Learn  /  What's new

Databricks Genie One grows up: Ontology and Search are GA, Genie answers in Slack, Teams, ChatGPT and Claude, and ZeroOps fixes broken pipelines

Databricks wants business users to ask their data questions wherever they already work, and data teams to spend less time fixing broken pipelines. One week of release notes shows how far it has got.

What's new8 min readUpdated 16 November 2026
DatabricksAI agentsWhat's new
All guides and cheat sheets

Genie One is the business-user front door to Databricks: you ask a question in plain English and Genie answers from governed data in Unity Catalog. In the week of 9 to 13 November 2026, Databricks' release notes added more Genie features than in any month before. Here is what landed, grouped by what it means for you. Our October Databricks update covers the previous month.

32Release notes9 to 13 November
6Places to ask GenieSlack, Teams, ChatGPT, Claude, Cursor, Chrome
50,000Requests per batchAI Functions asynchronous API
2Big GA launchesGenie Ontology and Genie Search

Context: Ontology, Search and Pages

An AI analyst is only as good as its understanding of your business. Three releases attack that problem from different sides.

Genie Ontology (GA, 9 Nov)
The context layer for Genie One, Genie Code and other agents. It joins what you model in Unity Catalog semantics (metrics, relationships) with what Genie learns from your tables and how people use them.
Genie Search (GA, 9 Nov)
Indexes documents in Google Drive, Jira, Confluence and Workday. Databricks stores each document's permissions with its content, so users only see what they already have access to.
Pages (GA, 11 Nov)
Governed definitions of business ideas such as "active customer", grouped by domain on the Discover page. Curators can certify or deprecate them; a new MANAGE PAGES permission controls who publishes.

Ask Genie where you already work

New ways to reach Genie One

  1. Slack and Microsoft Teams apps, GA

    Message the app directly, or mention it in a channel or group chat.

  2. Chrome extension, beta

    Every new Chrome tab opens Genie One in your workspace.

  3. ChatGPT, Claude and Cursor

    New Genie plugins connect these tools to the Genie One MCP server. An admin sets it up once; each user signs in with their Databricks account.

  4. Windows desktop app, beta

    The native Genie One app, already on macOS, now runs on Windows.

For the plugins, Databricks also lets security teams allow ChatGPT and Claude as "partner platforms" in context-based ingress rules (10 November), so you do not have to track their IP addresses yourself.

ZeroOps: an on-call engineer for pipelines

Genie ZeroOps (beta, 12 November) runs in the background. It monitors jobs, pipelines and tables, detects failures and data quality problems, investigates the root cause and proposes a fix. It proposes; a person still decides. That is the right design: an agent that silently rewrites production pipelines would be a governance nightmare.

  1. 1

    Start with one noisy pipeline

    Pick a job that fails often and whose fixes you know well, so you can judge the suggestions.

  2. 2

    Compare its diagnosis with yours

    For the first few incidents, write down your own root cause before reading ZeroOps'.

  3. 3

    Keep humans on approval

    Treat each proposed fix like a pull request from a new colleague: review, test, then merge.

  4. 4

    Measure

    Track time to fix before and after. That is the number that justifies keeping it.

For data engineers and admins

Generally available

Use now
  • Standalone materialized views and streaming tables from a notebook on serverless compute (11 Nov)
  • OAuth on-behalf-of for service principals, on by default: act as a service principal without creating a secret (12 Nov)
  • New jobs compute defaults to the Auto access mode (9 Nov)

Fewer secrets and fewer clusters to look after.

Beta and preview

Try in a test workspace
  • AI Functions asynchronous API: ai_parse_document, ai_extract or ai_classify on up to 50,000 requests from a JSONL file (9 Nov)
  • Date arithmetic in job parameters, for example {{(job.start_time - 1.day).iso_date}} for yesterday's date (9 Nov)
  • Pipeline flows as Unity Catalog objects with their own grants (12 Nov)
  • Budget alerts that call a webhook (10 Nov)

The date arithmetic alone removes a lot of glue code from daily jobs.

Plan for

Check your code
  • New models on Unity Gateway: Claude Fable 5.2 (9 Nov), Gemini 4 Carbon (10 Nov), Gemini 4 Hydrogen (12 Nov)
  • Gemini 3.6 Flash is retired on 19 November; calls move to Gemini 3.8 Flash at the same price

See Fable 5.2 vs Gemini 4 Carbon before you switch.

What to switch on first

  1. 1

    Write ten Pages

    Define your ten most argued-about business terms. Everything else, Genie included, gets better.

  2. 2

    Connect Genie where people already ask

    Slack or Teams first; most data questions start there.

  3. 3

    Pilot ZeroOps on one pipeline

    Low risk, quick to judge.

  4. 4

    Check the AI tools' access

    If you add the ChatGPT or Claude plugins, confirm in Unity Catalog what those users can read.

Check your understanding

1/3What does Genie Search index?

2/3What does Genie ZeroOps do when a pipeline fails?

3/3Which model is retired on Databricks on 19 November?

More: Snowflake vs Databricks for AI agents, Microsoft Fabric vs Databricks and this week's data platform news.

Sources: Databricks platform release notes, November 2026 (entries dated 9 to 13 November 2026). Features in beta or preview may change, and Databricks rolls releases out in stages, so your workspace may get them a week or more later.

Alessandro Ecclesie AgazziWritten by Alessandro Ecclesie Agazzi, analytics engineer in London. Updated 16 November 2026. About me

Found this useful?

New guides and "what's new" explainers on data and AI land here regularly. Follow along on LinkedIn, or share this one with a colleague.