August 27, 2023
NEWS
February 24, 2025
- Services
Datasheet
TL;DR: Our CEO made the case that proactive stewardship separates a maintained Drupal platform from a managed one. He is right. But stewardship lives on the operations side of the login screen. Once an editor signs in, a second question takes over: is the platform working for the people using it? Most enterprises have not asked it.
Stewardship Stops at The Login Screen
In "Proactive Drupal Support: Stewardship Beyond Patching," our CEO drew a clean line between maintenance and stewardship. Maintenance is reactive. Tickets get closed, SLAs get met, the platform is technically fine. Stewardship is the ongoing practice of treating a website as a living organism that gets safer, faster, and more capable month by month.
He ended with a sentence that has been on my mind since:
"Imagine an editor publishing an article in Drupal and getting on-brand-or-not feedback in the moment. Those tools exist, and we are building the bridge into the CMS now."
That sentence is the door. He opened it for the engineering buyer.
Now, we walk through that door.
Most enterprises have been rebuilding the operations side of their digital practice for the last decade. Observability, CI/CD, security posture, response SLAs, all of it has matured. The editorial side has not kept pace.
Marketing and content teams are running modern campaigns on a publishing layer designed for a quieter web, with AI bolted on the side and brand consistency held together by a PDF guideline document and goodwill.
The cost of that gap used to be invisible.
In 2026, it is not.
Image
WCAG 2.2 is increasingly written into procurement RFPs. Accessibility is no longer an afterthought. It is a commercial gate.
On brand, 60% of marketers using generative AI say they are concerned it could harm brand reputation through bias, values misalignment, or inconsistency, per Adobe's 2025 research. 12% of organizations report that content quality has decreased since they implemented AI, even as 84% report productivity gains.
Forrester research has long held that companies with consistent branding across touchpoints see 23% higher customer lifetime value, and 52% of senior marketing leaders at mid-size and large businesses report that poor brand consistency costs them $6 million or more in lost revenue annually.
The discovery you are trying to win, the legal exposure you are trying to avoid, and the brand consistency you are trying to maintain all converge on the same surface: the moment a piece of content gets composed and published.
That moment lives inside the CMS.
Three Shapes of Reactive Content Operations
Mohammed described three shapes of reactive support: the lone freelancer, the offshore ticket queue, the internal team being squeezed. After fifteen years of Vardot inheriting digital operations from other partners, we see three corresponding shapes on the editorial side.
The first is the disconnected toolchain. Scott Brinker's 2026 Marketing Technology Landscape lists 5,384 active vendors, and a typical enterprise stack runs around 120 of them. 62.1% of marketers report using more tools than they did two years ago.
The brand voice lives in one tool, the SEO checker in another, the accessibility audit in a third, the translation memory in a fourth, the DAM in a fifth, the analytics in a sixth. None of them speak to each other.
Editors spend 6 to 8 hours per week tracking down "where things are."
Adobe Summit 2026 coverage put the average annual waste from inefficient content processes at $2.5 million per enterprise.
The second is the bolted-on AI. Almost every CMS and DXP in the market now claims to be "AI-native." For many, that has meant adding a generate button to a CMS that was not originally designed to govern, structure, or trust the output.
The editor gets faster. The content gets less consistent. The brand absorbs the cost.
The third is the agency-dependent workflow, where every meaningful editorial decision has to leave the building. Brand checks happen on review cycles measured in days. Translation goes out as a project.
Accessibility is something a freelancer audits quarterly. SEO and GEO are someone else's department. The marketing organization is constantly waiting on a deliverable instead of producing one.
Each of these models produces output. None of them produces compounding capability.
The same pattern Mohammed described on the operations side, "work is not getting done," shows up here as a different symptom: 60% of marketing content goes unused, a number Forrester (formerly SiriusDecisions) has been reporting for over a decade and that has not budged.
AI did not fix it. In some cases, AI is making it worse.
What Changed in 2026
Two things changed at the platform layer that make a real fix possible now.
The first is Drupal CMS 2.0, released January 28, 2026. The Drupal Association called it the biggest evolution of the platform in 25 years, and the language fits. A new visual page builder called Canvas. A unified design system called Mercury. AI page generation. An admin chatbot for site-building tasks. AI-assisted alt text generation. An AI Dashboard for visibility into available providers and features. A recipe-based architecture that lets organizations adopt only the capabilities they need. AI features are provider-agnostic, supporting OpenAI, Anthropic, Google Gemini, Mistral, self-hosted Ollama, and sovereign EU options. None of it requires a license.
The second is the Drupal AI Initiative, which by February 2026 was backed by 28 organizations contributing more than $1 million in cash and engineering hours, with roughly 50 dedicated developers.
Vardot is one of those organizations, contributing as a Gold Sponsor. As a Drupal Diamond Certified Partner with 6,889 contribution credits, Vardot ranks among the top 20 Drupal contributors worldwide.
