Create AI trackers

Who can use this: Any team member

Available on: Gong Foundation

Required permissions: Manage AI Trackers

Ideal for: Business admin, RevOps, Enablement manager

AI trackers use Gong's AI to detect whether a specific concept or intent appeared in a conversation, regardless of the exact words used. Create one by using our question builder, or by writing natural-language questions that describe what you want to detect. Publish it to start analyzing conversations.

Use a keyword tracker when exact words or phrases matter. See Create a keyword tracker. Use an AI tracker when the topic can be said many ways.

AI trackers consume Gong credits based on the calls and emails they analyze. Before publishing, you can preview results and see the estimated credit cost for the start date you choose.

Step 1: Open AI Tracker settings

AI Tracker settings is where you create, view, and manage all your company's trackers. Open it to access the trackers table and create new trackers.

To open AI Tracker settings:

1. From the left sidebar, click Admin center.

2. In Agent Studio, hover over AI Tracker and click Settings.

3. Click New tracker and select AI tracker.

Step 2: Choose how to build your question

Two question-building methods are available. Choose before writing. You can switch between them later, but selecting the right one first saves reconfiguration.

• Use question builder (default, recommended): Guides you through predefined phrase options. Gong pre-fills the supporting context based on your choices, which you review and refine. Use this when you want structured help phrasing the detection question.

• Write a free-text question: Write your detection question directly in natural language, with full control over the wording and context. Use this when the concept is highly specific or doesn't map to the predefined options in the builder.

Note:

Switching from question builder to free-text carries over any pre-filled details. Switching back restores the original builder selections.

Tip:

If your company uses the example-based builder, that option is available from the creation page. See Create trackers using the example-based builder for guidance.

Step 3: Configure your detection question

The question builder and free-text paths work differently.

If you chose Use question builder:

1. Construct a question by selecting from predefined phrases for three sections. In each section of the builder, select the phrase that best describes the concept you want to detect.

2. Review the supporting context Gong pre-fills in the Add details field. Replace any placeholder values, such as competitor names, product names, or specific terminology, with your actual terms.

3. Provide extra information here to improve tracking accuracy.

4. Refine your question with in-app AI suggestions.

If you chose Write a free-text question:

1. Write a natural-language question from the perspective of someone reviewing the conversation. For example: "Did the rep ask about the customer's current vendor?" or "Was pricing discussed during this call?" A good concept is specific, not broad. For example, "pricing" is too general. Your concept does not have to be a question, though a question is recommended.

2. (Optional) Add supporting context in the Add details field to help the AI distinguish relevant matches from false positives.

3. Refine your question with in-app AI suggestions.

4. Choose which side of the conversation you want to track: Our rep (your side); Customer; Any party. Limiting to one side avoids false positives and reduces the volume analyzed.

Note:

Gong checks your question against existing trackers, both custom and by Gong. If a similar one exists, a warning appears. You can open the matching tracker to review it, dismiss the warning, or continue creating your own.

Step 4: Configure scope and filters

Narrowing scope reduces credit consumption without affecting the quality of results on relevant conversations. The more focused the tracker, the less it costs to run.

To configure scope and filters:

1. Under Choose where to track it, select Calls or Calls and emails. Choose calls only if the signal you're tracking is unlikely to appear in email.

2. (Optional) Click Edit filters to narrow the tracker to specific call types, deal stages, participants, or other criteria. Filters are applied to both the preview and the live tracker.

3. (Optional) Choose part of the call to analyze: Anytime (the entire call), or limit to the first or last portion by minutes or percentage. Analyzing only a specific part of a call consumes fewer credits.

Step 5: Preview results

Before publishing, preview your tracker against a sample of recent calls to check whether it detects what you intended and to review any false positives or misses.

To preview tracker results:

1. Click Preview results. Gong samples 100 calls and 400 email threads and shows the results in the right pane.  Each result includes relevant details, such as the account, a summary of the tracked concept, or if relevant— the email subject and sender.

   1. Select an email result to view the full thread context.

   2. For calls, Click Based on X mentions to view the relevant conversation snippets.

2. Review the matched calls and emails. If results include false positives or unexpected matches, refine your questions, filters or add context and preview again.

3. (Optional) Scroll to the bottom of the list to search beyond the initial 100 calls/ 400 emails.

Tip:

Refining scope or filters in Step 4 also narrows the sample used for preview and reduces the estimated credit cost shown at publish.

Note:

You can save a tracker as a draft without publishing it. Click Save, enter a name, and save the tracker. It appears on the AI Tracker settings page with a Draft label. When you’re ready to publish it, click Edit.

Step 6: Name and publish

Publishing activates the tracker and starts analyzing new conversations going forward. A start date in the past also applies the tracker to historical calls from that date.

To name and publish your tracker:

1. Enter a name for your tracker. The name must be unique within your company and will appear wherever tracker results are shown across Gong.

2. Click Publish.

3. (Optional) Set a Start date and End date to limit the tracker's analysis to a specific period.

   1. Trackers with a future start date are saved as Scheduled and do not begin analyzing until that date.

   2. A start date in the past applies the tracker to historical calls from that date.

   3. Setting an end date stops analysis automatically when that date is reached, and its status changes to Paused.

4. Review the estimated credit cost for the dates you selected. The system validates that your company has sufficient credits before proceeding.

5. Confirm the publish. The tracker moves to Published status, and starts analyzing conversations.

6. (Optional) Select Create and subscribe to a stream based on this tracker to collect matching conversations and get notified when new ones are added.

Note:

Publishing is blocked if your available Gong credits don't cover the estimated cost. If publishing is blocked, choose a later start date to reduce the backfill, or contact your admin.

Historical calls from your start date onward are queued for processing after publish. New conversations are analyzed as they are recorded.

Note:

When you update a published tracker, its past results are kept. The update applies to new conversations going forward.

Where tracker results appear

Tracker results show up in these places across Gong. For more, see Analyze tracker results.

• Search

• Team stats (Team insights)

• Streams

• Initiative boards

• Playbooks

• The deal activity panel

• The Calls API and CSV export

Frequently asked questions