Phase 1: Macro Strategy & Knowledge Infrastructure

The new rules of marketing online: Why your website should be built on an AI Harness

Martin Anderson Clutz, Senior Product Marketing Manager, Acquia

The increasingly pervasive adoption of Artificial Intelligence has left few digital experience platforms untouched. But while many are quick to add AI-driven capabilities, there's a large question we should be discussing: In an AI-first internet, how does that change what a website should be?

As customers increasingly begin their buying journey using AI chat interfaces and zero-click searches, the strategies for online success that have served so many brands for so long are seeing diminishing returns.

In this talk we'll explore a variety of ways AI will impact the nature of what constitutes success: not just for content creation, but also how and in what format it should be discoverable by LLMs. We'll also explore how AI transforms how we should think about topics like personalization, analytics, and more. We will also identify key capabilities that are imperative to succeed in this new landscape, as well as the attributes to look for in a content platform that will help you meet the moment.

How to Optimize Your Content for AI at Scale

Christoph Breidert, co founder 1xINTERNET GmbH

Last year at FOST Paris, we introduced a framework for evaluating content against six signals that determine whether AI systems can find, trust, and cite it — Citation Readiness, Query Alignment, Content Structure, Authority & Trust, Semantic Clarity, and Technical Accessibility. This is the follow-up, and it's hands-on: how to actually build an AEO system like this in your own Drupal CMS.

We'll walk through the architecture step by step. How to model an assessment with clearly separated context fields (brand guidelines, target questions, etc.) so the AI evaluates against your organization, not generic best practice. How to structure prompts for each signal so an LLM can score reliably and consistently rather than producing vague, unusable feedback.
The hardest part of this system has nothing to do with the CMS: it's prompt engineering. AI is probabilistic, not deterministic, and getting it to produce reliable, actionable suggestions instead of generic praise or noise takes real work. We'll go through example edges that can break your prompts, explain why context engineering sometimes matters more than the prompt wording itself, and how to calibrate the right level of detail in a response — too little and the feedback is useless, too much can be perceived as AI slop.

You'll leave with a concrete blueprint for building this pattern yourself, in Drupal or any structured-content CMS, and the traps we hit along the way.

Phase 2: Editorial Governance, Quality & Starter Tools

Rethinking Knowledge Management in the Age of AI Agents

Wolfgang Ziegler, Managing Partner at CTO bei drunomics

AI agents are becoming part of the workforce, and accurate, up-to-date knowledge is what makes them powerful. Get it right and agents deliver expert work; get it wrong and errors come back as facts, complete with a citation. Most teams feel the problem already: knowledge scattered across tools, nobody sure what is still current, and people and their agents now producing updates faster than anyone can review.

We introduce a workflow that works: People and AI agents work side by side in one place, with approval that always stays with people. We show the supporting tooling: Markdown-first content, MCP built in, all open source and complemented with AI that keeps the knowledge consistent as it grows.

Knowledge management is shifting from productivity tooling to strategic AI infrastructure. The companies that rethink it now will get the most out of their AI agents in the years ahead.

Keeping Content Standards Consistent Across Hundreds of Pages: AI as Editor, Not Author

Bruno Bruno, Frontend Developer at 1xINTERNET

Most AI-and-content talks start with generation - the AI writes, humans clean up. This one makes the opposite case.

AI has pushed content production to warp speed, but oversight hasn’t kept up: on large Drupal estates, keeping standards consistent still means a senior editor reviewing hundreds of authors by hand. Slow, and impossible to scale. So instead of retroactive review, the highest-value AI doesn’t write it all - it reads, scores content against your own standards, and proposes fixes on Drupal’s open AI tooling, while editors keep authorship and the final word.

The AI Safety Net: Let AI Coordinate Changes Across Your Site Without Risking Production

Fabian Franz, AI specialist, VP of Software Engineering at Tag1

What would you let an AI agent change on your Drupal site if none of it reached your users until you were ready?

