From data to knowledge: a ladder almost nobody finishes climbing
To understand why data on its own is not worth as much as we think, it helps to revisit a classic model of knowledge management: the DIKW hierarchy (data, information, knowledge and wisdom). It is a four-rung ladder in which each level adds value to the one before.
Data are raw facts: a sales figure, a sensor reading, an email from a customer. On their own they say nothing. Information appears when that data is processed and organised with a purpose: March sales fell 12 % compared with February. Knowledge emerges when context and experience are added: sales fall every March because our main customer closes its financial year and freezes orders. And wisdom is knowing what to do about it: bring the sales campaigns forward to January.
Most companies are settled on the first two rungs. They accumulate data in scattered systems and generate reports, but knowledge (the context, the reasons why, the accumulated experience) still lives in people's heads, in email threads and in documents nobody can find. That is exactly the gap that structured knowledge comes to fill.
Why accumulating data no longer sets anyone apart
If data were the competitive advantage, the success figures for enterprise AI would be very different. According to a Gartner survey published in 2025, 63 % of organisations do not have (or do not know whether they have) the right data management practices for AI, and the consultancy predicts that through 2026, 60 % of AI projects unsupported by AI-ready data will be abandoned. Along the same lines, the MIT study "The GenAI Divide: State of AI in Business 2025" concluded that 95 % of generative AI pilots in companies generate no measurable impact on the profit and loss account. The problem, the study itself notes, is not the models: it is integration with each organisation's knowledge and processes.
There is a second compelling reason. Gartner estimates that between 80 % and 90 % of the data a company generates is unstructured: documents, emails, presentations, minutes, contracts, conversations. In other words, most of an organisation's knowledge is in formats that traditional databases cannot interpret. Having a lot of unstructured data is now the norm, not the exception. What distinguishes some companies from others is their ability to transform that scattered material into something their systems (and their people) can exploit.
Added to this is a silent cost that almost nobody measures. A McKinsey study had already estimated that professionals spend around 19 % of their working day searching for information they need to do their job, and later studies by Forrester and Atlassian suggest the figure has worsened. Every hour spent looking for a document is an hour in which the knowledge exists but is not available.
What structured knowledge is (and what it is not)
Structured knowledge is an organisation's knowledge converted into an organised, connected and queryable format: information with context, with explicit relationships between concepts and with an attributable source, so that both people and AI systems can find it, understand it and apply it. It is not a document repository or a data lake. It is the difference between having a thousand contracts in a folder and being able to ask which penalty clauses we have signed with suppliers in the energy sector and get a correct answer with its sources.
It is worth clearing up a common misunderstanding: structured knowledge does not just mean structured data. An Excel table contains structured data, but it may lack context. And a free-text document is unstructured data that may contain extremely valuable knowledge. Structuring knowledge consists of extracting what is valuable in both worlds and organising it with three properties: context (what each thing means and in what circumstances it applies), relationships (how it connects with customers, projects, standards or processes) and traceability (where each statement comes from and when it was updated).
At LurNova we insist on that third property because it is what separates a useful AI from a risky one. A system that answers without being able to point to its source forces you to verify everything, and that cancels out much of its value. Our premise is that artificial intelligence, if it is not secure and traceable, is not intelligent.
RAG and knowledge graphs: the technology that makes it possible
Two technologies have established themselves as the practical way to exploit structured knowledge with AI. The first is RAG (retrieval-augmented generation). Instead of relying on what the language model memorised during training, a RAG system first searches the company's knowledge base for the fragments relevant to the question and generates the answer from them, citing the sources. The result is an AI that answers with your organisation's knowledge, not with generalities from the internet, and that can be audited.
The second is knowledge graphs. A knowledge graph represents the company's reality as a network of entities (customers, products, projects, standards, people) connected by explicit relationships: this customer belongs to this sector, this project applied this solution, this clause responds to this regulation. Compared with a classic document search, the graph makes it possible to answer questions that require chaining several reasoning steps, such as which projects we have done for customers affected by the EU AI Act and what we learned from them.
The combination of the two, known as GraphRAG, is one of the clearest trends in enterprise AI today, precisely because it brings together the best of both approaches: retrieving specific fragments and understanding the relationships between them. At LurNova we apply RAG technology in our AI agents and in products such as MIA, because we have seen that it is what makes it possible to move from an assistant that sounds convincing to an assistant that gets it right and can prove it. And when that knowledge is deployed on private models and the company's own infrastructure, the company also retains sovereignty over its most valuable asset.
Corporate memory: the knowledge that walks out of the door
There is an angle of structured knowledge that usually goes unnoticed until it hurts: corporate memory. Every person who leaves an organisation takes knowledge with them that is almost never documented: how processes really work, the particularities of certain customers, the mistakes of past projects and why certain decisions were taken. That tacit knowledge rarely appears in a manual, and its loss forces companies to repeat mistakes and relearn what they already knew.
Structured knowledge is the systematic answer to that problem. When the decisions, lessons and context of projects are captured in an organised and connected way, the company's memory no longer depends on who is sitting in each chair. And here AI changes the rules of the game: documenting was always costly and so it was not done, but today's systems can extract knowledge from minutes, emails and working documents with a fraction of the effort it took to do it by hand. The organisation that structures its knowledge does not only respond better today: it builds up an asset that gains value with every project, something no competitor can copy by buying the same tool.
That is the fundamental difference from the race for data. Data can be bought, AI models are increasingly accessible to everyone, but an organisation's structured knowledge (its relationships, its experience, its reasons why) is unique and non-transferable. That is why it is the real competitive advantage.
How to start structuring your company's knowledge
You do not need a gigantic project to get started. What we have seen with our clients is that an incremental approach guided by specific use cases works best.
- Identify a knowledge domain with direct impact: customer support, commercial proposals, technical documentation or regulatory compliance are usually good starting points.
- Audit where that knowledge lives today: what is documented, what is scattered across emails and folders, and what exists only in people's heads.
- Organise and clean up the sources: remove obsolete versions, unify criteria and define who is responsible for keeping each block of knowledge up to date.
- Deploy a RAG system on that domain, with source traceability, and measure its accuracy with real questions before opening it to the whole organisation.
- Evolve towards the graph when the case calls for it: if your team's questions require connecting entities (customers, standards, projects), it is time to structure relationships, not just documents.
- Govern the asset: security, access permissions, auditing and periodic review. Structured knowledge only keeps its value if it is looked after like any other critical asset.
An important nuance: order matters. Deploying an AI on disorganised knowledge tends to produce disorganised answers that look credible, which is the worst possible scenario. Structure first, automate later.
Conclusion: the asset nobody can copy
Data has become an abundant raw material, and abundance was never a competitive advantage. What sets an organisation apart today is its ability to turn that data into structured knowledge: organised, connected, traceable and exploitable by AI systems. The figures from Gartner and MIT point in the same direction: AI projects do not fail because of the models, but because of the lack of a prepared knowledge base.
At LurNova we support companies along that entire path: from initial consultancy and organising knowledge to deploying AI agents with RAG technology on private models, with security, traceability and technological sovereignty as starting requirements, not as add-ons. If you want to know where your organisation's knowledge stands and how to turn it into your competitive advantage, let's talk.
CEO of LurNova, with extensive experience in technology and digital transformation. Involved in developing technology projects and applying artificial intelligence in organisations, he combines technical knowledge with business vision.
