Why I’m Excited After ENDORSE 2025

We are at one of those rare moments where a set of technologies that have been evolving for years—semantics, ontologies, knowledge graphs, and large language models—are finally starting to align in useful ways. Add to that the emergence of Model Context Protocol (MCP) servers with HTTP endpoints, and suddenly the pieces fit together into something powerful and practical.

At Point Topic, we have been quietly building toward this for some time. Our work in broadband data, context assembly and ontology-based models gives us a front-row seat to what is coming next.

I’d also like to highlight the Common Telecoms Ontology that we’ve published - Login - PT Ontology Viewer.  You can check in as a guest and review some of the data, relationships and how knowledge graphs and associated ontologies are implemented in our data flows.

Lessons from ENDORSE 2025

The ENDORSE conference series (Homepage ENDORSE - Endorse - Publications Office of the EU) is an ongoing European forum dedicated to data, metadata, semantics, and interoperability. It brings together public sector experts, researchers, and technologists to explore how shared vocabularies, ontologies, and open standards can make data more usable and connected across institutions.

The goal is to move beyond isolated datasets toward linked, trustworthy, and reusable information ecosystems that support better decision-making, transparency, and innovation across Europe’s digital landscape.

ENDORSE 2025 brought together an extraordinary mix of data architects, ontologists, and applied AI thinkers. Some of the standouts included:

- EU Publications Office – impressive progress on tools already available for data publication and metadata management.  https://op.europa.eu/documents/d/endorse-2025/denis-dechandon_aniko-gerencser_vassilis-tzouvaras?download=true

- Semantic drafting for legal texts – open-source tooling that is transforming how laws are written and linked to structured data.  Real world, in use and a major time saver (according to those using it)  https://op.europa.eu/documents/d/endorse-2025/rui-ferreira_laurent-vinesse?download=true

- SHACL – a model-driven approach for validating ontologies and RDF data. It is essentially unit testing for data graphs—something we will be embracing soon.  https://op.europa.eu/documents/d/endorse-2025/thomas-francart_a-model-driven-approach?download=true

- Collaborative ontology development – used by the semiconductor industry but ready to adapt for telecoms and broadband.  https://op.europa.eu/documents/d/endorse-2025/mark-vanin?download=true

- AI-enhanced data modelling – still catching up with MCP, but showing how model-aware systems can structure and query complex domains.  https://op.europa.eu/documents/d/endorse-2025/emilien-caudron_georges-lobo?download=true

If you want to see our presentation, you can watch a version Endorse Oliver.  There’s plenty more now though.  We put this together nine months ago, and that seems like an age in the current evolution of LLMs and related machine learning tools.

Beyond the AI Bubble

There is a lot of froth in the current AI hype cycle. Much of what is being marketed as AI is really just well-packaged machine learning or statistical regression. It is very reminiscent of the internet boom of the late 1990s — plenty of noise and inflated promises, but also the seeds of something genuinely transformative.

Once the hype clears, there will certainly remain real value from connecting knowledge to context, and that is where ontologies, semantic models, and protocols like MCP will shine.

What This Means for Point Topic

Our mission has always been to research, clean, link, and contextualise broadband data. We have built dashboards, data windows, and analytical pipelines—but we are now moving into a new phase: context-driven interaction.

Using our UPC Query Agent (https://www.upc-query-agent.com/), subscribers can ask natural language questions about broadband availability, operators, or technologies and get precise, validated answers.  Our beta release is already in use, and we’re extending its capabilities and adding to its domain knowledge weekly.

This is a quick view of a query I ran earlier.  In reality this took around 30 seconds to generate.  I know this area very well and I can attest the data is accurate (I signed the wayleaves for the MDU myself).

Behind that simple interface sits a knowledge graph of UK broadband coverage, an ontology describing operators, technologies, and locations, and a Model Context Protocol server that dynamically assembles the right context for each question.

It is the bridge between our raw data and intelligent decision-making tools.

The Broader Picture

This is not just about broadband. It is about a wider shift in how data is published, linked, and validated. Using semantic technologies like SHACL and knowledge graphs, organisations can move from static datasets to living, queryable knowledge systems.

And with MCP integration, those systems become self-aware enough to talk to each other; reducing errors, explaining their reasoning, and even testing their own data integrity.

The Road Ahead

If we can deliver better data with richer context, our clients gain clearer insights and make faster, smarter decisions. Whether it is as simple as knowing how many households are in a postcode, or as advanced as a chatbot that predicts revenue potential for fibre rollouts, the path is the same: structure, semantics, and context.

The future is looking bright; we’d better wear shades.

If you want to book a meeting or find out more about anything in this post, just let us know and we’ll be right in touch.