Not all Graphs are Created Equal, Part 1
Moving beyond Property Graphs and Semantic Web for Contextual Automation
By Dave Duggal
November 2, 2026
Originally posted on Medium
Agentic AI Redefines Graph Requirements
Agentic AI has raised expectations for real-time, enterprise-grade contextual automation at scale with strong governance. This has moved the goal posts for semantics from pre-optimized queries and analytics on static domain knowledge and data to dynamically interpreting intent and optimizing decisions and actions based on living domain knowledge and data.
While bringing order to AI has generated tremendous interest in deterministic graph technologies, neither property graphs (e.g., Neo4J, Neptune and other Graph Databases) nor the Semantic Web standards (e.g., RDF, OWL, SPARQL, SHACL) were designed for today’s operational use cases. They are limited by their foundational design choices 25 years ago.
The distinction is not trivial. Contextual automation requires graph-based technology that supports data and behavior, entities and business objects, reasoning and execution, database transactions and complex event-driven distributed processes. These capabilities cannot be simply bolted on as an afterthought.
Defining Context
Context is background information (i.e., related, out-of-band knowledge and facts), humans, systems and agents need to interpret an event, request or prompt in order to optimize a response or action. A lack of context makes a query response irrelevant and makes a business process inflexible and impersonal leading to sub-optimal experiences and outcomes.
By this definition, context is information that is dynamically assembled, interaction-specific and ephemeral. It is just the relevant knowledge and facts, identified by relationships and policies, filtered by permissions to just the right human, system or agent, delivered just in time.
From a systems perspective, context is derived from a form of event-driven in-memory processing; a dynamic pipeline to translate intent into contextual responses. Conceptually, all data, information and knowledge, current and historical, contributes to 360-degree situational awareness. However, to be useful, it needs to be organized, interrelated and accessible so they can be efficiently retrieved, queried and reasoned over.
The Curse of Dimensionality
Contextual automation requires fine-grained knowledge of how business concepts, data, tasks, tools and rules relate to a specific event or request. However, reasoning across complex real-world business operations remains technically difficult and computationally “expensive”.
Business operations are typified by highly complex and ever-changing relationships between people, objects, services, environments, and infrastructure. It leads to an explosion of pointers, which makes modeling and processing the relationships difficult. In computer science this is called The Curse of Dimensionality.
It requires a solution that provides highly performant situational awareness across people and processes, systems and services, databases and devices with predictable low-latency, transaction guarantees and strong governance for event-driven, distributed processes at scale.
Consider that Semantic Web and Property Graphs have been around for twenty-five years; if they were the solution, then this problem wouldn’t exist. In the meantime, most organizations still struggle with developing a unified model, let alone operationalizing one for consistency and performance with strong governance at scale.
It’s Time for Something New
From the start, EnterpriseWeb has been motivated by these challenges. The company took a Computer Science first-principles approach and built its platform from the ground up to model complex real-world operations and enable real-time, enterprise-grade, contextual automation. It is the subject of EnterpriseWeb’s 23 granted US patents and the power behind its many award-winning industry projects.
EnterpriseWeb’s founders were inspired by Sir Tim Berners-Lee, James Hendler and Ora Lassila’s elusive vision of seamless agent interoperability, expressed in their famous 2001, Scientific American article on “The Semantic Web”. The company studied early technologies related to semantic web services (e.g., OWL-S, UDDI, and BPEL), which had strong ideas, but lacked practical implementations.
EnterpriseWeb took a different path. The company’s IP was developed from the ground up to support the demands of real-time, enterprise-grade contextual automation. This led to different design decisions.
EnterpriseWeb shifts Graph From Data Science to Computer Science.
The technology is rooted in Control Theory, Complex Adaptive Systems, Cybernetics, Systems Thinking, Metaprogramming, Concurrent and Parallel Processing Languages, Type Theory, etc. The company’s academic touchstones are Ashby, McCarthy, Bellman, Strachey, Backus, Milner, Beer, Dahl and Nygaard, Senge amongst others.
EnterpriseWeb is not a Property Graph or Semantic Web language, rather it is a no-code platform for contextual automation with its own language, “GOAL”(i.e., Graph Object Action Language ™). GOAL is a dynamically typed Lisp dialect with an abstraction that enables it to project hypergraphs from lists (i.e., list processing). It supports, as part of EnterpriseWeb’s systems and methods, the efficient modeling and processing of complex, n-ary, multi-dimensional and conditional relationships as abstract sets (i.e., Category Theory). EnterpriseWeb represents graph relationships mathematically and flattens graphs computationally for hyper-efficient global context.
Designed for knowledge-driven behavior.
