Meeting agent

Overview

The meeting agent is a life hack for people who spend lots of time in meetings, but don't want to spend time following up and post-processing the notes. Add the meeting agent to your calls, and it will transcribe what each participant is saying, and extract useful, actionable items that are immediately sent to the relevant context - without you lifting a finger.

Sync your meetings with Tana straight from your calendar. Our calendar integration brings all your meetings into Tana, and shows them on your daily notes page. For more on this, see Calendar integration. If you don't use Google Calendar you can add meetings manually by tagging them as meetings.

Allow Tana to take notes during your meetings. Send a meeting agent to any virtual meeting with a link, or have it automatically show up if your meetings are scheduled via your calendar. Works with Google Meet, Zoom, and Microsoft Teams.

See a summary of what was discussed, and all the action items mentioned during the meeting. Everything is linked to the transcript so you can see where it came from.

Meeting agent is an exclusive Tana Core feature. It uses Tana AI to transcribe and process meeting notes, which is only available on Tana Core.

Basics

The meeting agent is a set of special functions that can be added to a supertag.

• First time users: Setup of your first meeting agent. If you subscribe to Tana Core from an existing account, a setup process helps you select existing tags and fields for the meeting agent. For new users subscribing to Tana Core, the meeting agent will already be installed from the beginning, no setup required.
• It's easy to set up a meeting agent manually. On any supertag, set the base tag to Meeting and start the setup process in tag config. For more detail, see How to install another meeting agent
• Add an agent to a meeting. Click on the meeting (anything with the meeting tag is considered a meeting), make sure it has a meeting link and press "add meeting agent". It has now been invited and will ask to join the meeting once it's started.
• During the meeting, the agent is transcribing. Everything that participants are saying is being transcribed to Tana, live, during the call.
• Finish your meeting to start processing. Once you close the video call, the meeting notes processor begins. This uses Tana AI to distill important items from the meeting. A short meeting takes less than a minute, longer meetings can take several minutes.
• When processing is done, the meeting notes are complete. Notes include a Summary, found tasks, future agenda items, new and existing entities—all with links to relevant parts of the transcript.
• Enrichment of your knowledge. The meeting notes processor adds background information about persons and companies who attended and/or were talked about under their nodes. This helps with jogging your memory on what they've said about themselves before going into the next meeting you have with them!

Detail

The synergy between the Calendar integration and Meeting agent

The Meeting agent and Calendar integration are two intertwined features that are meant to work together.

The calendar integration is not just a convenient way to see your events on the day node. It is an opportunity for AI to help build out your knowledge graph:

• It syncs your meetings to Tana so you have a place to prepare for meetings, write notes and agenda points beforehand
• A meeting agent uses AI to attend and take notes during the meeting for you
• All notes are saved back in Tana on the meeting node, and connects back to your knowledge base.

When the Meeting agent creates notes, it does the following:

Writes a concise, interconnected meeting summary: Your meeting summaries are no longer siloed on a different AI transcription platform just for meeting notes. They now live in Tana with the rest of your knowledge, connected to things in the transcript and in your graph.

Creates action items based on the conversation: It picks up all the tasks, agenda items and decisions that come out of a meeting

Creates new nodes for entities if they don't exist yet: Things will get mentioned in meetings that you don't have an existing node for. AI detects cities, states and countries, and will add them to your graph as references wherever they are mentioned. It also detects products (like "ChatGPT") and companies (like "OpenAI").

Builds a CRM: Meetings usually have Attendees. By syncing your contacts, AI can enrich the information on these contacts with facts that it learns from meetings. Joanne is not just a generic contact in your CRM anymore, they went to an escape room last weekend with their husband and really enjoy a good mystery novel (good to know when getting a birthday gift).

Go here for more on the Calendar integration.

Setup

The meeting agent is a supertag that has meeting functionality under the hood.

To get the meeting agent, you can install it in one of two ways:

• New user: you don't have to do anything, it is already installed!
• Existing user before May 29: when you start your trial/upgrade to Tana Core, you can opt to install a new #meeting supertag, or to map an existing tag and fields to be used for your meeting agent. For more detail, see How to install another meeting agent

The meeting agent is meant to work seamlessly with Tana's calendar integration:

• Connect with your calendar to sync your meetings to Tana where you can use the meeting agent. For more, see Calendar integration
• You can still use the meeting agent without a calendar integration by adding the meeting tag manually.

Quick start guide

These are the steps for using a meeting agent. It looks like a lot but it's not actually that difficult, there are more steps here than you actually have to do:

1. Create/Find a meeting
   1. If you're syncing a meeting from your calendar, it should appear in your related content section on your daily page. Click on it to open it.
   2. If you're manually making a meeting, apply the #meeting supertag to any node to make a new one (or the name of the supertag you specified during setup)
2. Setup meeting. In the meeting node, make sure the Meeting link field is filled with the link to a video call.
3. Add meeting agent. Click the button "Add Meeting Agent".
   1. If there's no time in the Date time field, the agent enters immediately (usually within 30s after pressing button).
   2. If there is a time specified, the agent will be scheduled to automatically enter at that time.
4. Allow meeting agent in. When the meeting is open, the agent will ask to join the meeting; let it in. Once it's in, it immediately starts live-transcribing the meeting to Tana.
5. Processing meeting. After the meeting has ended, the meeting agent will automatically start processing the transcript. This usually takes 1 - 3 minutes. Do not close or reload your browser while the meeting agent is processing the transcript.
6. Done. When it's done, you can explore the results which include:
   1. A summary that has links to the full transcript.
   2. An overview of action items (if any) such as tasks, agenda items, and decisions.
   3. An overview of discovered entities (if any), including countries, cities, states, companies, and people.
   4. Background information about meeting participants (including their role, where they are from, what they are working on etc.) as well as on company entities.

