Inna Tokarev Sela, the CEO and Founder of Illumex, is transforming how enterprises prepare their structured data for generative AI. Illumex enables organizations to deploy genAI analytics agents by translating scattered, cryptic data into meaningful, context-rich business language with built-in governance.
The platform automatically analyzes metadata to locate and label structured data without moving or altering it, adding semantic meaning and aligning definitions to ensure clarity and transparency. By creating business terms, suggesting metrics, and identifying potential conflicts, Illumex ensures data governance at the highest standards.
With Illumex, analytics agents can interpret user queries with precision, delivering accurate, context-aware, and hallucination-free responses. Under Inna’s leadership, Illumex is setting a new benchmark for AI readiness, helping businesses unlock the full potential of their data.
What inspired you to found illumex, and how did your experiences at Sisense and SAP shape your vision for the company?
The vision for illumex emerged during my studies, where I imagined information being accessible through mindmap-like associations rather than traditional databases – enabling direct access to relevant data without extensive human consultation.
My time at SAP taught me the fundamentals of building enterprise software and scaling operations. Working across product development with SAP HANA cloud platform and business initiatives like the startup partnership framework gave me deep insights into enterprise customer needs. It revealed a significant gap between how companies approach data practices and what end users actually need.
At Sisense, building the AI practice from scratch demonstrated the immense value AI could bring to customers. Seeing this impact, combined with the rise of SaaS and GenAI technologies, convinced me the timing was right to launch illumex in 2021.
illumex focuses on Generative Semantic Fabric. Can you explain the core concept and what motivated you to tackle this specific challenge in AI and data analytics?
illumex pioneered Generative Semantic Fabric – a platform that automates the creation of human and machine-readable organizational context and reasoning. This platform unifies the experience of both LLM-based generative AI and business applications for technical and non-technical users around shared context.
This single fabric delivers two major benefits: it streamlines data management through the automation of up to 80% of data engineering tasks and enables non-technical users to access analytics with built-in governance, explainability, and accuracy. Both of these benefits address a multi-billion dollar market for enterprise decision-making.
Think of it as a digital playground where machines, humans, and applications interact spontaneously without pre-programming. This aligns with our vision of an application-free future, where instead of juggling multiple tools like sheets, analytics, financial systems, and customer amanagement, you simply express your task, and it’s completed seamlessly. Generative Semantic Fabric is the foundation for this future.
What were some of the key challenges you faced in the early days of illumex, and how did you overcome them?
In 2021, despite the fact that generative AI semantic models have existed since 2017, and graph neural nets have existed for even longer, it was a tough task to explain to VCs why we need automated context and reasoning. Even defining it back then was a tough task.
I would say the biggest challenge was to really spring this excitement about this future technology and future market. And I was very fortunate to meet forward-thinking investors who believed in me.
How does illumex empower organizations to become AI-ready, and why is this transition critical in today’s business landscape?
The business world is splitting into two camps: companies that recognize and capitalize on AI as a transformative force akin to the Internet and those that miss or delay understanding this opportunity.
illumex meets organizations wherever they are in their AI journey. We prepare their data for generative AI implementation, augment and govern organizational logic and context, and enable the deployment of agent analytics and orchestration.
Our full-stack GenAI implementation platform for structured data elevates any company’s landscape to effectively leverage these advanced technologies.
illumex emphasizes “hallucination-free” generative AI responses. How does illumex ensure deterministic and reliable outputs?
illumex builds on pre-existing business ontologies – knowledge graphs capturing industry-specific terminology, workflows, and processes across sectors like pharma, retail, and manufacturing, as well as business functions like finance, HR, and supply chain.
When onboarding customers, we automatically retrain these ontologies on their metadata. Within days, companies can search their data, validate results, and identify issues like duplicates or conflicts.
