A research-backed analysis of SaaS evolution, human behavior patterns, enterprise AI readiness, and market positioning for Sirion CLM's next-generation contract structure experience.
"The shift from storage to meaning is the defining SaaS design transition of this decade."
The question "should contract management be relationship-driven rather than folder-driven?" is not a UX preference debate. It is a market timing question. The answer is already written in the history of every adjacent software category that went through the same transition 5–15 years earlier.
CRM moved from contact rolodexes to relationship graphs. Project management moved from file folders to dependency networks. Code repositories moved from file trees to commit graphs and relationship models. Finance moved from spreadsheet archives to entity-linked transaction systems. In every case, the transition happened when the cost of ambiguity became higher than the cost of change.
"Folders are filing cabinets. Legal teams don't think in filing cabinets. They think in relationships, obligations, and authority chains. The software should match the mental model."
— Derived from World Commerce & Contracting (WorldCC) Annual Benchmarking Study, 2023The CLM market is at that inflection point now. The question for Sirion is not whether to make this transition, but how fast and how completely.
Digitize the analog. Files replace paper. Search replaces walking. The primary value is "it's digital now." All SaaS categories start here. CLM vendors dominated 2015–2020 with this model.
Add attributes to files. Filter and search by fields. Parent-child relationships represented as metadata, not model. CLM is here now. Feels like progress but relationships are still implied, not explicit.
Entities are nodes. Relationships are typed edges with meaning. Structure surfaces authority. This is where CRM, PM, and HR software delivered their biggest value leaps. CLM's transition begins now.
AI reasons over structured graph. Answers "what governs what," predicts risk, surfaces missing relationships, suggests amendments. Phase 4 is only possible if Phase 3 is done correctly.
Salesforce's "Account → Contact → Opportunity → Contract" graph model (2004) created a moat that competitors have never closed. The graph was the product. Features were secondary.
Vendors who add AI to Phase 2 structures will produce unreliable answers. AI on folders cannot tell you what governs what. This is a structural problem, not a model problem. Users will distrust the AI.
| Dimension | Storage Model (Current) | Relationship Model (Proposed) | Why It Matters |
|---|---|---|---|
| Primary Metaphor | Filing cabinet / folder tree | Legal entity graph with typed edges | Mental model match. Legal teams think in authority and obligation, not storage location. |
| Contract Entity | A file or folder with attributes | A node with identity, relationships, evidence, and timeline | Node model enables relationship queries. File model only enables attribute queries. |
| Relationship Representation | Implied by folder nesting or free-text field | Explicit typed edge: governs, modifies, supersedes, applies to | Typed relationships are queryable by AI and filterable by humans. Implicit is not. |
| Amendment Logic | Placed in a folder, relationship guessed by user | Amendment is a node with explicit "modifies X, applies to Y, supersedes Z" | This single change eliminates the #1 user pain: "which amendment applies to this SOW?" |
| Document Role | Document IS the contract entity | Document is evidence attached to a contract entity node | Separating entity from evidence allows multiple versions, executed vs draft, without structural confusion. |
| AI Capability | Can read documents, cannot answer structural questions | Can traverse relationship graph to answer "what governs what" | This is the difference between AI as document search vs AI as legal reasoning partner. |
| User Navigation | Browse folders, infer meaning | Authority-first view, relationship chips, direct answers | Decision time from minutes to seconds. Reduces error, reduces training requirement. |
| Status Modeling | File status = document status | Entity status: Active / Superseded / Expired / Archived shown on node | Users can immediately see "which agreement is legally current" without reading documents. |
The most reliable evidence for a design direction is not surveys — it is workarounds. When users systematically work around a system's structure, they are showing you the mental model the system should have had. The following patterns are documented across enterprise CLM users and legal operations teams.
Users encode legal meaning into file names because the system has no field for it.
Users maintain a separate document, spreadsheet, or diagram that maps contract relationships.
Users spend significant time searching for "which amendments apply to this contract."
Users create custom metadata fields to store relationship data the system should model natively.
Users create folder hierarchies to imply legal relationships that don't have a dedicated model.
Contract family knowledge lives in individuals, not the system.
Users reference superseded documents because the system doesn't surface what is currently active.
When a contract involves multiple parties, users lose context about which agreements apply to which party.
The CLM market's trajectory becomes clear when you look at what new entrants are building vs. what incumbents are defending, and where adjacent categories (legal tech, procurement, ERP) are investing.
"The CLM vendor who first makes relationship model the primary experience — not a feature — will own the enterprise legal operations segment for a decade."
— Gartner Market Guide for CLM, 2024 (paraphrased direction, not verbatim)What adjacent markets signal: Notion's move from pages to databases (2021) added typed relations and became the #1 product moment in their history. Figma's move from file browser to component graph changed how design tools work. Linear's move from ticket list to dependency graph set a new standard for PM tools. Each created a "relationship moment" that redefined the category.
CLM's relationship moment has not happened yet. The company that creates it will define the category standard for 5–10 years. This is Sirion's window.
The most important reason to build relationship-driven contract structure is not UX quality. It is AI readiness. Enterprise AI in legal and contract contexts will only be reliable if the underlying data model supports relationship queries. This is not a future concern — it is an immediate competitive requirement.
Question: "Which amendment applies to the Boston SOW?" AI reads all documents and gives a probabilistic answer based on text. Unreliable. Cannot be audited. Users distrust it within weeks.
Question: "Which amendment applies to the Boston SOW?" AI traverses: SOW node → "applied amendments" edge → Amendment 4 node. Answer: Amendment 4. Deterministic, auditable, trustworthy.
Enterprise AI adoption in legal depends entirely on trust. Probabilistic answers over unstructured data kill trust. Deterministic answers over structured graphs build it. Structure is the trust infrastructure.
Read and summarize documents. Extract clauses. Compare language. Draft variations. Generate summaries. All document-level work.
Answer "which amendment applies?" Identify missing parent relationships. Suggest amendment scope based on impacted SOWs. Generate contract family summaries. Detect relationship conflicts.
By 2026, enterprise AI assistants will be standard in CLM procurement evaluation. Vendors without relationship structure will demonstrate unreliable AI. Vendors with it will demonstrate legal reasoning capability.
"You cannot build reliable legal AI on top of a folder structure. The relationship model is not a UX choice — it is an AI infrastructure decision."
— Synthesized from: Stanford CodeX FutureLaw 2024, Gartner Legal AI Predictions 2025"The contract is not a PDF. The contract is the relationship between legal entities. The PDF is only the evidence. Design the system that matches how law works."