How to install another meeting agent

You can have specialized meeting agents for the different types of meetings you have during the day. Here's how to set up another meeting agent that you can then customize:

• Create a new supertag. This will become your new meeting agent
• In the template, add the fields from your main meeting tag if you want to reuse them, or keep it empty if you want to have brand new fields created in the process.
• Set the supertag's base tag as Meeting.
• A new config option pops up at the bottom called Meeting agent. Open it and you'll see this:

• Press Set up agent and you'll see this. If the template is empty or no fields seem to be suitable, it will default to "Create new". If there are suitable fields, they will be pre-filled:

• After the setup, the meeting will have a default command set up that dictates how the meeting agent works. To customize that, see the next section!

Advanced: How to customize your meeting agent

Customizing your meeting agent requires you to have some knowledge of advanced supertag configurations and how to build commands from scratch. We think the best way to show you how it can be customized is to show you how it's built.

So, this tutorial is actually on how to build a meeting agent from the ground up, with explanations on all the parts along the way. Grab your favourite beverage and let's dive in!

1. Have the required fields in your meeting tag

A meeting agent tag requires the following fields:

• Meeting link: A URL field
• Transcript: A plain field

Optional but quite common in many meetings synced from the calendar:

• Attendees: Usually a Options from supertag field linked to a supertag that represents people in your graph, and that uses the Base tag: Person. In the calendar integration, this is automatically made as a #person tag. In the absence of calendar integration, your own tag for persons with Base tag: Person is fine.
• Date time: A date field

Note: These are automatically configured/created for you if you run any meeting agent/calendar setup and you don't have suitable fields from before. Follow the steps above for How to install another meeting agent to get all the right fields.

2. The Add meeting bot command

This is what controls the transcription operation during the live meeting (adding a bot, transcribing, where things go, what to do after)

To edit the command running this, press the configure button:

This opens the command in an adjacent panel. The name of this command node is what appears on the purple button on the meeting. Here it's called "Add meeting agent".

Nested within is the Add meeting bot command node. This is setup included in the default meeting tag:

Node filter: this grabs the Meeting bot status system-defined field and makes sure it isn't running yet by checking there's no status value set.

Meeting link, Transcript, Attendees, and Meeting date fields: These are self-explanatory in the description of the field. Plug in the field definitions for the fields you created above.

Post process command (important): This command will be run on the transcript

Other config settings that are not in this example

• Meeting bot name: The name for the agent when it appears as a participant in the meeting.
• Recording target: Field for inserting link to recording, if applicable
• Transcription provider: Choose a transcription provider, currently Gladia, or Deepgram

For all the possible settings, see the Add meeting bot command.

3a. The Text processing agent command: Default meeting

This is what controls the meeting notes processing. It defines what happens after the meeting is done and the transcript is ready for processing. A lot of flavour can be added here, depending on your needs.

This is the setup included in the default meeting tag:

Node context: This command will reside inside the Add meeting agent command, on the config level of the main meeting node. In order to target the main meeting node correctly, we need to specify GRANDPARENT here.

Node filter: The command will only run if the following criteria are met:

• There is some contents in the Transcript field
• The Meeting bot status is Done (i.e. it has left the meeting and done transcribing)
• The Text processing agent status (basically the status field for this command) does not exist yet

Transcript source: The field that contains the transcript

New entities target: The field that will contain Discovered items from the call

Action items target: The field that will contain tasks, and could contain other things like agenda items or decisions

Tags to use for entities: Entities that AI will look for when processing the transcript.

Tags to use for action items: Action items that AI will look for when processing the transcript.

Note: For the Tags to use for entities/action items, all the tags here would have special configurations (base tags set at a minimum, AI description for precision) to increase likelihood that AI correctly identifies these entities.

For all the possible settings, see the Text processing agent command.

3b. The Text processing agent command: Onboarding agent

Here's another example of how another type of text processing can be set up. This is an Onboarding agent that sits in on manual onboarding meetings with new users:

The node context/filters are the same

Text processing agent mode: Generic

• Generic mode does not take into consideration attendees, and is better at non-meeting processing

Tags to use for item extraction: Add a list of supertags the agent will try to identify from the transcript. Optimize each tag for AI by setting a base tag (if applicable) and AI instructions.

Action item/entities/extracted items prompt (optional): Additional prompt to use for extracting above items

Summary/New entities/Action items/Extracted items target: Set field to insert found items into

For all the possible settings, see the Text processing agent command.