Custom AI processing for meetings

When you click to stop transcription, Tana can run a post processing command. The post processing command is set up under the audio-enabled toggle in the supertag configuration.

The default command is the Text processing agent, which is adapted to work well for meeting transcripts. You can also set up your own custom post processing using command nodes.

By default the Text processing agent gives you a summary based on the full meeting transcript. You can further customize its behavior, below are some of the customization you can do.

Extract action items

To automatically extract action items from your meeting.

  1. Edit the Text processing agent in the meeting tag configuration.
  2. Add the Action items target parameter, and reference the field where the items should be placed.
  3. Add the Tag to use for action items parameter and reference the supertag to use for extraction.
  4. Optional: Add the Action items prompt parameter to specify what should be extracted in more detail.

Action items will be extracted from the transcript and created as individual nodes with a tag suggestion. You can simply click the tag suggestion to apply the tag.

If you want extracted action items to be tagged directly, not just suggested, then you can use the Add tags directly parameter.

Custom summary notes

You can add a notes field to your meeting to include your own meeting notes as context to the summary that the Text processing agent creates. This allows you to highlighting key points during the meeting or add additional context to be included in the summary.

  • Add a field to your audio-enabled supertag (ex. Notes).
  • Add the Meeting notes field parameter to the Text processing agent and reference the field.

Any notes written in the field will be included as context to the summary when the live transcription is stopped.

Other item extraction

In addition to action items, you can extract other kinds of information from your meeting transcripts.

Item extraction is better suited for extracting things that are not directly actionable like tasks, follow-ups, etc. This functions well as a more general extraction, for example user observations, learnings, insights, etc. There are no restrictions to what you can extract.

  1. Edit the Text processing agent in the meeting tag configuration.
  2. Add the Extracted items target parameter, and reference the field where the items should be placed.
  3. Add the Tags to use for item extraction parameter and reference the supertag to use for extraction.
  4. Optional: Add the Extracted items user prompt parameter to specify what should be extracted in more detail.

You can also extract entities with using the same method and the parameters: New entities target, Tags to use for entities, and optionally Entities prompt. Entities can be things like companies, people, cities, countries, etc.

Custom prompts

You can add a custom prompt to the Text processing agent to tailor the generated summary or extracted items. All functions have system prompts that are tailor made for the task, but you can override these by using the different prompt parameters if you want to customize the output.

  1. In the supertag configuration, edit the Text processing agent.
  2. Add the prompt parameter that you want to override: Summary prompt, Action items prompt, Extracted items user prompt, Entities prompt, Augment prompt.
  3. Write your custom instructions.