Vassil Momtchev on Why Knowledge Graphs Are Critical to Building Context-Rich Enterprise AI

If your enterprise data is disparate and incomplete, no matter how big your AI investments are, you can't escape LLM hallucinations.

In this conversation, Vassil Momtchev, SVP Product & Technology of Graphwise, takes us through the growing significance of semantic AI and knowledge graphs, highlighting how these technologies inject meaning and context into generic LLMs and transform them into real business drivers.

Tell us about your role as SVP Product & Technology at Graphwise. How does it differ from your previous position as CTO at Ontotext?

As SVP Product & Technology at Graphwise, I drive the company’s product vision and strategy from conception to market success. I lead the complete product lifecycle, from market understanding to customer values, guiding our engineering teams in the execution and delivery of our innovative product roadmap. In my daily work, I align the needs of our customers with our technology and innovation plans to ensure they achieve measurable benefits. Working together with my peers in sales, marketing, and customer success, our goal is to create roadmaps that will benefit the demands of our users.

This role differs significantly from my previous position as CTO at Ontotext, where my primary focus was on technical architecture, research & development, engineering excellence, and building the most robust, scalable, and cutting-edge semantic technology possible. It was about perfecting the 'engine' of our solutions.

At Graphwise, I have the same objectives; however, my role has expanded to include overseeing the entire product offering and its market success. My focus is not just on how we build, but critically on what we make and why it matters to our customers and our business growth. I am more directly involved with overseeing market fit, customer experience, and ultimately, the commercial performance of our products. Put simply, my role is to bridge the gap between deep technology and tangible business outcomes, ensuring that our innovation directly fuels our competitive advantage and revenue generation.

What unique capabilities does Graphwise bring to the data landscape following the merger of Ontotext and Semantic Web Company?

Graphwise combines the two leading products in data—Ontotext's GraphDB and Semantic Web Company's semantic content management system, PoolParty Semantic Suite—to deliver a complete knowledge graph platform. This unified platform enables users to build knowledge graphs from disparate structured, unstructured, and semi-structured enterprise data, turning it into a rich, interconnected, and semantically aware format. This 'semantic layer' is crucial for grounding AI models, which significantly reduces LLM hallucinations and improves the reliability of generative AI applications.

Another key solution we offer is a business application called Graphwise for Microsoft 365. Leveraging our semantic AI tools, we revolutionize how organizations manage and interact with content within M365 and Copilot by:

• Enriching Search & Findability: Going beyond simple keyword search, we help organizations understand the meaning and relationships within their M365 content. This enables accurate semantic search, auto-tagging, and intelligent classification, making it fast for users to find what they need. Users can perform semantic queries beyond keywords, finding content based on its conceptual relevance. We also enable contextual discovery by providing rich metadata, facets, and relationships. This allows users to explore information intuitively, uncover connections that traditional search might miss, enhance productivity by reducing the time spent searching, and increase confidence in the retrieved results.
• Supercharging Copilot: Our knowledge graphs provide accurate domain-specific context to Copilot, ensuring efficient adoption in new commercial areas. We turn Copilot into an enterprise-specific expert rather than simply a general-purpose AI tool, which maximizes its value and the resulting impact for our customers.

For instance, a leading global pharmaceutical company with tens of thousands of employees has been leveraging our semantic AI platform for over four years. Their head of knowledge architecture in research credits their early adoption of knowledge graphs for preparing their company-wide data, combining insights from multiple heterogeneous data sources, and for giving them a significant technological advantage. Better yet, this foresight allowed them to deploy LLMs faster and more effectively, even within the highly sensitive pharmaceutical sector.

They currently have millions of internal files prepared in knowledge graphs, with plans to scale this to 20-30 million soon. The company emphasizes that while they constantly evaluate other technologies, our platform remains in the 'top league.'

In an industry where bringing a new molecule to market can take a decade and involves vast, varied datasets across different systems and languages, our tools help them link this diverse data and make it truly searchable. They stress that an LLM alone isn't sufficient for internal data due to its lack of context, praising our knowledge graph for providing the essential context and traceability to derive AI-driven insights and ensure answers are accurate and verifiable.

How critical are knowledge graphs for modern enterprises looking to unlock the full potential of their AI technologies?

Knowledge graphs are an important piece of the puzzle for modern enterprises looking to unlock AI's full potential. By building a solid and mature data foundation, knowledge graphs can:

• Control LLM Hallucinations: By grounding LLMs with precise, verified, and contextual enterprise data, knowledge graphs dramatically reduce factual errors and increase trust.
• Provide Context & Meaning: Knowledge graphs and their explicit model based on relationships between disparate information sources enable AI to control the meaning and connections within their business information.
• Explain Decisions to & from AI: Knowledge graphs capture data lineage and provenance of information, allowing users to understand why an AI algorithm reached a certain conclusion.
• Unify Disparate Information Sources: Knowledge graphs are an application-agnostic, data-driven architecture that serves as a central hub. They integrate all enterprise data and turn it into a coherent, queryable network.
• Supercharge Retrieval Augmented Generation (RAG): They enable exact and contextual retrieval for LLMs, feeding them semantically rich information for superior output. Knowledge graphs provide the AI-ready knowledge infrastructure that transforms raw data into intelligent, actionable insights, making AI reliable, accurate, and truly powerful.

