Content Operations Agency: We Build the Supply Chain That Publishes to People and to AI

Six findings for the leader who reads only this box

  1. Agent access to the enterprise CMS became standard in 2026, with Storyblok shipping its MCP server on March 23, Agility CMS on May 5, and Contentful taking Remote MCP to general availability on July 21, followed by Salesforce completing its acquisition of Contentful on September 1, 2026.
  2. 45% of martech leaders with AI agents in pilot or production say vendor agents do not meet the business performance they were promised (Gartner, 413 respondents, fieldwork June to August 2025).
  3. Gartner predicts over 40% of agentic AI projects will be cancelled by the end of 2027, citing escalating cost, unclear business value, and inadequate risk controls (June 25, 2025).
  4. Marketers use about one third of their martech stack's capability, down from 58% in 2020 and 42% in 2022 (Gartner 2023 Martech Report, as reported by MarTech).
  5. 61% of organizations sit at content operations maturity levels 2 and 3 (Content Science Review, 2025, more than 100 respondents). In the 2022 wave, 2% reported level 5.
  6. In many cases a brand's own site is 5 to 10% of the sources AI search references (McKinsey, October 16, 2025). Every publish event now serves two audiences: people and the AI systems that decide whether people find it.

A content operations agency builds the supply chain from idea to publish

Content operations is the system that moves content from strategic idea to a published page, post, email, kiosk screen, or ad, and keeps it accurate afterward. We build and run that system for large organizations. The enterprise content operations guide covers the discipline in depth; this page covers what we do about it.

Agentic content operations means AI agents perform the repeatable parts of that system under a named human owner: research, first drafts, tagging, metadata, schema, localization prep, link maintenance, and QA against a written standard. People decide what to say, set the bar, and approve. Agents do the volume.

So what makes us different from a content marketing agency is the unit of work. An agency sells you articles. We build the operation that produces them, measured on cycle time, coverage, and whether AI engines cite the result. We run our own marketing this way, which is how this page was produced.

Two audiences read everything you publish now. People see the rendered page. AI systems read the whole stack underneath it and decide whether to cite you. McKinsey found that in many cases a brand's own site is only 5 to 10% of the sources AI search references (McKinsey, October 2025). A supply chain built for one audience leaves the other to chance.

Agent access to the CMS is standard; the operation underneath decides the result

All three platforms we build on exposed an agent interface within four months of each other in 2026. Storyblok published a production MCP server on March 23. Agility CMS launched its MCP server with 19 tools on May 5. Contentful took Remote MCP to general availability on July 21, and on September 1 Salesforce completed its acquisition of Contentful. The tooling question is settled.

Exhibit 1: Every Dotfusion partner platform shipped agent access in 2026
Platform What shipped Date Vendor wording
Storyblok Storyblok MCP Server March 23, 2026 "connects any MCP-compatible AI agent directly to Storyblok's content and management layer"
Agility CMS Agility CMS MCP Server May 5, 2026 "19 tools that cover the full content management lifecycle"
Contentful Remote MCP, general availability July 21, 2026 "bringing Contentful into AI workflows at enterprise scale"
Contentful Acquisition by Salesforce completed September 1, 2026 "the Company completed the acquisition of Contentful pursuant to the Merger Agreement"

Sources: Storyblok; Agility CMS; Contentful; Salesforce Form S-8. All opened September 17, 2026.

What is not settled is whether the operation around those interfaces is ready. Gartner's survey of 413 martech leaders found 45% of those with AI agents in pilot or production say the vendor agents do not meet the business performance promised (Gartner, October 29, 2025). Gartner separately predicts over 40% of agentic AI projects will be cancelled by the end of 2027 for escalating cost, unclear business value, or inadequate risk controls (Gartner, June 25, 2025).

Our reading of those two figures is operational. An agent connected to a CMS with no typed content model, no written editorial standard, and no owner for accuracy produces more content faster and no better. We treat the agent as a multiplier on whatever operation it lands in.

