Salesforce Buys Contentful, and the Question Is Who Owns Agent-Ready Content

On 1 June 2026, Salesforce signed a definitive agreement to acquire Contentful. The deal is set to close in the third quarter of Salesforce's 2027 fiscal year, subject to regulatory approval. What stands out is not the buyer, and not the size of the cheque. It is the reason given.

Salesforce is not treating Contentful as a marketing add-on. It wants Contentful as a content layer for Agentforce, its agent platform, so that content can be assembled by machines and delivered across channels without a person shaping each output. The framing came most plainly from Contentful's own founder, Sascha Konietzke, who wrote that "AI agents now outnumber humans on the Web", and that companies have to rethink how digital experiences are made and delivered as a result.

That premise is the one the Machine Experience community has worked from since the start. Content is increasingly read by a machine before, or instead of, a person. A page built only for human eyes leaves the machine to guess at provenance, context, and meaning. The market has now put a price on the problem.

The open question is not whether content must serve machines. A founder said it on the way to an acquisition, and a platform paid for the capability. The open question is who owns the answer.

The Shape of the Answer

Salesforce's answer is a feature. Agent-ready content becomes something an enterprise buys, configured inside one vendor's stack and delivered through that vendor's APIs. For organisations already committed to that stack, it will work well. It also ties the machine-readability of your content to a commercial platform and to the roadmap of its owner.

There is another shape - one the MX books call content that manages itself. Not content administered through a platform, but content that carries its own context in the file: what it is, who produced it, when it changed, how it should be used. The file introduces itself to every system that reads it. The platform dependency falls away because the content no longer needs the platform to explain itself.

Provenance, context, and usage are recorded in the data file itself, in an open format that any system can read, governed by a body that sells nothing. That is what a COG is: a carrier-neutral container for a record in a data file, with the format published under an open licence and the schemas held by The Gathering. Attestation - the Ed25519 signature that lets a reader confirm where a file came from and that it has not been changed - sits in a separate trust layer, and stays narrow. It speaks to provenance and integrity, never to whether the content is true.

The distinction matters most at the point of lock-in. A platform feature works while you stay on the platform. An open record travels.

There is a discovery corollary. Agent-readiness as a platform feature means discovery happens through Salesforce's stack, on Salesforce's terms. Withdraw from the platform and the signal goes with it. A declared, attested record is readable by any agent that chooses to read it. Discovery follows from what the site has stated - not from membership in a register the site cannot see or apply to join.

The Sovereignty Question

There is a second-order point that European readers will notice first. Contentful was founded in Berlin. Once the acquisition closes, a German-founded content platform sits under United States law, including CLOUD Act jurisdiction. Open source and public-sector voices have already raised it.

For a regulated industry or a public body, the location of your content's governance is not a footnote. A standard that can run on-premises - including the air-gapped Local REGINALD configuration - keeps that governance where the organisation needs it. An open standard owned by no single jurisdiction is harder to acquire out from under you.

Built for Machines, Read by People

The case for machine-readable content is usually made in the language of agents and automation. The quieter part is worth saying out loud. Content structured well enough for a machine to parse is content structured well enough to meet WCAG. The Convergence Principle holds in both directions: an interface built for a machine tends to serve the person using assistive technology, and the reverse. Building for the agent and building for the screen-reader user are, more often than not, the same task done once.

There is an efficiency argument alongside it. A machine that can read a clear record does not have to infer one. Inference costs compute, and compute costs energy. A web that states its provenance plainly asks less of the machines reading it.

What Is Settled, and What Is Not

The Contentful acquisition helps the argument, whatever one makes of the buyer. A CMS founder has said in public that the web now reads differently, and a platform has paid to act on it. The premise is no longer the contested part.

What remains open is whether agent-ready content becomes a capability you rent from a platform, or content that manages itself - a property of the open web that any system can read and any organisation can govern. The Machine Experience community has placed its bet on the second.