When you give AI full freedom to rewrite pages, populate sections, and coordinate changes across your whole site, then there is one catch. Whatever it builds can be visible to users the moment it finishes. That limitation is what should stop teams from putting AI near production.

Workspaces solves this. Everything inside is safe, reversible, and self-contained, because the complete result remains contained in the Workspace. There it can be reviewed, revised, or discarded before anyone publishes it.

Build the next version of your site with AI while the published site stays exactly as it is. When a generation fails, nothing breaks for your users and nothing is permanent until you say so.

Fabian Franz, one of the original architects of Workspaces and Tag1’s VP of Engineering, will demonstrate this live. Watch an AI agent reshape content across a Workspace and see the one deliberate step that puts those changes on your live site.

Come see what it looks like to build a Drupal site with AI and Workspaces.

What will you create first?

How Open Source Ecosystems Are Powering the Next Wave of AI

James Abrahams, Director at FreelyGive Ltd

We are all seeing an explosion in the power of true Open Source Ecosystems that will accelerate through 2027! Individuals are getting subscription fatigue and fed up with yet another AI tool they have to manage. Businesses and Enterprise are getting nervous of the threat of lock-in and lack of interoperability from proprietary software. Governments are finding that AI sovereignty is increasing a matter of national security that they would otherwise be at the mercy of changed in the geopolitical landscape.

As a result more and more they are choosing open-source first and even investing significant money into it. Open-source models are getting better than ever, the tooling and orchestration around AI is moving to open-source and as a result finished open source products are winning more than they ever have.

• Why Open Source AI is Winning: Discover how subscription fatigue, vendor lock-in fears, and national AI sovereignty concerns are driving individuals, enterprises, and governments toward open-source ecosystems.
• How Drupal Capitalized on Open AI: Hear how Drupal's structured architecture enabled instant wins with innovations like MCP, automated tools like Jev, and specialized AI cybersecurity agents that protect the entire ecosystem beyond core.
• Where the Technical Ecosystem is Heading: Get a front-row look at Drupal's latest integrations, including Symfony AI, headless front-end AI tools, and advanced evaluation suites like DeepEval and Inspect AI.

Jamie is a co-founder of a Drupal agency called FreelyGive and is obsessed with AI. Though not a programmer, he has been involved with the founding of the AI module, the AI Initiative and will come and share with you the history of AI in Drupal, what you can do today and where things are going next.

Phase 3: Sovereign AI & Regulated Enterprise Proof Points

How Clients Use Drupal AI in Production While Keeping Their Data Sovereign

Dan Lemon, Software Engineer at amazee.io

Drupal AI demos are everywhere, but what happens after the demo, when real editors, real compliance teams, and real content hit the system? For many organisations adopting Drupal AI, such as universities, government bodies, and enterprises, one rule came before anything shipped: their data must never be used in public AI training.

This session walks through real client use cases from teams we work with around the world. The stories are anonymised. Editorial teams use AI to categorise and tag thousands of pieces of content. Public sector teams rewrite complex text into simple language to meet accessibility rules. Some clients pair Drupal with our vector database to enhance their search with Search AI, so results understand meaning rather than just keywords, and answers respect who is allowed to see what. Others have replaced the traditional search button with a chatbot persona that guides visitors to the right place, whether that is the right page, a sign up, or a purchase. Agentic workflows are moving from experiments into daily routine. The common thread is that all of this runs on private, sovereign AI with zero data retention. That is what turned "we are not allowed to use AI" into "we use AI in production every day."

We will explore these patterns together in a live demo built on the same sovereign stack our clients run. You will leave with a realistic picture of what Drupal AI does well in production today, and how to ship it without losing control of your data. If your organisation is still stuck at "we are not allowed to use AI", this session is for you.