For businesses struggling with data integration and quality, why is it worth investing in semantic AI technologies?

Data integration is the most expensive enterprise problem. In the era of AI automation, solving integration problems is becoming essential for future success. With their decentralized approach, knowledge graphs solve one of the biggest hurdles for many organizations—harnessing fragmented data spread across countless systems. They can intelligently integrate diverse structured, unstructured, and semi-structured data by representing their meaning through relationships and classes. This capability opens up traditional data silos, providing a single, consistent, and holistic view of enterprise information. Knowledge graphs also eliminate the need for complex, manual data integration and reconciliation processes, saving significant time and resources.

Marked by inconsistencies, inaccuracies, and a lack of context that can severely cripple AI initiatives, knowledge graphs help overcome traditional data quality issues as well. Leveraging semantic AI, knowledge graphs enrich data with meaning and enable automated classification, intelligent entity extraction, and the consistent application of business rules. This ensures that data isn't just present, correct, consistent, and deeply contextualized, but it is also truly AI-ready. This high-quality data is fundamental for reliable analytics and is particularly vital for large language models (LLMs), helping to significantly reduce hallucinations by feeding them precise,  factual information.

This is important because generic AI often falls short in complex environments due to a lack of specific domain understanding. Semantic AI provides this crucial context. By integrating with LLMs, it enables generative AI to retrieve highly accurate, verified information from internal knowledge graphs. This results in more relevant, reliable, and explainable AI responses that are closer to data and content summarization. Our approach boosts the ROI from our users' AI applications by delivering tangible business value rather than just plausible but incorrect outputs.

So, while data and content management solutions will continue to evolve with new technologies and formats, organizations will always have a need for high-quality information that is represented in a flexible format. That is the foundation of knowledge graphs, as they offer remarkable portability between various domains, integrating new data sources seamlessly and evolving their data models with greater agility. This ensures the data infrastructure has long-term relevance and effectiveness regardless of the current IT technology stack.

Investing in semantic AI technologies is about transforming data from a costly liability into a strategic asset. It's the critical step for businesses looking to move beyond simple data collection and processing. It also allows them to truly understand, leverage, and derive exponential value from their information in this dynamic age of AI.

Could you walk us through the new Graphwise for Microsoft 365 solution? In what ways does it drive enterprise-wide interoperability?

Graphwise for Microsoft 365 drives interoperability by creating a unified semantic layer across the business’s entire Microsoft 365 ecosystem. This provides:

• Consistent Data Understanding for All AI: We establish a shared understanding of an organization’s content by applying consistent, enterprise-wide semantic tags and building a central knowledge graph connected to the Microsoft Term Store. This shared vocabulary ensures that Copilot and any other AI application or business system that leverages this M365 data can interpret and utilize the information consistently, simplifying integration and maximizing the utility of their AI investments.
• Connecting the Unconnected: Our knowledge graph bridges the gap between siloed content and structured data. This allows AI to connect insights from an internal report in SharePoint to customer data in a CRM or to link a project discussion in Teams to relevant product specifications. We also create a more holistic view of enterprise knowledge that AI can leverage.

Graphwise for Microsoft 365 transforms the Microsoft 365 environment into an AI-ready knowledge hub by addressing the challenges of content sprawl, fragmented information, and the inherent limitations of AI operating on unstructured data.

At its core, Graphwise for Microsoft 365 leverages advanced semantic AI and knowledge graphs to:

• Automate Contextual Content Enrichment: We move beyond simple keyword search by applying sophisticated semantic AI to automatically analyze and tag all M365 content, such as documents, emails, chats, and more. These tags are derived from an enterprise's unique taxonomies and ontologies, ensuring consistency and deep contextual understanding. This process transforms raw, unstructured data into a rich, interconnected knowledge asset, making it fundamentally AI-ready.
• Enhance Microsoft Copilot with Grounded AI: While Copilot offers significant potential, its responses can sometimes lack precise, domain-specific context and/or occasionally generate inaccurate information when relying solely on the vast, undifferentiated sea of enterprise data. Graphwise for Microsoft 365 addresses this by feeding Copilot with verified, precise information from their enterprise knowledge graph. This means Copilot's answers are factually grounded because it leverages the organization's single source of truth data, significantly reducing inaccuracies and increasing the factual accuracy and trustworthiness of its outputs.