The agentic analytics chatbot provides complete transparency – showing how questions are interpreted and mapped to the customer ontology and then to data. This transparency, combined with automated data validation, ensures deterministic, hallucination-free answers. Additionally, governance teams can pre-validate potential responses since the context embeds all possible questions and their permutations in advance.
How does illumex differentiate itself from traditional approaches like Retrieval-Augmented Generation (RAG)?
While RAG attempts to customize off-the-shelf AI models by feeding them organizational data and logic, it faces several limitations. It’s a black box – you can’t determine if you’ve provided enough examples for proper customization or how model updates affect accuracy. It also relies on data scientists who may lack business context, making it difficult to fully capture organizational logic.
Additionally, RAG consumes around 80% of AI infrastructure and tokens just for fine-tuning rather than actual use, raising ROI concerns. It also lacks built-in governance – there’s no way for compliance teams to validate training adequacy or ensure proper access controls.
illumex’s Generative Semantic Fabric (GSF) addresses these challenges through automated context building without consuming external AI tokens. It eliminates the need for specialized data scientists and provides complete transparency in mapping and reasoning through web, Slack, or Teams interfaces. GSF includes built-in governance and explainability, clear indicators of organizational coverage and data quality, and automated quality assessment for question-answering capabilities.
Many businesses struggle with making data-driven decisions despite investing heavily in data infrastructure. Why do you think this gap exists, and how does illumex address it?
The gap between data investment and effective decision-making continues to widen as data volumes explode, both internally and externally. Organizations now face not just their own growing data but also an array of external sources – from weather APIs to industry cloud platforms sharing healthcare data across European institutions, plus synthetic data for various use cases.
The challenge is that organizations still rely on humans for critical data tasks like modeling, quality assessment, and dashboard creation. Yet the scale and complexity of modern data environments make it increasingly impossible for human teams to effectively classify data, assess its quality, and ensure it’s suitable for AI-driven analytics and automation.
illumex bridges this gap by automating these traditionally manual processes, enabling organizations to effectively manage, validate, and utilize their expanding data landscape for meaningful business decisions.
What industries have been the quickest to adopt illumex’s platform, and what unique challenges or opportunities have you observed in these sectors?
We’re seeing the fastest adoption in industries that sit at the intersection of data intensity and heavy regulation, where companies need robust automation of data quality monitoring, usage tracking, and conflict detection. Financial services, pharmaceuticals, and retail/e-commerce are leading the charge, as these sectors aim to rapidly reinvent themselves using their existing data assets while navigating complex regulatory requirements.
With generative AI evolving rapidly, what advice would you give to enterprises looking to integrate AI effectively and responsibly?
Start by developing a clear strategic plan that identifies specific use cases and the business imperatives driving AI adoption. It’s crucial to avoid creating new silos of AI technology that operate in isolation from existing systems.
Instead, build a unified platform that integrates data management, analytics, and generative AI capabilities. Keeping AI initiatives disconnected from established governance practices not only creates significant risks but also leads to increased costs. The key is to create a shared infrastructure that supports all these functions while maintaining proper oversight.
With AI adoption accelerating, what trends do you see shaping the enterprise AI landscape over the next 3–5 years?
Two major trends are emerging in the AI landscape. First, agentic analytics is gaining momentum, allowing for more sophisticated data analysis and insights. Second, we’re seeing a shift toward agentic orchestration, which enables workflows based on collaboration between multiple AI models with diverse functionalities.
This orchestration moves us beyond single-purpose applications toward more comprehensive solutions. For example, in healthcare, instead of isolated applications for specific tasks, think about automation of the entire physician office workflows – combining image scanning, prescription processing, and drug recommendations in one seamless system.
These advancements rely on a robust generative semantic fabric to ensure accurate data access, shared context and coordination between AI agents. This foundation will be crucial for enabling both agentic analytics and orchestrated AI solutions to reach their full potential.
Thank you for the great interview, readers who wish to learn more should visit Illumex.
Credit: Source link