Adopting AI at the European Commission: Sovereign, zero-hallucination AI across 24 languages

Dominique De Cooman, Founder & CEO at Dropsolid AI

Most organisations treating Drupal as a CMS are sitting on an untapped advantage. EPSO and Dropsolid discovered it the hard way: when you need to deploy production-grade AI in a regulated institutional environment with full data sovereignty, auditability and zero tolerance for unverifiable answers, Drupal's API-first architecture and content modeling capabilities turn out to be exactly the right foundation. Not despite being a CMS but because of what lies underneath it.

This session tells the story of how the European Personnel Selection Office deployed a sovereign AI Answer Engine serving tens of thousands of recruitment candidates across all 24 official EU languages, with every answer cited back to the exact source document and correctness enforced at the model gateway level. No commercial AI platform. No proprietary lock-in. Built on Drupal, Mistral and a RAG pipeline running entirely on EU-sovereign infrastructure.

We will walk through how Drupal functions as the operational control plane: managing content ingestion and chunking, orchestrating the vector retrieval pipeline, exposing structured data to the language model and rendering cited responses back to users: all through a single administrative interface that EPSO's own team operates and owns. We will show why Drupal's content modeling made multilingual RAG tractable across 24 languages and how its API layer served as the bridge between institutional knowledge and AI inference.

AI Assisted Medical Research Reviews with Drupal

Vijay Mani, Solution Architect at Reading Room

The Clinical Practice Research Datalink (CPRD) provisions anonymized clinical data to research institutes and companies worldwide via its electronic Research Application Portal. Because of the sensitive nature of primary care data and linked datasets, every research application undergoes a rigorous, multi-stage review process involving administrative screening, Information Governance, lay reviewers, and subject-matter experts.

This case study explores how we integrated a standalone "AI Reviewer Service" into the Drupal portal to augment this highly complex, human-in-the-loop workflow.

By building an API-first bridge between our Drupal platform and an Node.js AI service powered by the AI SDK, we created an intelligent assistant capable of validating incoming research applications. The AI evaluates applications against strict domain constraints.

In this session, we will share the practical reality of making open-source AI integration work in a highly regulated environment. We will cover:

• APIs for AI : How we structured and served complex Drupal application data (including multi-study licenses, dataset requests, and protocol details) to an external AI agent.
• Human-in-the-Loop Workflows : The UX of integrating AI inside the CMS so that it assists, rather than replaces, the human review hierarchy.
• Structuring Domain Knowledge : Techniques for grounding the LLM in complex business logic, such as CPRD's strict rules around data usage, datasets.
• Governance & Trust : How we ensure AI-assisted outputs remain transparent and trustworthy in the context of healthcare data provisioning.

Phase 4: Agentic Architecture & Summit Wrap-Up

AI-Enhanced Is Not AI-Native — Rethinking CMS Architecture for the Agentic Era

Artem Dmitriiev, Senior Developer Drupal AI and 1xINTERNET

Most "AI-powered CMS" products today are existing systems with AI features bolted on: the human is still the operator, AI is still the assistant. But content volume, personalization demands, and the rise of AI agents as primary content consumers are breaking that model.

This talk draws on real architectural work to unpack what actually changes when AI, not a human editor, becomes the primary operator: content-as-source-of-truth over templates, governance and compliance checks built into every generation step, multi-provider AI integration (Anthropic, OpenAI, Gemini, self-hosted models) as a first-class requirement, and structuring content for AI agents (AIO) alongside humans.

Drupal's API-first architecture, structured content model, and open governance make it one of the strongest foundations available for AI-native systems, arguably ahead of most proprietary platforms racing to retrofit AI onto legacy cores.

The same architecture has a further advantage that few teams have yet recognized: with content and governance sitting behind Drupal’s typed API, external AI agents can act on it directly and can create or edit content under the same rules as a human editor, with no additional integration work required.In this presentation, we will share the technical challenges and considerations for preparing your CMS for brand consistency at scale, governance observability, and provider dependency in an agent-orchestrated web.