GPT-trainer uses a multiple-agent architecture for its chatbot design. To conduct lead collection under this framework, you will either need to create a dedicated Lead Collection Agent or integrate its capabilities inside another, existing one.

To create a simple Lead Collection Agent, head to the “AI Agents” tab. You can find the template for a Lead Collection Agent when you click on “Create new Agent”.

After you create the new Agent, go into its edit screen by clicking on it. Scroll down to review the base prompt. Depending on what information you would like to capture, configure the base prompt appropriately.

Click “Save changes”. Next, go to the “Variables / Tags” tab. For each piece of information you want to capture, make sure there is a corresponding Variable created to store it.

When defining the variable, be as explicit and concise as possible. Where you can, try to provide examples for what the intended values look like. You can also pre-define the variable’s default state. If left empty, its default value will be “NOT PROVIDED”.

You can also make it mandatory for the Variables to have non-empty (non-”NOT PROVIDED”) values before the AI Supervisor switches away from the Agent. You can do so using AI Supervisor Overrides. If you apply the appropriate settings there, the Lead Collection Agent will not switch away until all the Variables defined in this Agent have been assigned proper values.

Next, head over to the “Knowledge” tab. Depending on what information you’d like the Agent to have access to when requesting user information, you can link the associated training data with this Agent. In general, Lead Collection Agents are designed to specialize in collecting personal information from users, so they may not need any knowledge base access. To disable all knowledge, simply toggle off “All Sources” and leave “Agent Sources” empty.

Click “Save changes” and close the Agent edit dialog. Return to the Agent selection view. You can now “Connect” your new Lead Collection Agent. On the top right of the Agent card, open the “three dot” dropdown menu, then select “Connect”.

You should now see a glowing blue dot next to “Lead Collection”, indicating that the Agent is being trained. Since you may want your Agent to collect user information before performing any service, you can set this as the “Initial Agent” to handle the first incoming user query.

After the blue dot turns green, your Agent will be good to go. You may need to refresh the page to check for status updates.

Go to the “Preview” tab. You can now test your new chatbot, now equipped with conversational Lead Collection capabilities!

To review the data you collected, open the “Debug mode” while you are still in Preview.

You can also head to the “Inbox” tab, find the test conversation session, then open “Debug view”.

You can export the data you collected across all chat sessions by going to the “Inbox” tab, then clicking the “Export” button. In the subsequent dropdown, select “Collected Data”.