We’ve attended The Knowledge Graph Conference (KGC) numerous times and yet we’re always delighted to find that each year brings a refreshing perspective to a long-standing yet quickly-evolving industry. In this article, our Community Growth Manager, Larry Swanson, shares his report of this year’s KGC, specifically noting two talks that underscored the human-centric theme that recurred throughout the conference. Keep reading for his takeaways!
metaphacts at the Knowledge Graph Conference 2025 by Larry Swanson
I'm still basking in the afterglow of the Knowledge Graph Conference idea-fest and keep coming back to the talks by Tara Rafaat and Mara Inglezakis Owens, which I mentioned in my post-KGC LinkedIn post.
The Knowledge Graph Conference is one of the largest and most influential conferences for the knowledge graph community, bringing together folks from industry, research and academia who all simply love knowledge graph technology. But more than that, it is an opportunity to share cool knowledge graph applications and explore ways to advance the understanding and adoption of knowledge graphs across all sectors.
Human-centricity was a common theme at the conference, but Tara and Mara’s talks really drove home the importance of a people-first mindset in Enterprise Information Architecture practice. Tara Rafaat is the Head of Metadata and Knowledge Graph Strategy in the CTO Office at Bloomberg, and Mara Inglezakis Owens is an enterprise architect at a major US airline. Each has worked in both their current and prior roles on multiple enterprise-scale knowledge architecture initiatives.
Both Tara and Mara are thoughtful and diligent stewards of the resources in their purview. Tara’s role specifically focuses on strategy and long-term knowledge asset building, while Mara's current architecture role keeps her closer to the day-to-day concerns of building knowledge infrastructure for operations and commercial services that mature over time. So they brought nicely complementary perspectives to the topics of enterprise architecture and organizational change management.
Tara's talk, "Vision to Reality: What it Truly Takes to Build a Knowledge Graph in an Enterprise," tied together the complex variety of human activities that are needed to connect the semantic and data layers in an enterprise architecture.
Mara's talk, "Describing It All: Lessons Learned from Documenting Organizational Life," shared lessons from her extensive experience in enabling teams to capture and model enterprise knowledge.
Table of contents
- Pragmatic approaches
- Connecting knowledge work to business
- People first
- Grounding knowledge architecture in teamwork . . . and biology
- The importance of governance
- Conclusion
Pragmatic approaches
Mara and Tara focus on taking a pragmatic approach to knowledge architecture and engineering. That pragmatism revolves around understanding how users and stakeholders make decisions, how they approach knowledge management, and how to tie human effort to business outcomes.
Mara's pragmatism is driven by the need to understand and connect a diverse variety of business units and activities. She calls her approach "digital anthropology." Like any good anthropologist, she spends a lot of time in the field sorting out actual human behaviors from the things that people say that they do and want, and then writing up a report. This is the genesis of the subtitle of her talk: "Lessons Learned from Documenting Organizational Life."
Image: Slide from Mara Mara Inglezakis Owens' talk: Describing It All: Lessons Learned from Documenting Organizational Life
Tara's pragmatism is driven by the need in her role to align her strategically driven knowledge management with concrete business outcomes. This starts, of course, with people—more on that later—but she is also careful to focus on specific business benefits of her work over flowery business jargon. She avoids the "unicorns and rainbows" talk about how cool the technology is and focuses instead on how technology like knowledge graphs can concretely solve real business problems. She also tailors her messaging to the different stakeholder personas she encounters. Business, engineering and IT each have their own perspectives, and individual attitudes can range from skepticism to enthusiasm. She has found that accounting for this range of interests and attitudes is crucial to the long-term success of any knowledge-architecture project.
Connecting knowledge work to business
Mara also ties her work to business outcomes, focusing on documentable improvements in operational efficiency and on financial outcomes like cost reduction and quantifiable operational savings. In particular, she has found that tying knowledge work to the company profit and loss sheet is the best way to show the value of her work to the rest of the organization..
Tara has a pragmatic take on the conventional wisdom that iterative solutions are always the best. She points out that, while you are likely to iterate many times in any enterprise initiative, sometimes it makes sense to do significant up-front work. For example, some domains are small and uncomplicated enough that building out a complete ontology ahead of time might make sense. She also has a pragmatic and thoughtful framework for evaluating high-level architectural considerations like whether to bring a centralized, federated, or hybrid philosophy to a semantic architecture.