With domain-specific intelligence, Copilot gains a deeper understanding of the company's unique terminology, products, projects, and processes to provide highly relevant, context-aware responses tailored to the business’s needs. The solution also helps to evolve Copilot into a more effective organizational assistant. Thanks to the structured knowledge graph, we provide explainability to AI by providing Copilot with traceability. This helps users understand the source of Copilot's answers, fostering trust and enabling validation for critical business decisions.

How do you translate customer challenges into product opportunities, and then into scalable technology solutions?

We do this by keeping the customer’s needs first, starting with deeply immersing ourselves in the customer's world. We go beyond simple feature requests to truly understand underlying pain points. Through interviews, usage analysis, and feedback from sales & support, we actively listen to them. We also continually monitor market trends and competitors to identify gaps and validate that we continue to solve widespread, impactful problems.

Next, we synthesize these insights to define clear product opportunities. We articulate the specific problem, its target audience, and our solution's unique value propositions. Prioritization is next and is determined based on customer impact, market potential, and strategic alignment with our core Graphwise strengths, which results in a validated product roadmap.

Finally, we translate these opportunities by building robust, scalable solutions. Our engineering teams apply architectural design principles like modularity, cloud-native approaches, and API-first development for flexibility and integration. We use agile methodologies for iterative releases and continuous feedback gathering. We also monitor performance and user feedback post-launch to ensure continuous optimization and effective scaling.

What does the future hold for knowledge graphs and semantic AI? How do you plan to steer Graphwise in that direction?

The future for knowledge graphs and semantic AI is incredibly bright and increasingly central to the success of enterprise AI. We're moving beyond the hype of general-purpose AI models towards a recognition that context, accuracy, and explainability are paramount for real-world business applications. Knowledge graphs are the key enablers of this shift and will become the critical grounding layer for enterprise Generative AI (GenAI). While LLMs are powerful, they're prone to hallucinations and lack specific domain knowledge.

Knowledge graphs provide the verifiable, contextual, and up-to-date factual information that’s needed to make LLMs reliable for enterprise use cases, such as advanced RAG. This combination unlocks accuracy in chatbots, intelligent assistants, and provides automated content generation that can be trusted. The focus also changes from "big data" to creating "AI-ready data." By unifying disparate data sources, knowledge graphs break down silos, and create a holistic, interconnected view of all enterprise information including structured, semi-structured, and unstructured data. They also enable automated knowledge curation and enrichment, making the knowledge graphs more dynamic and adaptable and ensuring robust data quality and governance in the age of AI.

As AI moves towards autonomous agents that perform complex tasks, knowledge graphs serve as their long-term memory and reasoning engine. They provide the structured context these agents need to make intelligent decisions, interact effectively, and automate workflows across complex business processes.

With increased regulatory scrutiny and a demand for transparency, knowledge graphs will also be fundamental for building explainable AI systems. Their explicit representation of relationships allows for clear traceability of AI outputs back to their source data and reasoning paths. Ultimately, by providing intuitive, semantic interfaces over complex data, knowledge graphs will democratize knowledge access, making it easier for business users to understand and leverage enterprise knowledge without needing deep technical expertise.

Graphwise aligns perfectly with these trends, positioning us at the forefront of the Graph AI revolution. We'll continue to invest heavily in advancing our GraphRAG capabilities and developing more sophisticated methods for integrating knowledge graphs with LLMs. This will enable even more accurate, nuanced, and explainable generative AI applications for our customers across diverse industries. We also aim to make it easier for enterprises to build trustworthy GenAI assistants and applications.

Our core focus is building the most comprehensive and robust semantic layer for enterprises. This involves enhancing our platform to seamlessly ingest, enrich, and unify more diverse data types at scale. We'll also continue to provide tools for automated knowledge engineering, ensuring that creating and maintaining enterprise knowledge graphs is efficient and agile. While our platform is horizontal, we see significant opportunities to develop targeted, vertical domain models. Graphwise for Microsoft 365 is a prime example, and we'll explore similarly tailored solutions for other critical enterprise platforms and industry verticals where the need for AI-ready data and contextual AI is most acute.

We understand that knowledge graphs can appear complex, so our strategy will continue to include simplifying adoption and accelerating time-to-value. This means creating a pleasant user experience, providing more out-of-the-box solutions, and enhancing developer tools to accelerate deployment and demonstrate rapid business value. We accomplish this by placing a greater emphasis on intuitive UIs, leveraging low-code/no-code options where appropriate, and establishing robust integration frameworks. Finally, we believe in the power of an open ecosystem; consequently, we'll continue to ensure our platform is highly interoperable, with strong APIs and connectors to key enterprise systems and AI frameworks that enable our customers to integrate Graphwise seamlessly into their existing technology stacks.