GOAL is not just for graph domain modeling, queries and inferences, it was specifically designed for knowledge-driven behavior and autonomic, policy-controlled close loops. As with other Metaprogramming languages, GOAL allows for the declarative modeling of Code-as-Data. EnterpriseWeb leverages “Reflection” to generate data structures and behaviors at runtime based on interaction context.
GOAL is part of the runtime for EnterpriseWeb’s design, management and execution environments. It encodes and decodes all relationships in the background so Ontologists and Developers can focus on business domains and use cases.
GOAL efficiently enforces ontological entailments, business rules, workflow logic and governance policies. The runtime is event-driven, asynchronous and concurrent with massively parallel processing, designed for performance, scalability and resilience.
By bridging Knowledge Engineering and Systems Engineering, EnterpriseWeb has realized the holy grail — semantic web services.
EnterpriseWeb:
1. Bridges decidability and expressivity, description rules and production rules (constraints), data and behavior
2. Features a type-system that dynamically enforces both TBox (i.e., types) and ABox (i.e., instances)
3. Establishes metadata as a first-class citizen
4. Supports whole-part relationships and prototypal inheritance for flexible collections and evaluation of free variables
5. Projects ontological entailments, logical partitions, contextually realized objects, and multiple perspectives including namespaces
6. Enables declarative modeling of business logic, functions and workflows in the same graph with domain concepts, objects, relationships and constraints so they can be reasoned over together
7. Provides Massively Parallel Processing (MPP) enables hyper-efficient ad-hoc named queries, inferences and multi-factor constraint satisfaction problems without pre-optimization or database tuning
8. Performs Hybrid Analytic and Transaction Processing (i.e., HTAP) for declaratively modeled complex events, realized as a semantic service, non-locking, non-blocking with distributed transaction guarantees
9. Dynamically generates interfaces based on interaction context and permissions and late binds implementations
10. Executes ontology-driven automation and intent-based orchestration
11. Renders precise Context-as-a-Service to enrich and validate LLM inferences and optimize and control agents
12. Delivers one-to-many payloads per event, request or prompt
13. Supports branch-history (i.e., Change Data Control) based on immutable, log-style, append-only storage for replay, audit, troubleshooting, predictive analytics and learning
One Runtime for Description and Production Rules
In EnterpriseWeb’s GOAL, closed world Production Rules (e.g., business rules, workflow logic, governance policies) are processed first. The constraints scope the application domain, interaction context and security and identity. Constraints are defined in terms of conditional relationships, so Production Rules are flexible collections applied at any scope (namespaces, node types, node instances, properties, paths).
Once scoped, GOAL uses open world Description Rules (i.e., ontological entailments) to dynamically type the participating entities and relationships, to realize objects as a semantic service. It is achieved as a pipelined set of rules and transformations, a progressive evaluation, to deterministically materialize contextual meaning. Strong dynamic typing also supports deterministic mapping of properties via non-hierarchical Prototypal inheritance. Meaning is late bound — process scopes it, ontology controls it.
All processing takes place in one graph, with one runtime, which optimizes performance and ensures syntactic and semantic interoperability. The ontology remains open world through a separation of concerns. The domain model is logically partitioned from the application layer in the hypergraph. This duality ensures meaningful things, not just coincidental strings.
Metaprogramming and Category Theory
EnterpriseWeb systems and methods leverage Metaprogramming and advanced Functional Programming for generating precise Context-as-a-Service. They replace manual Context Engineering and primitive loops, with pure functions, that use recursion to reduce its hypergraph (i.e., highly dimensional, conditional, and sparse array) to a list. As a result, EnterpriseWeb can hyper-efficiently hydrate global context for stateless, ephemeral functions without linear serialization or over allocating resources.
One of EnterpriseWeb’s key differentiators is its support for Category Theory (i.e., abstract sets), which enables efficient higher-level reasoning across highly dimensional domains. In EnterpriseWeb this includes concepts, entities, objects, behaviors, namespaces, temporal history and system controls in the Knowledge Plane.
Conventional graph technologies, which are built on generally “flat” graphs (i.e., even if they support n-ary relationships, are limited to a single concept or data dimension), can only linearly serialize one object in one dimension at a time, which inhibits reasoning over complex domains, objects and transactions.
EnterpriseWeb uses higher-order logic to dynamically compose and contextually filter one or more sets, wrangled across dimensions, in a single operation (i.e., Flexible Collections). Flexible collections support complex queries, inference and constraint satisfaction algorithms with n-ary, multi-dimensional and conditional relationships.
Flexible Collections are an EnterpriseWeb Super Power!
