Agentic AI is driving digital agency evolution
Over the past year, we’ve observed a major transformation among our digital agency partners. They are evolving from building websites to AI orchestrators.
We are supporting their transition by providing an AI-ready foundation that allows AI agents to safely perform complex tasks, such as content orchestration and real-time SEO automation on their clients’ sites.
Providing an AI-ready foundation
As a company rooted in the open-source community, our AI strategy is to nurture the growth of the ecosystem.
We’re sticking to our core values of transparency and flexibility. Twenty years’ experience in the open-source community has repeatedly shown how this enables innovation.
We believe AI works best as a partner to digital craftsmanship. To avoid locking our digital partners into a single large language model, we’re providing the architectural plumbing that allows digital agencies to choose the best models for their clients’ specific requirements.
Our AI strategy follows two distinct tracks:
AI in Umbraco – We provide a modular, model-agnostic foundation (Umbraco.AI) that empowers organisations to select which LLMs they use. This ensures transparency, control over token use, predictable costs, and strict governance over brand tone and data privacy, keeping the human-in-the-loop by design. Governance is central to the way we work with AI within the CMS.
Umbraco in AI – To make the CMS agent-ready, we’ve utilised the Model Context Protocol (MCP) and Agent Skills to enable external AI agents to understand and interact with Umbraco’s structured data. This allows developers and editors to work within AI-native environments, while Umbraco acts as the trusted source of truth. As an example, BackOffice Extension Skills ensure that AI agents connect to modern frameworks, the latest software, documentation, and examples.
Creating an AI blueprint
Umboto, our AI support agent, was developed as a blueprint, to prove that our AI-ready foundation works at scale.
By grounding Umboto in our proprietary documentation and ensuring it operates within our strict transparency frameworks, we’ve achieved a 24/7 global service model.
Umboto proves that, using AI-ready foundations to offer 24/7 support and maintenance, our digital agency partners can deliver instant value to a global audience without compromising on data security. It demonstrates how they can use Umbraco to scale their own businesses.
Open-source and AI
I often describe Umbraco’s open-source community as our superpower. As a case in point, the Umbraco MCP Server originated from work by open-source contributors who were exploring AI-assisted workflows. Their project demonstrated such technical depth and alignment with our own values of strong AI governance, transparency, and flexibility, that we invested in bringing their expertise in-house.
The community continues to be a crucial partner for scaling AI sustainably and reducing risk. We recently formed two community-based advisory teams comprising lead engineers and business leaders, who are working closely with us on the development of the agent-ready CMS, based on real user needs.
Enabling model sovereignty
Our users have absolute control over the models they’re using, where their data is going, and how data is used. We’re integrated with all of the major model providers, allowing users complete control over every aspect of their own integrations. That’s really important for governance, and cost management.
During a recent webinar discussing membership organisations’ application of AI, James Jemmett-Page, senior consultant at Cantarus, observed, “I’ve seen AI implementations in other products, including content management solutions, and some of them have a model where you’re baked into the CMS provider’s choice of LLM. Also, sometimes you see a CMS vendor reselling tokens and making a profit. So, I can certainly see a few benefits of the Umbraco approach, namely, that you have precise control over the models that you’re using. If a model supplier suddenly increases costs, you have the flexibility to jump to a different LLM. You have the ability to use different models for different purposes. You could, potentially, use a frontier model for some use cases, and a cheaper model, like Gemini Flash, for more prosaic purposes, like producing alt text for an image. It gives you that flexibility.”
Token efficiency
During the Cantarus webinar, Phil Whittaker, AI Staff Engineer at Umbraco HQ, predicted that token efficiency is likely to become increasingly important as organisations rely more heavily on AI-powered workloads, “I anticipate that models may become commoditised over the next couple of years. I wouldn’t really expect to use a PhD-level physics model for creating some text for a website. This will allow you to tailor and change your models and bring the cost of your tokens down over time as better models come out. Token efficiency is going to become far more important because, currently, we’re all being subsidised. Unfortunately, when model providers start to charge the real cost, it will increase.”
Sustainable-by-Design
The issue of token efficiency goes hand-in-hand with model efficiency. We remain dedicated to our philosophy of sustainability-by-design, to ensure that our company and community growth does not come at an unchecked environmental cost.
Our guidelines for AI coding agents help developers to ensure that they are energy-efficient and minimise carbon outputs.
Our digital agency partners are making use of Umbraco Cloud and our AI-ready architecture to reduce their clients’ carbon footprint and manual overhead of managing their global digital presences.
Optimising for longevity not lock-in
I’m excited to see how our partners use these tools as a starting point to create even greater impact for their clients.
We’re not trying to predict how everyone will use AI. Our role is to build an AI-ready platform that continues to work and deliver value when predictions change. We’re optimising for choice, longevity, and real-world use. Because once AI is part of normal infrastructure, that’s what partners and end clients will rely on.
