Innovation beyond the hype

The world of AI is evolving at a rapid pace. New tools, models, and capabilities follow on swiftly. This is inspiring for us as developers, but it demands a careful approach for our clients. They expect innovation - but within a stable, secure, and scalable foundation. Topics like privacy, compliance, security and continuity are critical.

The challenge is clear: how can a digital agency continue to innovate with AI without falling into hype-driven experimentation? How do you ensure that every step forward is responsible, fits into the existing tech stack, and contributes to a reliable digital environment for the client? This tension forms the basis of our AI architecture and development strategy.

Microsoft, Azure and Umbraco

Our foundation consists of Microsoft technology, Azure cloud services, and the Umbraco CMS. This stack is robust, secure, and designed for scalability - this is what our clients rely on, built on years of proven agency experience. The challenge is to add new AI capabilities without making compromises to these fundamentals. We cannot simply experiment randomly or replace components.

Within our existing architecture, integrations must remain safe, manageable, compliant, and future-proof.

Our approach

This demands a tailored way of working for the development of portals, platforms, and sites. Our architecture emphasizes the following key principles:

• Human-in-the-loop
  AI supports the work, but the developer always retains control and makes the final decisions.
  
  

• LLM agnostic
  We don’t tie ourselves to a single language model, so we remain flexible and can always use the most suitable model.
  
  

• Tool agnostic
  The framework works with various AI tools, allowing experimentation without impacting our core architecture.
  
  

• Pragmatic adoption
  We combine tools and models in ways that align with existing workflows so AI becomes a natural and accepted extension of the work process.

Results

We have recently invested significantly in this approach and are now applying it in several projects, with clear outcomes:

• Up to 30% reduction in development effort
  AI supports structural and repetitive tasks, allowing developers to ocus on more complex challenges — increasing output and quality.
  
  
• Shorter time-to-market
  Beyond time savings, we see that we can go from idea to a working version faster. Where previously multiple iterations were needed to clarify a concept, we can now often show a clickable prototype or functional setup within one or two sessions; helping clients make decisions more quickly and enabling us to continue development sooner.
  
  
• More flexible teams
  Because AI helps generate both front-end and back-end components, the boundaries between different development skill sets diminish. Teams can work more flexibly and contribute more broadly to projects.

Next steps

Because AI is evolving rapidly, we continually adjust and refine our approach. This flexibility is intentional: our model must keep pace with new tools, insights, and capabilities so we can build smarter and more efficiently. In the coming period, we will focus on strengthening testing and quality assurance so that AI-generated code can be validated more quickly and reliably. We will also expand our CI/CD processes with new controls and automations that specifically address AI-assisted development. This will further professionalize our development workflow and increase the quality and predictability of every release.