Twenty years of martech says tools get bought and operations do not get built

Martech utilization fell from 58% in 2020 to 42% in 2022 to about one third in 2023, by Gartner's own annual survey, while stacks kept growing (MarTech, reporting the Gartner 2023 Martech Report). The content operations maturity data tells the same story from the other side. In Content Science Review's 2025 study, 61% of organizations sat at levels 2 and 3 of a five level model, and in the 2022 wave only 2% reported level 5 (Content Science Review, 2025; 2022 fact sheet).

Exhibit 2: Stack capability went unused for three straight years before agents arrived
Year Share of martech stack capability marketers report using
2020 58%
2022 42%
2023 About one third

Source: Gartner Martech Reports 2020 to 2023, as reported by MarTech. Opened September 17, 2026. Gartner's 2023 report is behind a paywall; the figures here are the secondary source's.

The lesson we take from this is operational, not technical. Each wave of tooling landed on teams whose process, model, and governance were never designed to use it. Agents are the fourth wave, and the fastest, so the gap between what was bought and what is operated widens faster too.

We build five parts of the content supply chain, each with an owner and a metric

A supply chain that publishes to people and to AI systems has five parts. We design, build, and run all five, and every one leaves the engagement with a named owner on your side and a number that person reports.

Exhibit 3: The five parts of an agentic content supply chain
Part What we build Owner on your team Metric
Content model Typed, governed content types in Contentful, Storyblok, or Agility CMS; author and organization entities; structured fields that produce schema at the template level Content architecture lead Structured field population rate; schema coverage
Editorial standard A written definition of done: evidence rules, voice rules, answer first structure, source requirements Editorial lead Share of published items passing the standard first time
Agent workflows Research, drafting, packaging, QA, and publishing agents connected to the CMS through its MCP server, each with a scoped task and a human approval gate Content operations manager Cycle time from brief to publish; agent output acceptance rate
Governance Roles, approval paths, partner and compliance rules, an audit trail for every agent action COO or marketing operations Items published outside process (target: zero)
Measurement Query coverage, brand mention rate, and citation share on your priority query set; GA4 AI referral channel; Search Console CMO Queries moved from gap to cited

Source: Dotfusion practice model, September 2026. Measurement definitions follow our AEO measurement framework.

The platform matters here, and we say so as a partner of all three. A composable, API first CMS produces clean delivery, template level schema, and author entities as default outputs. A legacy monolith produces them only through sustained manual effort. Our enterprise decision framework covers how to choose among the three, and our headless CMS practice builds the platform when a migration comes first.

An engagement moves from assessment to a running operation in four stages

We start with an assessment, because 61% of organizations are at maturity levels 2 and 3 and the fix at level 2 is different from the fix at level 3. Then we design, build, and run.

  1. Assessment, four weeks. We map the current supply chain from idea to publish: cycle time, handoffs, tool count, model quality, and where AI already touches production. Output: a one page current state map and the maturity level, with the three constraints that matter most.
  2. Design, six to eight weeks. We write the editorial standard, design the content model changes, and specify each agent workflow with its human gate. Output: the supply chain specification your team and ours build against.
  3. Build and pilot, eight to twelve weeks. We build the model, connect agents through the CMS MCP server, and run one content stream end to end in production. Output: a working workflow with measured cycle time and acceptance rate.
  4. Run. A monthly retainer covers agent maintenance, standard updates, measurement reporting, and expansion to the next content stream. Your named owners run it; we keep it sharp.

Timing depends on platform state. An organization already on Contentful, Storyblok, or Agility CMS with a clean model moves faster than one mid migration. When the platform work comes first, we sequence the content operation into the migration rather than after it.

What CMOs, CDOs, CIOs, and COOs each do first

CMO. Pick the twenty commercially important queries your buyers ask AI engines and test them this week in ChatGPT and Gemini. Record where you are named and where you are absent. That list becomes the priority query set every content decision is measured against.

