The explorer, the parrot and the world map: why your CLM goes astray without ontology

At the beginning of the 19th century, Alexander von Humboldt crossed South America. He observes, measures, notes, collects... but he advances without a reliable map. The territories he explores, the Orinoco, the Andes, the Amazon, are still white areas. He has instruments, a notebook but no representation of the world to rely on. However, it has clear objectives, in the Upper Orinoco, for example, its objective is to confirm the presence of a natural channel between the Orinoco and the Amazon.

Today, “CLM” Projects are another form of expedition for Legal Departments. Let us dare to draw a parallel.

• The CLM, it is Humboldt the Explorer : organized, rigorous, able to follow an itinerary and document each stage. But limited by the absence of a card.
• The LLM, it is The parrot that Humboldt carries on his shoulder: he repeats, imitates, vocalizes... he has a lot of data because he has squared the ground, but he doesn't understand anything about it
• Ontology, it is The map of the world that does not yet exist : the structure of the territory: rivers, mountain ranges, outbuildings, passages... and pitfalls

Today, most AI-powered CLMs work like Humboldt before drawing his maps: a brilliant explorer, accompanied by a talkative parrot... but without a map.

‍

1.Why is your CLM disappointing where it should excel

Legal departments are investing heavily in motorized CLMs in LLMs, with the hope of automating the generation, review and execution of contracts. And yet, a paradox persists: these systems seem to be effective on confidentiality agreements, real estate leases or standard service contracts... but collapse as soon as the challenges become truly strategic.

The symptoms are well known: proposals for “exchangeable” clauses when they do not have the same legal scope, the inability to manage the fine dependencies between clauses within the same contract, events on contracts that are misinterpreted in terms of risks. This has the effect of creating frustration for lawyers who have to go back behind the machine.

In other words: where legal value is concentrated, most current CLMs fail.

‍

2. “Data models” don't make sense

The cause is neither the underlying LLM nor the prompt engineering. It is deeper: CLMs rely on data models that describe the class in which information belongs, but never its deeper meaning

CLM fields, forms, workflows, and tables indicate where information is stored, but never:

• wherefore a clause exists,
• which distinguishes a serious breach from a simple deviation,
• why a late notification must sometimes be reclassified as a variation order,
• how the clauses affect each other.

Deprived of this reasoning context, LLMs fill the gaps through statistical correlations.

The result: fanciful interpretations, especially in complex contracts where words have a very precise meaning, linked to the business sector, not a generic meaning and where the risks may themselves be specific. These systems come up against the uses considered to be the most promising for Legal Departments: where the burden and added value of legal teams are concentrated.

CLMs fail because they don't understand the business domain.

‍

3. Contract ontologies finally give AI systems a legal brain

The solution that is needed today in advanced AI architectures is clear: add an ontological layer, i.e. an explicit representation of legal and business reasoning.

Where a data model answers “where is the information stored?” , an ontology answers “what does this information mean, and why does it exist?” ”.

This symbolic layer allows:

• to anchor the reasoning of the LLM in legal reality,
• to avoid hallucinations,
• to manage edge cases,
• to correctly interpret post-signature events,
• to add new concepts without breaking the existing ones.

With an ontology, AI no longer recognizes only motives: it reasons.

‍

4. How ontologies transform a CLM into a reasoning engine

A contractual ontology is based on four fundamental building blocks:

• The entities: domain concepts (for example, in the audiovisual sector: works, rights, licenses, operating windows, ready to broadcast...).
• The attributes: qualifiers (exclusivity, territory, risk level...).
• The relationships: verbs that structure the real (creates, gives in, depends on, impacts...).
• The rules: legal mechanisms (a co-production involves the sharing of derivative rights; a modification of the deliverable requires an amendment...).

Together, they form a semantic contract which becomes the source of truth in the field of activity.

Structured experimentation showed it: with a simple data model, LLMs were systematically hallucinating about complex contracts. In contrast, with an ontology of 20 questions encoding definitions, constraints and rules, the same LLMs made correct pairings, including in borderline cases.

Beyond precision, ontology unlocks new abilities:

• semantic flexibility (adding new concepts without cumbersome reparameterization),
• research based on meaning,
• logical constraints automatically applying legal rules,
• systemic coherence between clauses, obligations and events.

It's a change in nature, not in degree: systems are really becoming intelligent.

‍

5. Waldenlab orchestrates the ontological transformation of your CLM

For organizations that deploy or renew a CLM, the challenge is no longer just to configure a tool (data model, approval circuits, authorizations, reporting, etc.) it's about architecting an AI system capable of reasoning.

Waldenlab intervenes precisely at this intersection between functional consulting, AI architecture and knowledge engineering. We support your teams so that their tools fully inhabit their world:

• frame the scope and priority use cases,
• animate legal workshops to identify concepts, rules and dependencies,
• structure the first version of the contractual ontology,
• guide translation in Knowledge Graph,
• orchestrate the integration between CLM, LLM, and ontology layer.

Are you in the process of acquiring or strengthening a CLM? Waldenlab can manage the architecture of your project to transform your CLM into a real engine of legal reasoning and manage ALL your contracts, even the most complex ones.

Humboldt ended up producing what was missing: a coherent map of the world, connecting phenomena, landforms, climates, ecosystems. He wrote on the map that a natural canal linked the Orinoco to the Amazon, in the same way that two clauses of a contract can be dependent without being obvious. This map was not a simple drawing: it was a structure for understanding.

This is exactly what a contractual ontology does today.