This guide walks you through each step of creating and deploying an AI Agent. Not all agents share the same steps, but many share similar setup options, all of which are outlined in this guide.
For a high-level overview of the agents and agent categories Solink offers, see the Agents: Getting Started guide here.
Note: While most steps in this guide are universal to all agents, some agents require specific configurations for schedules and triggers. The difference in available options between agents are highlighted throughout this guide.
Agent Creation Steps
This section provides an overview of each of the agent creation steps, regardless of selected agent type.
To start creating an agent:
Navigate to the Agents Overview page in the Solink platform > Create New Agent. Take note of the requirements needed for agent creation access.
Next, select from one of Solink's pre-built agents based on what you want to accomplish; once selected, you can further customize how the agent behaves on a more granular level.
See the sections below for a high-level overview of each agent configuration step.
Agent Name (Optional)
You can rename the agent in the Agent Description section. You can also leave the default name as is.
Agent Skills
Skills define exactly what activity the agent should flag. Each agent has its own unique set of activity it monitors for.
All skills are selected by default, but if you do not want an agent to monitor for a specific activity you can deselect any of the selected skills.
Note: Deselecting skills can sometimes help reduce the AI quota the agent uses.
You can also provide Additional Context in the Skills section to provide further guidance for the agent:
Add information to the Additional Context field to help the agent understand certain terms or situations relevant and/or unique to your business, such as "Look for voids for small purchases - this is important because theft may occur more frequently with these kinds of transactions".
Tip: See our agents prompting guide for best practices on creating effective prompts.
Agent Schedule
Depending on the agent, you can configure how frequently the agent runs based on a schedule.
Scheduling: Video Monitoring Agents
Schedules can be configured such that the agent monitors cameras on an interval (for instance, every 5 minutes). This is recommended for use cases requiring spot checks.
Scheduling: Transaction & Video Monitoring Agents
Some agents, such as the Loss Prevention Agent, monitor both video transactions and transaction data.
Schedules for these agents are system defined based on when transactions become available in Solink. By default, the schedule is set to run daily at 5 AM and fetch transactions from up to 3 days in the past; you can review these events once they are fetched via the Agent Logs section (see Agent Logs) or sent via a notification (see Choose Agent Action).
Monitored Areas
Select specific locations you want this agent to monitor.
If you do not want the agent to monitor all cameras at a given location, you can expand the location in the location selection window and select/remove cameras:
Select Save Selection once you have selected the locations and/or cameras you want the agent to monitor.
Confidence Threshold
The Confidence Threshold allows you to fine-tune the level of activity the agent monitors for. You may need to re-visit this setting after the agent has been deployed if you find the agent is capturing too much or not enough activity of interest.
The lower the threshold, the more activity the agent will flag. For instance, if the agent monitors for suspicious activity, the agent may flag whenever someone wearing a hoodie sweater appears in the video.
The higher the threshold, the less activity the agent will flag. For instance, if the agent monitors for suspicious activity during closed store hours, the agent may only flag whenever there is a clear threat to person or property, such as a forced entry, shattered window, etc.
It is recommended to start at the Medium Confidence Threshold option and re-adjust the agent later if needed.
Event Structure
Select the + Add Output button to add an event output. Each output determines what the event card displays when the agent flags activity you specify in the output.
You can define the following for an event output:
The exact criteria needed for the agent to create an event. Options include:
Every time the agent runs: Creates an event every time the agent runs, as per the schedule (for example, every four hours between 9-5). Some agents such as the Overnight Guard run constantly and therefore this option may not be available.
Only when a skill is matched: Creates an event only if the agent finds activity based on one of its skills (for example, for the "Back Door Security Check" agent, it creates an event when it detects an unsecured/opened back door).
Only when specific conditions are set: Set the conditions yourself to ensure the agent only triggers an event under those conditions. Type out a description of what you want the agent to flag.
The output provided in the event card. Output setup includes the following:
Output: The name of the event displayed in the event card.
Type: Determines how the agent responds based on the following options:
Yes/No: Provides a "Yes" or "No" response based on the description.
Duration: The duration of the agent event (EG: How long did it take for an employee to greet a customer).
Single Select: Agent chooses an answer based on provided answers (EG: What type of car came through the drive-thru? Minivan, SUV, Sedan or Other).
Numerical: Total count of items or occurrences of monitored activity (EG: How many customers were in the queue OR how many times did the customer interact with the display).
Percentage: An estimate of the percentage of monitored activity (EG: How full is the donut display?).
Text: A written response describing in detail a point of focus (EG: Describe the customer's interaction with the display).
Description: Write the associated question/description that pairs with the Type field, as outlined in the bullet points above.
You can create reports from your agent events and then build out a dashboard of agent insights/analytics. See Agent Events for more information.
Agent Action (Notifications)
Select one or more actions to determine how the agent notifies you and/or other Solink users when the agent identifies activity of interest.
You can select no actions if you want to simply restrict agent responses to events and the Agent Logs, or you can select one or more actions if you wish to enable other notification methods, such as SMS or Email.
Publish Agent
Once you have finished customizing your agent, select Publish to create the agent:
You can view and edit the agent from the main Agents page once published. More information about viewing and editing your existing agents is outlined below.
View Created Agents
You can view and edit agents you have created at the bottom of the main Agents page:
For a full view of all agents, select My agents:
The agents list shows:
Agent Name: Name of the agent.
Status: Displays whether the agent is currently running or not (enabled/disabled). See Disable/Re-Enable Agents for more information.
Last Activity: How long ago the agent last ran.
Runtime: The amount of hours the agent has used.
You can also sort by:
Recently created agents
Agent name
Edit Agents
You can perform a few actions to edit agents. Select the My agents button to start editing your agents.
Disable/Re-Enable Agents
You can disable an agent to stop the agent from running without completely deleting it.
To disable an agent:
From the My Agents page, select the More Actions icon (three dots) next to the agent you want to disable.
Select Disable to immediately disable the agent.
You can also re-enable a disabled agent. Repeat the steps above and instead select Enable.
Delete Agents
To delete an agent:
From the My Agents page, select the More Actions icon (three dots) next to the agent you wish to delete.
Select Delete.
Confirm deletion in the next window. Please note that once an agent is deleted, it cannot be recovered.
Duplicate Agents
You can create a duplicate of an exiting agent. To duplicate:
From the My Agents page, select More Actions icon (three dots) next to the agent you wish to delete.
Select Duplicate.