CDO. Audit whether content is typed by meaning or by layout. Open ten high value pages in the CMS and count how many carry a populated author entity, a structured summary field, and a machine readable type. Agents inherit the model they find.

CIO. Confirm whether your CMS exposes an MCP server or equivalent agent interface, which version, and who holds the credentials. If the answer is a monolith with no agent surface, the platform decision is now on the content operations critical path.

COO. Put content governance on the org chart before agents add a second producer to the model. One named owner for accuracy, one for the editorial standard, one for the agent audit trail. Gartner's 45% figure is the reference point for what an unowned operation reports back.

Three ways the next 18 months play out

Scenario 1, agents bolted on. The team connects vendor agents to the existing CMS with the existing model and no written standard. Volume rises within the first quarter. Watch for: rising correction rates, duplicate pages, and no movement on the priority query set. Gartner's 45% and 40% figures are the reference points for this scenario.

Scenario 2, operation first. The team spends the first quarter on the model, the standard, and one governed workflow, then scales. Volume is flat for 8 to 12 weeks, then rises; our target is an acceptance rate above 80% on agent output. Watch for: the first priority queries moving from gap to cited within two quarters.

Scenario 3, wait. The team holds until the platform market settles. Given the Salesforce close on September 1, 2026 and three MCP launches between March and July, the period being waited for has already passed. Watch for: competitors appearing in AI answers on your category queries.

Who we do this for

We have run the digital presence of First Canadian Place for Brookfield Properties for more than twenty years, our longest client relationship. We deliver content operations and AEO for Westman Communications, a telecommunications cooperative. Our headless practice has shipped production work for Mitsubishi Electric Canada, Oxford Properties, and Financial Horizons Group on Contentful, Storyblok, Agility CMS, and Magnolia.

We are a Contentful Silver tier partner, a Storyblok Silver Partner, and an Agility CMS Implementation Partner. Dotfusion Technology, Inc. has been a Certified B Corporation since May 2021 with a B Impact score of 89.4 (B Lab). We have been building for large organizations from Toronto for 27 years.

Method note

Analysis as of September 17, 2026. Every external figure on this page was fetched from its source on that date and is linked inline. Gartner's 45% figure comes from a survey of 413 martech leaders with fieldwork June to August 2025. Content Science Review's 2025 study drew on more than 100 content leaders and team members; its 2022 fact sheet did not disclose sample size, and the two waves are reported separately here for that reason. Martech utilization figures are Gartner's, taken from MarTech's report because Gartner's 2023 study is paywalled. The engagement durations in the four stage model are our standard ranges and vary with platform state.

Frequently asked questions

What does a content operations agency do?

A content operations agency designs, builds, and runs the system that moves content from idea to publish and keeps it accurate: the content model, the editorial standard, the workflows, the governance, and the measurement. It is measured on cycle time, coverage, and discoverability rather than on articles delivered.

What is agentic content operations?

Agentic content operations is a content operation in which AI agents perform the repeatable work, such as research, first drafts, metadata, schema, and QA, under a named human owner and a written standard. People decide what to say and approve what ships. Agents do the volume.

Do we need a headless CMS before starting?

No. The assessment and the editorial standard apply to any platform. Agent workflows connect most cleanly to a CMS with an agent interface and a typed content model, which is why platform state sets the timeline, and why we sequence content operations into a migration when one is coming.

How long until the first agentic workflow runs in production?

Our standard range is 18 to 24 weeks from the start of the assessment to one content stream running end to end in production, faster on a clean model and an existing MCP enabled platform.

How do you measure success?

Cycle time from brief to publish, agent output acceptance rate, structured field population, and movement on your priority query set: the count of commercially important queries where AI engines name you, tested in ChatGPT and Gemini and reported monthly.

Sources

All pages opened September 17, 2026.

This page was conceived and written by Chris Bryce, CEO and Partner at Dotfusion, with our stack of AI research agents doing the heavy lifting on sources and data. It is an example of the operation described above. If you want to talk about content operations, AEO, or enterprise web strategy, talk to us. We love this stuff.