Masterclasses at CMSCamp Mallorca 2026

Enterprise CMS and DXP architecture has undergone significant change over the past decade, moving from traditional monolithic platforms through decoupled and headless architectures, microservices and MACH, towards increasingly pragmatic and context-driven approaches.

This masterclass takes a broad look at where CMS and DXP architecture is heading, while examining some of the assumptions and architectural fashions that have shaped the industry along the way. Rather than advocating a particular architectural model, we will explore the trade-offs between flexibility, complexity, cost, governance and organisational capability, and consider how these should influence architectural decisions.

The session will explore four key areas:

Content structure and modelling
How approaches to structured content have evolved, including the relationship between content, experience and presentation. We will consider how much content should be modelled for reuse and omnichannel delivery, where content modelling becomes overly complex, and how different organisations can balance editorial needs with the demands of digital channels.

Stack composition and the evolution beyond MACH
The rise of composable architectures and best-of-breed technology has enabled organisations to select specialised capabilities such as CMS, PIM, DAM, search, personalisation and commerce. We will examine the benefits and hidden costs of increasingly fragmented architectures, and explore the emergence of more pragmatic "big lane" approaches, where clear capability boundaries exist without decomposing every function into an individual service.

Delivery models: headless, hybrid and templated
Headless delivery has become a mainstream architectural pattern, but it is not without cost. We will examine the trade-offs involved in headless and SPA-based approaches, including engineering effort, performance, caching, preview, governance and operational complexity. We will also consider why hybrid and templated delivery models are increasingly relevant, and how organisations can choose the right delivery model for each channel rather than adopting a single approach everywhere.

AI and the changing role of the CMS
AI is beginning to change both the capabilities expected of CMS and DXP platforms and the way content is consumed. We will look beyond AI-generated content to consider the architectural implications of AI agents, emerging standards such as MCP, and the possibility that CMS platforms will increasingly serve not only websites and applications, but also machine and agent consumers.

How do we choose the right architecture for the problem, rather than adopting the architecture that happens to be fashionable?

The session will conclude by considering the broader direction of enterprise CMS architecture: a move away from architectural certainty and prescriptive models towards pragmatic, context-driven decisions based on business outcomes, organisational capability and the value that architectural complexity actually delivers.

William Borg Barthet

Professional Services Director for CMS, Architect

William Borg Barthet is Professional Services Director for Bloomreach’s CMS and a senior architect within the Bloomreach team. With 16 years of experience at Bloomreach, William has helped shape some of the platform’s most significant customer implementations, leading architecture and delivery for flagship digital experiences across global organisations.

Beyond solution architecture, William has played an active role in driving product strategy, innovation, and customer success, working closely with customers, partners, and internal teams to evolve modern CMS capabilities. His expertise spans enterprise content management, digital experience architecture, and building scalable, personalised platforms that deliver business value.

Most AI coding demos stop at the diff.
This one doesn’t.

In this workshop we wire up a complete, governed software development lifecycle in which every moving part stays inside mittwald. The model that reasons about your code is a mittwald-hosted model. The platform that gets provisioned, deployed, and configured is mittwald. And the workflow that keeps the whole thing accountable is Spec Kitty.

Put plainly: mittwald’s inference drives mittwald’s platform — and Spec Kitty makes sure it happens in the right order, under your rules, with an audit trail you can hand to a reviewer.

What we plug together

mittwald AI Hosting — GDPR-compliant inference from German data centres over an OpenAI-compatible API. Qwen3-Coder-30B for implementation, gpt-oss-120b for reasoning and agent work, Ministral-3-14B for text and vision,
Qwen3-Embedding-8B behind RAG. No US data transfer, no prompt retention, no external providers in the chain.

The mittwald MCP server — your agent’s hands on the platform. Projects, apps, databases, domains and SSL, mail, security, backups, automation: all reachable in natural language, authenticated with OAuth 2.1 or an API token for CI. Not a wrapper around a few endpoints — full platform coverage.
   
Spec Kitty — the governance and workflow layer. A project Charter that encodes your rules, doctrine that carries your team’s hard-won conventions, specs and plans that survive a context window, work packages in isolated git worktrees, a deterministic runtime that decides the next step, and fresh-model plus human review before anything merges.

What we actually do together

1. Stand up a project Charter and doctrine — your architecture, security, and review policy written down where an agent will actually read it.