The initiative has shipped over 77 AI-powered modules and more than 30 AI agents.
In his October 2025 DrupalCon Vienna keynote, Dries Buytaert framed the moment in a sentence that has stuck with me:
"AI is the storm, but it is also the way through it."
Six months later in Chicago, he made the editorial argument out loud:
"The world is being flooded with AI-generated average. Average is cheap now, but expertise remains hard-earned and valuable."
The structural reason Drupal is positioned for that argument has nothing to do with marketing. Schema-enforced structured content has been shown to reduce LLM hallucination rates by 40 to 60% compared to free text, a finding that lines up with peer-reviewed work from ServiceNow's research team published at ICLR 2024. Drupal has been built around structured content for two decades.
That architectural choice, made long before LLMs existed, turned into the substrate AI now needs.
Image
The AI-Native CMS Maturity Model
After working inside Drupal editorial environments for sixteen years, I have come to think of CMS maturity in four levels. The model is not about features. It is about what the editor's day actually looks like.
Level 1: Reactive Content Operations
The CMS is a publishing surface. Brand checks, accessibility audits, SEO optimization, translation, and personalization happen outside the CMS, in separate tools, by separate people, often in separate weeks. Time-to-publish for a multilingual, accessible, SEO-optimized landing page is measured in weeks, not days. Editors copy and paste between systems. Versions drift. The brand guide is a PDF that nobody opens.
Most enterprises sit here and do not know they sit here, because their dashboards say uptime is fine.
Level 2: Augmented Workflows
AI has been added, but the CMS has not changed. There is a generate button, perhaps a grammar plugin, maybe an SEO assistant. Editors get faster at producing drafts. Quality variance gets wider. Brand consistency drops because every editor is now an unsupervised AI prompt. This is where the 12% of organizations reporting content-quality decline post-AI actually live. The platform is not aware of what is being published. The output is faster, not better.
Many of the "AI-native" capabilities visible in the market today still operate at this level. The architectural shift to Level 3 is what makes intelligence governable rather than just available.
Level 3: Integrated Intelligence
AI is wired into specific points in the editorial workflow. Accessibility checks run at publish. Brand voice review runs at draft. SEO and GEO suggestions surface during compose. Structured content is enforced. The CMS knows what is being published and reasons about it.
A team at Level 3 stops shipping out-of-policy content because the platform catches it before publish. Translation parity improves because the CMS knows when content has changed. Accessibility lawsuits drop because alt text and contrast and focus order are guardrails, not afterthoughts. Time-to-publish compresses because the editor is no longer a human router between disconnected tools.
Level 4: AI-Native Authorship
The CMS is the co-pilot. Editorial governance, brand context, accessibility, structured output, SEO, GEO, multi-language, and personalization are native primitives, not bolt-ons. AI is woven across the editorial lifecycle, governed inside the platform, with human-in-the-loop guardrails. The editor's job shifts from operating the CMS to directing the CMS.
A site at Level 4 is structurally ready to be cited by AI agents because it produces structured, sourced, governed content by default. It is structurally ready for the EAA because accessibility is enforced at publish, not audited quarterly. It is structurally ready for brand consistency because every editor publishes through the same context.
This is the level Drupal CMS 2.0 and Varbase are built for.
Image
What Level 4 looks Like in Practice
Varbase is Vardot's enterprise distribution of Drupal. The latest recipe-based generation of Varbase is built on Drupal CMS 2.0, and it ships with the editorial primitives that Level 4 requires.
Brand voice is governed inside the CMS. The Varbase AI Editor Assistant runs in CKEditor 5 alongside the editor as they compose. It drafts, summarizes, adjusts tone, checks grammar, all with brand context the team controls. The editor does not have to leave the page or paste their work into another tool.
Accessibility is automatic. AI Image Alt Text generates alternative text at upload and can run in bulk across an existing media library. The publishing workflow (Draft, Needs Review, Published) gives editors and reviewers a structured place to confirm before content goes live, not an audit cycle two weeks later.
Discoverability is built in. Metatag AI generates SEO titles and descriptions aligned to brand voice. Structured content underneath the editor surface is what makes content citable by AI Overviews, ChatGPT, Gemini, and Perplexity. Schema is generated from content models, not bolted on top.
AI is provider-agnostic. The underlying Drupal AI module connects to 48+ providers including OpenAI, Anthropic, Google Gemini, Mistral, Hugging Face, and self-hosted models like Ollama for sovereignty-sensitive deployments. The organization controls the model relationship.
Governance is the floor. AI-generated changes route through the same review workflow as human-generated ones. Audit trails exist. Roles enforce who can do what. The editor is not the last line of defense. The platform is.
None of this is theoretical.
