
Insight · November 16, 2026
What changes when an AI agent stops waiting to be asked?
On November 11, 2026, Databricks announced an always-on version of its Genie agent. Here is what it does and what a smaller business can take from it.
- Announced Nov 11, 2026
- Runs in the background
- Starts from a goal
- Asks a person to decide
- Admin approvals and budgets
- No price or date given
- Announced Nov 11, 2026
- Runs in the background
- Starts from a goal
- Asks a person to decide
- Admin approvals and budgets
- No price or date given
The story
An agent that owns the task, not only the answer
- November 16, 2026
- Databricks, November 11, 2026
- TechTarget, November 11, 2026
- Insight
- By Silas Node, AONE Creative
On November 11, 2026, Databricks announced the new Genie. Its blog post describes an autonomous AI coworker that handles long-running tasks on your behalf, within guardrails you set. The Databricks release, carried by SD Times, calls it the second generation of Genie, an “always-on, autonomous agent for enterprise work.” According to the release, you give Genie a goal and clear boundaries. Genie then owns the work, delegates to subagents, tracks progress, follows up on missing inputs and contacts a person when a decision is needed. In autonomous mode it runs in the background on its own compute, and it emails you when you are away.
The blog says Genie can run on a schedule or respond to triggers, such as a metric dipping or more than 10 customers requesting the same feature. Its examples include monitoring insurance claims for fraud patterns and running marketing campaigns while proposing weekly budget changes for approval. People can work with it through Genie One, email, Slack and Microsoft Teams, and it can check in for input or approval. The release says Genie runs inside the customer’s existing Databricks environment, with every action governed by Unity Catalog permissions and Unity Gateway approval rules and budgets, and described as fully auditable. The blog adds that admin controls cover connector access, approval requirements, per-user budgets, spending caps and alerts.
Databricks also reports that Genie with its Genie Ontology answered 91% of task runs correctly, 1.6 times the rate of the stronger of two frontier agents, while using less than half as many tokens. That is a benchmark Databricks ran with Turing, and the post notes the comparison ran in the same Databricks workspace; it does not detail the method. Neither the blog nor the release gives a price or a general availability date for the new Genie. TechTarget reports two analyst cautions. Mike Leone of Moor Insights & Strategy wants Genie to say when it is not sure. “A wrong answer that shows up unprompted can reach a lot of people before anyone thinks to question it,” he said. Sanjeev Mohan of SanjMo said enterprises “need to govern what agents do, not just what they read.” Genie is built for Databricks customers and their data, and we have no connection to it. This account rests on the three pages named above.
An example of how it could work. Northgate Sample Outfitters is a fictional online shop with eleven people. Today its owner asks an assistant for a returns report every Monday. With the always-on pattern, the owner would instead hand over a goal: tell me when returns on any product pass a number I set, and draft a reorder only for items I approve. Each morning the agent checks. When one product’s returns jump, it opens a ticket, posts a two-line summary in the team channel and emails the owner a question: pause the ad for this product, yes or no? Nothing is paused, bought or sent without that yes.
Theories, in our view. First, the shift is from answers to follow-through. The useful unit of work becomes a goal, a boundary and a check-in, not a single question. Second, the boundary is the product. The analysts’ cautions point the same way: before any agent runs on its own, write down what it may read, what it may do and what always needs a person’s yes. Third, plain definitions matter more than the model. Mohan says an agent only works if it understands the business, and a small team can start with a one-page list of its own terms, such as what counts as an active customer or a return. These are our views, not findings from the sources above.
How AONE Creative would set it up. Our AI Services cover this kind of work. AI Integration and Automation connects AI to Google Workspace, Microsoft 365, your CRM, your website and social tools so repetitive work runs on its own, and Custom AI Agents are assistants that qualify leads, answer customers and help your team find answers in your own documents. We would start from the software your team already uses, scope each agent’s access to what its task needs, log what it does and put a human approval step in front of anything public-facing or high-stakes. Builds are scoped and quoted project by project. We have no connection to Databricks, and none of this describes how Genie works.
Where to start. Book a strategy call and bring one process your team repeats every week. If you want a ranked list of what is worth handing to an agent first, the AI Readiness Audit is Snapshot $1,500 or Blueprint $4,500.
Join the conversation
What is the one task you would hand to an AI agent, and what must it always ask you about?
Send us your answer, or book a call and we will look at it together.
What this means for you
Give one process a goal and a boundary
Bring one process your team repeats every week and we will tell you what an agent could own and what it should ask about.
Related
Where an agent could start
-
Service
AI Services
AI integration and automation, custom AI agents, platform development and content production.
-
Service
AI Readiness Audit
A review of your workflows and tools that ends with a ranked list of automations and a written plan. Snapshot $1,500 or Blueprint $4,500.
-
Service
Web Platforms
Custom, WordPress, Shopify and Wix sites with care plans.
-
Service
Campaign Digital
Digital strategy, paid digital and social channel management.
AI Services
More from the News Desk
- ServiceAI ServicesAudits, automation, custom agents and AI platforms built around how your team works.See AI Services
- AI Services, November 15, 2026Is your business claimed in ChatGPT yet?OpenAI launched Businesses in ChatGPT in beta on Nov 10, 2026: claim a profile, add product feeds, see insights. What it says and what to check.Read the story
- AI Services, November 14, 2026Box will build your workflow from a description. Are you ready?Box announced Architect, which will design agentic workflows from a plain description, with testing and human review first. Beta from Q4.Read the story
Start with AI
Put AI to work on one process
An AI Readiness Audit maps how your team works and ranks the automations worth building, whether or not you hire us to build them. Snapshot $1,500 or Blueprint $4,500.
Insight
What would you let an agent decide alone?
Tell us the one task you would hand over and the one thing it must always ask about. We will tell you where we would start.