2. Point the agent surface at mittwald AI Hosting and connect the MCP server. First run: the agent inspects your live infrastructure and tells you what’s there.

3. Take one real feature from vague intent through specify → plan → tasks, with ambiguity surfaced instead of guessed.

4. Run implementation as parallel work packages in isolated worktrees, with a mittwald-hosted coding model doing the work.

5. Review with fresh eyes, accept, merge — then let the agent provision, deploy, and configure the result on mittwald through MCP.

6. Close the loop: mission review and retrospective, so what the team learned in this mission changes doctrine for the next one.

What you take home

A working, reproducible setup: Spec Kitty + mittwald inference + mittwald MCP on your own machine and your own project.

A Charter and doctrine starting point you can adapt to your stack — including legacy and vendor-specific constraints.

A defensible answer to the question your compliance people are going to ask: where did the code come from, who approved it, and did any of this leave the EU?

A deployment path where the agent doesn’t hand you a README — it does the provisioning.

Who this is for

Agencies and product teams shipping software that has consequences — customer systems, shops, platforms, regulated workflows. Developers who are already using AI agents and are tired of reviewing their output blind. Technical leads and CTOs who need speed without giving up control. Anyone who has to answer for digital sovereignty in writing.

Bring: a laptop, a project you actually care about, and we will provide mittwald demo accounts.

Robert Douglas

⁠Robert has spent decades building and scaling software organizations. Spec Kitty comes directly from seeing how easily intent drifts once teams, tools, branches, and AI agents all start moving at once.

One central change in Figma. Automatically updated and transferred directly into the CMS  — including all defined effects on components, states, and output channels.

Five brands can look completely different while still being built on the same underlying system. New websites can be launched faster. Changes no longer need to be discussed and implemented hundreds of times. They are decided once, centrally, and rolled out everywhere in a controlled way.

Used consistently, a design system is therefore far more than a tool for visual consistency. It becomes the central control logic for digital platforms — and a powerful driver of speed, scalability, quality, and cost efficiency.

In this masterclass, we provide a comprehensive and accessible demonstration of the true potential of a consistently structured design system.

Even companies that already work with a design system usually leverage only a small part of its potential. Key relationships are often not addressed systematically, while inheritance mechanisms and automation are missing.

As a result, significant value remains untapped.

We show what design systems are truly capable of. Step by step, a fully integrated system emerges: from atoms, molecules, and organisms to their practical application within a digital platform.

“I’m a button — and I can trigger 450 different discussions.”

Everything that should remain identical is systematically connected. Differences are created intentionally through defined values — rather than uncontrolled copies, manual adjustments, and individual one-off solutions.

This results in:

• far less maintenance and coordination effort
• significantly greater consistency
• manageable technical complexity
• reliable quality and testability
• faster continuous development
• significantly lower operating and development costs

Using a real-world example, we demonstrate how brands with very different visual identities can be based on shared components, which properties are centrally shared, and which values are deliberately defined for each individual brand.

This makes it clear how functionality only needs to be developed and tested once, while each brand retains its own distinct visual identity. The company can control its digital platform down to the smallest detail while remaining flexible enough to respond to new requirements, websites, and brands.

“Figma-to-CMS Pipeline”?

The transition from design to technology is particularly exciting. Using our “Figma-to-CMS Pipeline,” we demonstrate how changes made in Figma can be transferred automatically into the CMS design system. We explain which information is transferred, how it is processed technically, and how design and development become part of one integrated process.

We demonstrate the technical implementation partly through general principles and partly using Neos CMS as a practical example. However, the principles, ideas, and tools presented are relevant regardless of the CMS being used. They can also be applied to TYPO3 and other modern content management systems.

At the heart of the masterclass is one fundamental question: How should a design system be structured so that a company can continuously develop its digital platforms in a manageable, scalable, and cost-effective way?

The masterclass is led by Sven Ditz, Managing Director of sitegeist, and Martin Ficzel, Neos core developer at sitegeist. Alongside insights, inspiration, and practical examples, the instructors will also dedicate time at the end to individual questions, discussions, and conversations about participants’ own design system challenges.

Anyone managing multiple websites, platforms, or brands who wants to do more than simply document their design system — and instead use it as a truly systematic foundation — should not miss this masterclass.