EnterpriseWeb performs parallel, not serial processing, and it is based on prototypal inheritance not hierarchical class-based inheritance, so flexible collections can efficiently identify just the relevant properties, behaviors and constraints necessary to advance processing of an interaction. Instead of serializing discrete objects and then parsing them separately, EnterpriseWeb hydrates global context to realize a Semantic Web Service, which represents a contextual instantiation of a service.
Optimizing and Extending Graph Use Cases
By bridging knowledge, data and process, EnterpriseWeb optimizes existing graph use cases and extends them beyond retrieval, queries and inferences to ontology-driven constraint satisfaction, behavioral interfaces, workflow automation and intent-based orchestration using a single language.
Just as EnterpriseWeb’s serverless middleware capabilities eliminate the need for vertically-integrated application stacks, the platform does the same for the Semantic Web language stack (e.g., OWL, SPARQL, SHACL, Prolog/Datalog). EnterpriseWeb’s single runtime eliminates the stack related complexity, overhead and cost, while providing advanced new semantic capabilities,
Higher Order Queries, Inference and Constraint Satisfaction
Complex queries, inference and constraint satisfaction algorithms are modeled as declarative expressions (implemented as native JavaScript functions) that live in the graph as other typed objects. They can be registered in a managed system library (i.e., discover, invoke, log/trace, LCM).
EnterpriseWeb supports:
Ad-hoc Queries and Retrieval — efficient queries and traversals across dimensions, objects and sets (extends or replacesSPARQL and GQL).
Named Queries — custom, re-usable, named and parameterized queries across dimensions, objects and sets (extends or replaces SPARQL and GQL).
Named Inferences — inference policies across dimensions, concepts and objects (alternative to OWL ontological commitments and SHACL business logic).
Constraint Satisfaction — declarative goal-oriented algorithms across dimensions, concepts and objects (alternative to Prolog/Datalog and constraint-based language algorithms).
Advanced Data Transformation
EnterpriseWeb provides native operations for declaring simple, complex and temporal transformations over raw data/event streams (alternative to ETL data transforms and temporal DB queries). EnterpriseWeb advantages include:
- A single declarative ontology-driven language/tool for all data/event stream transformation that provides declaratively composable objects (no one-off, manual schema-based integration)
- Support for complex transformations over data that result in flexible, contextual collections (i.e., filtered and composed sets from a single operation, as opposed to multiple ETL steps)
- Common transforms (e.g., data slicing, filtering, grouping, enrichment, temporal windowing) implemented as fully declarative configurations
- Complex transformations are implemented as native JavaScript functions (managed libraries Functional-Declarative) modeled in the graph as declarative expressions
Rich Object Transformation
EnterpriseWeb provides native operations for automatically mapping raw data/event stream objects to domain concepts and types with optional persistence of instances (alternative to Custom Integration). EnterpriseWeb advantages:
- A single declarative ontology-driven language/tool for all workflows that provides declaratively composable objects and platform services and supports runtime queries and inferences (no one-off, manual integration)
- Replaces manual schema mapping exercises required in traditional ETL tools
- Enforces normalization and consistency of objects and types across discrete workflows
- Allows the modeling and reuse of rules across discrete workflows
- Supports persistence of object instances under full management (permissions, versioning/rollback, LCM) without the need for a separate DB or object store
Deploying EnterpriseWeb
EnterpriseWeb doesn’t use a Graph Database or Triple Store. EnterpriseWeb’s GOAL hyper-graph-driven metaprogramming language is a Lisp dialect; it naturally implements over long skinny tables (i.e., lists, key-value or columnar storage).
EnterpriseWeb deploys as a multi-model database (i.e., Key Value, Document, Object, Graph, Search, Time-Series) over a column-family database, eliminating the need for specialized databases and database administrators.
All queries, inferences and workflows are declaratively modeled and executed on-demand; no pre-optimized queries or database tuning is required.
EnterpriseWeb’s runtime handles all higher-level transaction controls and state management, as well as logging and tracing. It abstracts the persistence tier for developers, eliminating database administration concerns.
Column Family databases are one of the most basic, efficient and compressed forms of storage so they scale up naturally. Since EnterpriseWeb abstracts database management, the platform can also scale-out without sharding overhead or complexity.
Copyright 2026, EnterpriseWeb LLC. All Rights Reserved.
— — — — —
If you enjoyed this post, please follow me here on Medium or on LinkedIn.
Part 2 will continue the technical deep dive into EnterpriseWeb’s hypergraph implementation.
Related Links
Introducing the Semantic AI Harness
Sorry Telcos, graph databases are no way to run a business
Lowering the cost of context for real-time, data-driven systems