In her talk, Mara set out a number of practices around ontology design and governance. One that really jumped out at me was her advocacy for good UX in ontology design and the value of paying attention to system usability for business, engineering, analytics and other users. She ended her talk with very timely advice about how to think critically about the pragmatic needs of your organization versus the industry hype surrounding AI.
As a digital architect who earlier in my career was prone to budget-busting, “boil the ocean” solutions, I always welcome reminders of the importance of pragmatic thinking around socio-technical enterprise solutions. Tara and Mara bring that perspective in abundance.
People first
Both Mara and Tara gave among the most human-centered talks at the conference. (Not to take anything away from the other presenters. Comments about the crucial role of humans in the loop, the techno-social nature of semantic technology and other people-centric observations were common.)
The rise of generative AI has led to Silicon Valley hype and social media “thought leadership” speculating on the imminent replacement of human talent in the workplace. As a cantankerous human knowledge worker, I naturally bristle at such chatter. More to the point, as a close observer of the AI economy and as a long-time student of knowledge-management pioneers like Doug Engelbart and current researchers like Frank van Harmelen, I’m well aware that the more likely and desirable outcome will be hybrid human-AI collaborations that augment human intelligence. So my ears always perk up when presenters emphasize the human aspect of knowledge work.
Grounding knowledge architecture in teamwork . . . and biology
Tara opened her talk with a fantastic definition of a knowledge graph: "A knowledge graph is the DNA of information—it encodes not just facts, but the structure, relationships, and rules that bring raw data to life and allow it to grow, adapt, and evolve." This biological foundation got me thinking about biology and evolution and nicely contextualized her observation about the importance of finding and nourishing relationships with key stakeholders, especially in the C-suite, who have a "growth mindset."
In addition to highlighting the crucial role of executive support, Tara emphasized the importance of assembling the right kind of knowledge graph team. She accounted for the usual ontology, subject matter expertise, business and engineering roles but also added to this list a product owner to coordinate the team's activities and ensure the integration and adoption of the knowledge graph into the enterprise. She calls these five roles "the star team."
Image: Slide from Tara Rafaat’s talk: Vision to Reality: What it Truly Takes to Build a Knowledge Graph in an Enterprise
Mara also drove home the need for strong executive sponsorship of knowledge architecture projects. In addition to the financial metrics discussed earlier, Mara showed how measuring team performance based on impact delivered (as opposed to the volume of the work they do) both earns executive support and restores human dignity to productivity measurement.
The importance of governance
The main mechanism that enterprises use to encourage the human behaviors they want to see is governance. Both Tara and Mara had a lot to say about the topic.
Tara mentioned governance throughout her talk, showing how it aligns and helps to institutionalize all of the human effort that goes into a knowledge graph project and can keep everyone in the organization aligned on the original project vision.
Mara tailors governance practices to the nature of the enterprise. She'll take a very different approach depending on the culture of the organization. An agile startup that tolerates a "move fast and break things" mentality requires a very different governance program than an established, risk-averse company in a highly regulated industry.
When people talk about governance, they often sound like meddling bureaucrats or pedantic traffic cops. Not here. Tara and Mara both wove the topic matter-of-factly into their presentations. Another nice human touch.
Conclusion
The Knowledge Graph Conference in general and these talks in particular were a refreshing break from the current generative AI hype-o-sphere. Delivering reliable, trustworthy, explainable AI systems will require a huge amount of human insight and work. Together, these two talks offer a great roadmap for navigating the cultural and architectural challenges of building knowledge and AI architectures in big enterprises. As Mara said when she mentioned Tara's presentation in her talk, "You can use our talks together to get where you want to go."
Try it for yourself
Our enterprise knowledge graph platform metaphactory delivers semantic knowledge modeling and knowledge discovery capabilities to help customers to democratize and utilize knowledge across the organization. AI capabilities allow users to chat directly with the underlying data, extract trustworthy insights in context and build semantic models on the fly.
Speak with an expert to learn more about how metaphactory can support your organization’s information system needs!