Clankers and content operations

We’ve also said, “Content is data,” which is still true, but if we rewrote this tagline today, it would likely be “Content is context.”

It's basically the same idea, just with different words.

You can extract more value from your content when you author and query what your business knows—not what your website looks like.

The age of context management systems

CMSes have traditionally focused on workflows where one content operator modifies distinct web pages. Over the lifecycle of that piece of content, they regularly make more changes to that single page.

This doesn’t scale.

And it isn’t how developers work.

Developers write “source code,” which is simpler for humans to understand, and then pass it through a compiler, which creates the final output rendered by web browsers.

As development work becomes increasingly AI-enhanced, developers are increasingly writing product requirement documents or “specs” passed through an LLM to create output code. In an ideal world, the output code should not need editing, but updates to the spec should result in better output.

So why don’t content operators get to work like this?

It's time to let content teams write source content.

A place to write individual pieces of structured content that combine to create context and generate relevant and accurate content on demand on a massive scale.

We can change our thinking from a Content Management System towards a Context Management System.

What are content operations?

Another problematic sticking point of CMSes is that they’re often only used as a place for content teams to paste text they’ve prepared in some other location and press publish.

Think of all the work that is required leading up to that moment. All the research, tasks, approval processes, rounds of review, and more. These are all content operations.

Further, consider how much work needs to take place after pressing publish. Revalidate a cache, rebuild a site, perform workflow functions, translate, syndicate, and more. These are all content operations.

Sanity is for everything before and after pressing publish.

How Sanity fosters human and robot collaboration

Since launching in 2018, Content Lake has always powered real-time, collaborative authoring experiences, whether with multiple humans or robot tokens.

We didn’t know it then, but the availability of large language models came sooner than we thought. Thankfully, the underlying infrastructure was already in place to power content creation, where robots could work faster and at a greater scale than a human collaborator ever could.

This does not mean humans have a passive role while robots take the wheel. Instead, authors can fine-tune prompts and provide additional relevant context about their business and goals to better inform and guide content created by LLMs.

Sanity has built-in AI tools for schema-aware content creation, such as AI Assist, Agent Actions, and the new Sanity Agent.

Already in production

The concepts here aren’t theoretical or from the future. Sanity customers are already putting this into practice today.

Travel-tech start-up and Sanity customer loveholidays run almost their entire business logic through Sanity.

Their content model includes “prompts,” which are fine-tuned by content teams—not developers—to generate content at a scale otherwise impossible for such a small team.

Go beyond: Content Agent

The focus of this talk is about authors writing prompts that are stored within Sanity. But this is just one way to leverage AI in the content creation process—and creating content is only one benefit of AI tooling.

is a way to research, refine, query and create using your content as its context.