What is AI-powered content operations?

What content operations means, what changes when AI agents join the system, and the five core concepts behind it in Sanity: structured content, governance, content applications, automation, and agentic context.

Content operations is the practice of running content as a coordinated system rather than as a series of one-off publishing tasks. It covers how content is modeled, who is allowed to change it, how a change reaches production, where the result is delivered, and how you tell whether it worked.

AI-powered content operations extends that system to AI agents. Agents read, draft, and act on the same structured content your team works in, under the same schema and the same permissions.

Sanity supplies the parts that make this work: a queryable content database, applications for people to work in, an automation layer that reacts to changes, and machine-readable context for agents. One set of content serves websites, apps, and agents, so you don't maintain a separate copy per channel.

With Sanity, you can:

• Model content around your business instead of around a page template, and reuse the same fields across every channel.
• Give writers a tool that fits the work: a structured editor in Sanity Studio, or a freeform AI-assisted draft in Canvas.
• Automate the repetitive parts of publishing with Functions that run on Sanity's infrastructure whenever content changes.
• Let AI agents work on your real content through Sanity Context, a hosted MCP server, and Agent Actions for generating and transforming content.
• Deliver to every channel from one source: query your content with APIs and SDKs for websites, apps, digital displays, and agents.

Five ideas underpin content operations in Sanity. They are not steps to complete in order, and each one shapes the others: what you learn from delivery can change your governance policies, and agent-assisted creation can change how you model content in the first place.

Sanity stores content as structured data rather than as pages or blocks of markup. You define document types and fields in code, and every field is individually addressable: you can query it on its own, reuse it in another channel, and change it without touching anything else. See Content modeling in Sanity Studio to define the shape of your content.

Addressability is also what makes structured content measurable. You can attribute performance to a specific field or component in the analytics and experimentation tools you already run, rather than to an opaque page.

Governance decides who can change what; workflow decides how a change reaches production. Roles and permissions control access to projects, datasets, and documents. Content Releases group changes that span many documents—a localization pass, a product launch, a scheduled campaign—so they publish together as one unit.

Model both around how your organization actually operates, rather than around the limits of a content management system (CMS). This is the layer that decides whether the rest of the system is safe to open up to more people, and to agents.

People work in applications, not in a database. Sanity Studio is a customizable, code-defined editing environment. Canvas is a freeform document editor for drafting, with AI assistance built in. Media Library manages assets across projects. When none of those fit, build your own application with the App SDK and run it alongside the rest in the Sanity Dashboard.

Functions run your code on Sanity's infrastructure when content changes, so enrichment, validation, and syndication happen without a person triggering them. APIs and SDKs deliver the result: query your content with GROQ from a website, a mobile app, a digital display, or an agent.

Every channel reads from the same source of truth. Adding one means writing a new query, not maintaining a new copy of the content.

An agent needs more than API access; it needs to know what your content means. Sanity Context is a hosted Model Context Protocol (MCP) server that gives agents structured, read-only access to a dataset. Agent Actions work in the other direction, generating and transforming content through the API against your existing schema.

Together they let an agent answer questions about your content and make changes to it, without a bespoke integration for every tool you want to connect.