What you will be able to do
- Describe the admin and user steps for setting up the Databricks Genie app for Slack and for Microsoft Teams
- Configure workspace, message visibility, and Genie scope, and apply the three-level visibility precedence
- Embed a Genie Agent in an internal portal with an iframe
- Match an agent usage scenario to the right user-facing interface
1.Bringing Genie into Slack
Some users would rather not leave the chat tool they already work in. For data questions, Databricks ships the Databricks Genie app for Slack, currently in Public Preview. Users can DM the app or mention @Genie in a channel or group DM. The question is answered by Genie One, or by a specific Genie Agent if a channel owner has pinned one. Both paths go through the Genie One API. Pinning only tells Genie One to send the question to that one agent.
Setup needs two admins and then the user. First, an account admin enables the Genie for Slack preview in the account console. If the workspace uses an IP access list, the Databricks control plane's outbound IPs for us-west-2 must be added to it. Next, a Slack workspace admin installs the app from a direct URL and approves its permissions. After that, each user connects, either by sending @Genie in a channel or by clicking Connect to Databricks on the app's Home tab, and then signs in to a Databricks workspace. Pinning a Genie Agent is optional.
Checkpoint 1 of 4· Put it in order
Put the steps for getting the Databricks Genie app for Slack working in a channel in order.
- 1.A Slack workspace administrator installs the app using a direct URL and clicks Allow
- 2.A user opens a channel and sends a @Genie message
- 3.The user logs in to their Databricks workspace when prompted
- 4.An account admin enables the Databricks Genie for Slack Public Preview in the account console
The account-level preview must be enabled before installation. A Slack admin installs the app, and then users connect by mentioning @Genie and signing in.
“An account admin must enable the Databricks Genie for Slack Public Preview in the account console.”Source: docs.databricks.com
In use, Genie answers each question with the credentials of the Slack user who asked, so Unity Catalog decides what that person can see. Responses include a View in Genie button that opens Databricks for more context, and charts appear as images. In a 1:1 chat, the app picks the most suitable Genie Agent and replies with sources and suggested follow-up prompts. One limitation to remember: you can add the app to a new group DM, but not to an existing one. Channel settings live behind /databricks-genie config.
Apps can't be added to existing group messages. They need to create a new group DM, add the Databricks Genie app for Slack when creating it, and then send @Genie.
Sources1
2.The same pattern in Microsoft Teams
The Databricks Genie app in Microsoft Teams (also Public Preview) follows the same design. Users DM the app or mention @Databricks Genie in a channel or group chat, and it can also be used from Microsoft 365 Copilot. Installation is different: a Microsoft 365 admin installs it from the Microsoft marketplace and approves its permissions, after an account admin has enabled the preview. If the workspace uses an IP access list, it needs the control plane IPs for East US 2. The first time a user asks a question, they click Configure, enter their workspace URL, and log in.
| Aspect | Databricks Genie app for Slack | Databricks Genie app in Microsoft Teams |
|---|---|---|
| Installed by | Slack workspace administrator via a direct URL | Microsoft 365 admin from the Microsoft marketplace |
| Mention in channels | @Genie | @Databricks Genie |
| Open channel configuration | /databricks-genie config | config |
| Audit request_params.source value | slack | teams |
| Answering engine | Genie One API, optionally a pinned Genie Agent | Genie One API, optionally a pinned Genie Agent |
In a Teams DM, sending /newchat clears the conversation context so the next message starts a new Genie conversation. Earlier Teams messages are not deleted. If a channel has no pinned Genie Agent, it uses Genie One by default.
Checkpoint 2 of 4· Check yourself
A Teams user wants their next question to the Genie app to ignore the earlier conversation, without deleting any messages. What should they send?
/newchat clears the current context so the next message starts a new Genie conversation. config opens settings and doesn't reset context.
“To start a new conversation in a direct message, send /newchat.”Source: docs.databricks.com
Sources2
3.Message visibility and auditing in shared channels
This is the main design decision for a chat-tool interface. Each channel's configuration covers three things: the connected workspace, message visibility (public or private), and Genie scope. Responses are public by default. Setting visibility to private limits each reply to the person who asked, but it also turns off visualizations, which can only be shown in public replies. If charts are missing from public Slack replies, a Slack admin must reinstall the app to grant it the files:write permission. Visibility can be set at three levels, and a higher level overrides a lower one. A workspace admin can turn off Allow public messages in collaboration platforms, and then every response is private.
Checkpoint 3 of 4· Match them up
Match each visibility level to how it is set and how it ranks.
Tap a term, then the definition that fits it.
