It's 2024, where few and far between omit AI from their core business operations or roadmap. It's never been easier to request anything from a Large Language Model (LLM), and get the heart’s desire in mere seconds - be it a travel plan to Japan or a summarized outlook on last year’s leading stocks in various markets. However, this process is far from being reliable. LLMs are known for their inaccuracies and hallucinations, moreover, one could hardly trust them with sensitive data.
Nevertheless, organizations can't pass on the most important technological innovation of our time, and forgo its advantages - increased productivity, democratized access to knowledge, and endless data monetization opportunities. Can organizations leverage LLMs reliably and safely on top of their data and enjoy the Generative AI advantages without its pitfalls?
There’s a sprint to have AI tools utilize all the data that flows into the enterprise, but using it "as-is" is far from feasible. Due to varied data stacks, different naming conventions, permission issues, data silos and the fact that structured data, which sits in databases, warehouses and business systems, is seldom semantically meaningful, AI cannot be implemented without preliminary preparation. This might result in unreliable and unusable answers, failed projects, high costs and unnecessary manual labor. The latest MIT Technology Review Insights survey, with data intelligence giant Databricks, found that 26% of surveyed C-Suite executives have inadequate data governance frameworks and siloed legacy data systems. It’s hampering most of them from scaling their AI use cases.
"Data truly is messy," said Inna Tokarev Sela, founder and CEO of illumex, an Israeli Generative AI startup whose vision is to make organizations become Generative AI-ready and enable them to have reliable interactions with AI and D&A tools. "Data has to become semantically meaningful before AI tools come into play. However, leaving that to your data team is laborious, time consuming and error-prone. Outsourcing this effort, as was done with semantic labeling for computer vision implementation, is impossible due to the sensitivity of the data. With illumex’s Generative Semantic Fabric, we automate this process, and turn the whole data landscape across different sources and apps into a semantic knowledge graph in days, all of that by only touching metadata. This is the first step in unifying the business data language and 'semantification' of customer data which allows quick AI implementation and its reliable querying."
Tokarev Sela explained to me that she started illumex based on her work with strategic enterprise clients at SAP and Sisense and witnessing their struggles in scaling their analytics and AI practice. "Our customers were managing data manually, just to make sense of it, and it’s just not feasible or scalable." Predicting the timely surge in Generative AI tools, Tokarev Sela indeed had the foresight to bring an automated solution for scalable, semantic data practice.
illumex has a number of features designed to automate and augment organizations’ data practice and governance, and allow them to become truly AI-ready and data-driven. Their Generative AI-generated data dictionary feature shows a comprehensive view of the customer's data landscape in the context of the business usage and workflows, with detailed lineage information about databases and their components, and documentation, sensitivity tags and ownership.
illumex not only shows usage statistics and other informative stats over data and analytics objects, but automatically translates their semantic meaning into business terms and metrics stored in the Active Business Glossary. These terms are the foundation stones of a common language between data producers and consumers. Their Active Business Glossary functionality connects the semantics of accurate organizational business definition with its precise data definition. The active feature ensures that the glossary remains updated with the ever-evolving data stack. It further allows a reliable data querying which fosters data-driven decision making processes.
"We also introduced illumex Omni late last year to democratize the data and analytics access to the business users," explained Tokarev Sela on their new offering to enable governed and contextualized LLM deployments in organizations. Omni is a smart assistant designed to help your LLM service output governed and more accurate responses, without the need for fine-tuning.
The startup is an alumni of the Intel Ignite accelerator program, as well as the ICON program. They received backing from a number of VCs, including Cardumen Capital, JIBE Ventures, toDay.Ventures and Israel Colorado Innovation Fund, as well as notable angels, such as Giora Yaron, Moshe Lichtman, and Ameet Patel.
Their customers come from data-intensive and highly regulated industries, like financial services, gaming, retail and pharmaceutical companies. They also integrate with any structured data sources, including Snowflake, Databricks, Tableau, and a myriad of on-premise sources.
Gartner sized the data management software market at $10.4 billion in 2022. Moreover, some research projections expect the sector to grow at a CAGR between 10% and 17% by 2030. These forecasts account for the magnitude of Generative AI, which will demand even more sophisticated data management solutions. Accenture recently pegged the economic value added of Generative AI at $10.3 trillion by 2038.
"With data becoming increasingly central to business operations, effective data-driven decision making is crucial. illumex’s generative semantic fabric is not just about better data management; it's about anchoring it in the business practice, and augmenting both with AI, LLM chatbots and other tools for a truly data-driven organization."
