non-functional hyperlinks or inaccurate data may arise.

  1. Non-Functional Links: Sometimes, Copilot generates links that don’t work. If this happens, refresh the page or manually check the relevant footage on the timeline. Use the feedback button to report this.

  2. Incorrect Activity Identification: The AI may misidentify objects or activities (e.g., stating that cylindrical components were delivered on a certain date when they weren’t). Always verify this information manually and report incorrect answers using the feedback buttons.

  3. System Downtime: If responses are slow or don’t appear, refresh the page or wait a few minutes before trying again. Persistent issues should be reported to support@evercam.io.

  4. Validating Responses: Always verify Copilot’s outputs, such as checking project timelines or records, especially when links or details don’t appear accurate.