The Lens Agent

The Lens Agent Trial

Keeping up with changes across large portfolios of properties can require intensive manual reviews of imagery or time-consuming site visits. The Lens Agent helps teams dramatically scale their monitoring capacity by using AI-supported workflows to review imagery and generate draft observations.

The trial allows your organization to run the Agent 15 times with preset prompts at the individual property level. If you're interested in automated portfolio-scale monitoring, custom prompts, or managed monitoring, reach out to us at [email protected] to discuss further.

What is the Lens Agent?

The Lens Agent is an AI-powered land monitoring assistant built to streamline your team's day-to-day property oversight. It helps you interpret imagery and identify baseline conditions, changes, structures, and other visible features on properties you monitor.

You choose the imagery for the Agent to review and which property to monitor, then the Agent gets to work inspecting the selected scene(s) and drafting observations for you to review and save as notes.

The Agent is designed to support your team's monitoring work, not replace human judgment. Your team remains in control of when the Agent runs, what it reviews, which outputs are saved, and how they are attributed.

What's included in the trial?

The trial is designed to give your team a first experience with the Lens Agent through guided, property-level workflows. During the trial, your organization gets:

• 15 free runs at the property-level
• Access to preset prompts depending on the scene(s) you're viewing:
  • Compare mode prompts
    • Describe what's changed between two scenes – the Agent reviews two images and drafts notes about changes it identifies, ranging from vegetation disturbances to signs of trespassing, and more
    • Identify structure changes between two scenes – the Agent reviews two images and drafts notes focused on any changes to buildings, sheds, and other structures
  • Single-scene prompts
    • Describe what's present in this scene – the Agent reviews one image and drafts notes about visible features or conditions on the property
    • Identify structures in this scene – the Agent reviews one image and drafts notes focused on buildings, sheds, and other visible structures
• Draft notes by the Agent that can be reviewed, edited, saved, or dismissed
• Full visibility into the history of Agent runs on each property

How to start your free trial

All organizations get 15 free runs with the Agent to experience it first-hand. Once the trial has been enabled, any Member, Admin, or Owner user in the account can try out the Agent at the individual property level.

1. Enable the Agent: To begin, the account owner must opt-in. This can be done directly from the Settings page or by opting in on the Agent pane from any property page in Lens. Owners can also disable the Lens Agent from settings at any time.

2. Team access & free trial runs: Once enabled by the account owner, any user in a Member, Admin, or Owner role can try the Agent. To help you get a feel for the tool, your organization will receive 15 free runs.
3. Kick off your first Agent run: First, open one of your properties in Lens and click the "Agent" pane. Then, select the image (or two images in Compare mode) that you want the Agent to interpret. Finally, choose your prompt and the Agent will begin working. Runs typically take about 5–10 minutes to process.

4. Review results: After the run is complete, open the "History" tab and click your prompt to review the Agent-generated notes. You can edit the text if you wish before saving the note directly to your property, choose whether your name or the Agent is shown as the note author, or simply dismiss the note if it isn't relevant to your monitoring.

5. Share your feedback: We would love to hear how the Agent works for you! Please let us know if the observations were helpful and any other comments you'd like to share at the bottom of each run.

Want to use the Agent at scale?

The trial is just a starting point. If your organization wants to monitor many properties, run custom prompts, or configure the Agent around a specific use case, we'd love to talk with you. Expanded workflows could include:

• Custom prompts optimized for your monitoring needs
• Runs across a portfolio or subset of properties
• Automated runs across many properties at a certain cadence, such as quarterly
• Managed monitoring support from the Upstream team
• Integration with reporting or review workflows

The larger goal is to help your organization monitor land more efficiently, whether your team uses Lens Agent directly in Lens or works with Upstream to configure and support the workflow on your behalf. Please reach out to us at [email protected] to discuss how we can leverage this technology to help your team scale up your impact.

FAQs

What happens to my data if I use the Agent?

When you run the Agent, the imagery or image tiles needed for that run, the selected prompt, and relevant property area information are securely sent to the AI model or models used to generate the result. Only the information needed to complete the requested Agent run is shared.

What models are used exactly?

The Lens Agent uses different language and image segmentation models depending on the task. For example, describe prompts use large language models (LLMs) to interpret imagery and generate plain-language observations, while structure identification prompts use segmentation models to spot and outline features. Additional models and versions may be introduced over time.

How accurate are the results?

The Lens Agent provides AI-assisted monitoring observations. Outputs may contain errors, omissions, or uncertainty. The Agent is designed to help your team review imagery more efficiently, not to replace human expertise or final review. You should always validate Agent outputs against the source imagery and your own knowledge of the property before taking action or using results in reporting.

Is my data used to train AI models?

No, customer imagery, and outputs are not used to train third-party foundation models. Upstream may retain limited metadata or outputs internally to support debugging, quality evaluation, customer support, and improvements to the Lens Agent experience.

Is the Agent automatically on or always running?

No, the Lens Agent is optional and must be enabled by an account owner in the Settings page. AI models are only invoked when a user explicitly runs the Agent on selected imagery. During the trial, the Agent does not automatically capture any information about your workflows in Lens.

Are there any limitations during the trial?

During the trial, your organization has access to 15 Agent runs by default. Lens Agent is available for supported high-resolution imagery sources, but cannot be used with Nearmap imagery at this time. Each trial run can return up to 100 observations and if a run reaches this limit, the observations will be consolidated into one multi-polygon note for easier review.

For any other questions please reach out to us at [email protected] or refer to our Terms of Service. All use of the Agent is subject to our Terms of Service and End User License Agreement.