The levels form a strict hierarchy: workspace overrides channel, and channel overrides personal.
“Message visibility settings can be set at three levels. Higher levels override lower levels”Source: docs.databricks.com
Admins can track usage from these interfaces. Each message logs a createGenieChatResponse event under the genieChat service in system.access.audit. The request_params.source field records where the request came from: teams, slack, or webapp.
SELECT
event_time,
user_identity.email AS user_email,
request_params.conversation_id AS conversation_id
FROM system.access.audit
WHERE service_name = 'genieChat'
AND action_name = 'createGenieChatResponse'
AND request_params.source = 'teams'
AND event_date >= date_add(current_date(), -5)
ORDER BY event_time DESC4.Embedding Genie in a portal, and choosing an interface
A third option is to embed a Genie Agent in an existing internal tool or portal (Genie Agents were formerly called Genie Spaces). Three conditions must hold. A workspace admin has configured the allowed embedding surfaces. The agent's author has at least CAN MANAGE permission. End users have explicit access to the agent and to every data asset it uses. The author then clicks Share > Embed space and pastes the generated iframe into the portal. Embedded users can send prompts but can't change the agent's configuration. Users who aren't signed in are asked to authenticate.
<iframe src="<your-genie-space-url>" allow="clipboard-write" width="100%" height="600"></iframe>| Scenario | Interface | Who sets it up |
|---|---|---|
| Custom agent code you version in git, with a chat UI | Databricks Apps agent template with built-in chat UI | Developer creates the app |
| Data questions answered inside Slack | Databricks Genie app for Slack | Account admin enables preview; Slack admin installs |
| Data questions answered inside Microsoft Teams | Databricks Genie app in Microsoft Teams | Account admin enables preview; Microsoft 365 admin installs |
| Genie inside an internal website or portal | Embedded Genie Agent iframe | Workspace admin allows surfaces; author with CAN MANAGE generates code |
Checkpoint 4 of 4· Check yourself
An agent author pastes the Genie embed code into the company intranet, but the iframe won't load the agent. The author has CAN MANAGE and users have data access. What is the most likely missing piece?
Embedding requires a workspace admin to allow the embedding surfaces first. clipboard-write only enables copy features and doesn't decide whether the agent loads.
“A workspace admin must configure allowed embedding surfaces before a Genie Agent can be embedded.”Source: docs.databricks.com
Sources3
Exam traps
Each one states something that sounds right. Open it to see what is actually true.
1.Since Genie uses the asker's credentials, a public reply in a channel is safe to show to everyone.Why is that wrong?
Credentials decide what data Genie retrieves, not who reads the reply. Public replies are shown to the whole channel, including members without access to the data, so use private visibility for sensitive answers.
Covered in Message visibility and auditing in shared channels
2.Pinning a specific Genie Agent to a channel bypasses Genie One and calls that agent directly.Why is that wrong?
Both modes use the Genie One API. Pinning only tells Genie One to send questions to that single agent.
Covered in Bringing Genie into Slack
Sources
Every claim above is drawn from one of these pages, quoted as it was written on the date shown.
- 1.
“Genie answers each question using the credentials of the Slack user who mentions @Genie”
↩︎ Bringing Genie into Slack“Existing group messages: Apps cannot be added to existing groups.”
↩︎ Bringing Genie into Slack“Visualizations are not shown when message visibility is set to private”
↩︎ Message visibility and auditing in shared channels“To keep a response visible only to the person who asked, set message visibility to private.”
↩︎ Exam trap 1“Both use the Genie One API; pinning a specific Genie Agent directs Genie One to send the question to only that agent”
↩︎ Exam trap 2“An account admin must enable the Databricks Genie for Slack Public Preview in the account console.”
↩︎ Checkpoint“including members who do not have access to the underlying data”
↩︎ Prediction“Message visibility settings can be set at three levels. Higher levels override lower levels”
↩︎ Checkpoint - 2.
“A Microsoft 365 admin installs the Databricks Genie app in Microsoft Teams from the Microsoft marketplace”
↩︎ The same pattern in Microsoft Teams“When Allow public messages in collaboration platforms is off, all responses are private to the user who asked.”
↩︎ Message visibility and auditing in shared channels“To start a new conversation in a direct message, send /newchat.”
↩︎ Checkpoint - 3.
“embedded users can send prompts but cannot edit the agent configuration.”
↩︎ Embedding Genie in a portal, and choosing an interface“Genie Agents were formerly known as Genie Spaces.”
↩︎ Embedding Genie in a portal, and choosing an interface“A workspace admin must configure allowed embedding surfaces before a Genie Agent can be embedded.”
↩︎ Checkpoint