Vardot has delivered Drupal platforms for 200+ organizations over 15 years, including UNHCR, UNICEF, BCG, Georgetown University, and Al Jazeera. The publishing workflow that ships in Drupal CMS 2.0 has been refined across that body of enterprise work, and Vardot is among the first agencies building an enterprise AI powered distribution on top of it.
Image
Four Questions for Your Next Digital Experience Review
Mohammed's article asked seven questions a VP of Technology can take into a QBR to diagnose whether their platform is being stewarded or just maintained.
Those questions are the right instrument for the operations side of the engagement.
For the marketing side, four questions are usually enough:
1. Where in your editorial workflow does AI catch a brand or accessibility violation today, before publish?
If the answer is "in a separate tool, after the fact," you are at Level 1 or Level 2, regardless of how many AI features your CMS markets.
2. Can a single editor produce a multilingual, accessible, SEO-optimized, on-brand landing page inside the CMS in a single sitting?
If the answer requires three tools, two reviewers, and a week, you are paying the disconnected-toolchain tax.
3. Who controls the AI provider relationship for your editorial workflow?
Provider-agnostic AI is an open-source ownership advantage: your organization chooses the model, swaps it when something better ships, and keeps the platform's intelligence portable. That portability is one of the practical reasons mission-critical organizations choose Drupal in the first place.
4. How is your CMS structured for AI agents and AI search to cite it?
Citation by LLMs follows structured content, schema, freshness, and authority. The CMS produces those by default or it does not. Drupal's structured content architecture, refined over two decades, is part of why it is well positioned for the AI search era.
If your current platform answers those questions clearly, you are at Level 3 or above and you are in a defensible position. If it does not, the gap is showing up in your traffic mix, your brand audit, your accessibility exposure, and your editorial velocity, even if it is not yet showing up on any dashboard.
Mohammed's piece made the case that maintenance keeps a Drupal site alive and stewardship keeps it moving forward. The same distinction holds on the editorial side. A CMS can let an editor publish, or it can help the editor compound capability with every post.
Stewardship of the platform and AI-Native Authorship of the editorial layer are the two halves of an enterprise Drupal engagement in 2026. One without the other is half a system.
Or, in the line from Dries that I keep coming back to:
"AI gets you to a prototype fast. Drupal gives it the foundations that last."
The login screen is where the first half ends and the second half starts. Both halves matter.
Where does your CMS sit on this curve?
Vardot offers a Drupal audit that diagnoses exactly that. Reach out to us for an assessment of your Drupal website. Or explore Varbase on Drupal CMS 2.0 to see what Level 4 looks like in production.
FAQs
An AI-native enterprise CMS is a content platform where editorial governance, brand context, accessibility, structured output, SEO, GEO, multi-language, and personalization are built-in primitives rather than bolt-on features. AI is woven across the entire editorial lifecycle and governed inside the platform with human-in-the-loop guardrails, so the editor directs the CMS instead of operating it. This is the difference between a true AI-native CMS and a conventional CMS that simply adds a generate button to a publishing layer never designed to govern, structure, or trust AI output. Drupal CMS 2.0 and Vardot's Varbase distribution are examples built for this model.
Structured content is what makes a web page citable by AI Overviews, ChatGPT, Gemini, and Perplexity, because citation by large language models follows structured content, schema, freshness, and authority. Schema-enforced structured content reduces LLM hallucination rates by 40 to 60 percent compared to free text, a finding consistent with peer-reviewed research from ServiceNow published at ICLR 2024. Drupal has been built around structured content for two decades, which means an architectural choice made long before LLMs existed became the substrate AI search now needs. Generating schema from content models, rather than bolting it on top, is what positions a CMS for the AI search era.
Drupal CMS 2.0, released January 28, 2026, introduced a visual page builder called Canvas, a unified design system called Mercury, AI page generation, an admin chatbot, AI-assisted alt text generation, and a recipe-based architecture that lets organizations adopt only the capabilities they need. Its AI features are provider-agnostic, supporting OpenAI, Anthropic, Google Gemini, Mistral, and self-hosted options like Ollama, and none of it requires a license. Separately, the Drupal AI Initiative, backed by 28 organizations contributing more than $1 million in cash and engineering hours, had shipped over 77 AI-powered modules and more than 30 AI agents by February 2026. Vardot contributes to the initiative as a Gold Sponsor.
A marketing team can assess whether its CMS is truly AI-native by answering four questions. First, where in the editorial workflow does AI catch a brand or accessibility violation before publish, rather than in a separate tool after the fact? Second, can a single editor produce a multilingual, accessible, SEO-optimized, on-brand landing page inside the CMS in one sitting? Third, who controls the AI provider relationship for the editorial workflow? Fourth, how is the CMS structured for AI agents and AI search to cite it? If the answers require multiple disconnected tools and multi-day review cycles, the platform sits at Level 1 or Level 2, regardless of how many AI features it markets.
Join the conversation +
Related blogs
March 27, 2018
May 24, 2021