Information Architecture

We build integrated information architectures and knowledge graphs to enhance data and AI, improving efficiency and business knowledge.

Industries:

Finance & Insurance - Retail & FMCG - Energy & Utility - Life Science - Transportation - Industrial

Solutions:

AI - Data Governance

Technologies:

Blindata - Open Data Mesh

Background

Increase data usability to achieve greater value

Data becomes a liquid asset only when it is easily usable — and, even better, reusable — quickly generating value for the organization and maximizing the return on investments. Therefore, within Data Governance processes, it is essential to build an information architecture that supports the search and querying of the corporate data base.

In fact, to be truly reusable, data must be understandable — that is, enriched with context and domain semantics — so that they can be effectively interpreted and leveraged by users who are not familiar with the technical aspects of the systems.

The synergy between data and AI

Today, with the spread of generative artificial intelligence, having a shared knowledge base and a high-quality information architecture is essential. This enables the integrated representation and querying of structured data (e.g., tabular data following an E-R model), semi-structured data (nested JSON/YAML data), and unstructured data (documents, images, audio recordings).

AI can enhance efficiency, productivity, and user satisfaction only when data is complete and reliable. It can also assist humans, to some extent, in building and maintaining the information architecture itself.

AI uses the enterprise information stack to provide relevant answers and, in turn, helps build the architecture. [img.1]

Common problems without an information architecture

Without an integrated information architecture, organizations often struggle to:

→ Locate data assets
When responding to a business need or exploring the data base for new opportunities, it can be challenging to identify the relevant data assets and understand how to access them.
→ Interpret the meaning of data assets
Many datasets become difficult to use and maintain over time because their content and context are hard to understand — especially in the case of legacy data, whose ownership and knowledge have gradually been lost.
→ Share a common domain semantics
In organizations with complex business domains, different units often develop divergent knowledge and terminology. Without shared semantic elements, there is a high risk of creating data silos.
→ Align business language and data terminology
To scale data management and accelerate delivery, modern operating models aim to create cross-functional teams aligned with business domains. However, forming these groups is often challenging, as business and IT have traditionally worked in separate silos, with distinct processes, terminology, and objectives.
→ Difficulty reusing data in a versatile way
Even with modular architectures and a data-as-a-product approach, without a shared and explicitly defined semantics, it is not easy to leverage existing data assets in a composable and flexible way across different use cases.
→ Difficulty extracting relevant informational content from unstructured data
An organization produces content every day in the form of procedures, documentation, meeting minutes, or recordings. If this knowledge base is not integrated with the structured data generated by systems around a common semantic model, it becomes difficult to extract valuable insights from these contents to enhance productivity.

To unlock the value of data and make it truly useful to the organization, it is essential to be able to answer fundamental questions such as:
• What are the main data assets available to the company?
• What is the meaning of a dataset, and how can it be correctly interpreted?
• What are the most important domain concepts, and what relationships exist among them?
• Which terms help business users better understand the data?
• How can one determine whether it makes sense to combine existing data assets for new use cases instead of developing ad-hoc solutions?
• How can unstructured content be linked to related structured data and to the key domain concepts?

An integrated information architecture therefore enables organizations to maximize the value of their data by promoting consistency, reuse, and shared understanding across the enterprise.

Challenges

Building an integrated enterprise-level information architecture is based on a few essential pillars:

Domain knowledge as a differentiating asset

Domain knowledge is a unique asset of the organization, built from the learning and experience of its people. It is what differentiates the company from its competitors and is internalized, shared, and applied by individuals — it cannot be replaced or regenerated solely through artificial intelligence.

Linking domain semantics to data assets

Formalizing domain knowledge is not an end in itself — it serves to maximize business value by promoting the reuse of informational assets. It is essential to maintain a strong link between domain concepts and the data that represents them, fostering synergy and continuous communication among all stakeholders, both IT and business.

Interoperability between structured and unstructured data

Representing unstructured data so that it can complement structured data and be linked to the same conceptual domain model expands the information assets and allows the organization to extend its domain knowledge — leveraging what people have already produced, even if in an unorganized form.

A strategic and value-driven approach

Building an information architecture requires significant investment. Therefore, an integrated data & AI strategy is needed to direct efforts toward priority use cases, following an incremental and value-driven approach.

Solutions

What is an information architecture

An information architecture is built by developing several layers on top of raw data, enhancing their understanding and making them actionable — that is, usable for decision-making, execution, and value creation within the organization. Each layer adds contextual elements, which significantly increase the ability to use data assets both individually and in combination.

Building layers of information, knowledge, and intelligence improves the usability of data. [img. 2]

Defining a solid metadata foundation makes it possible to transform data into understandable information, at least at a basic level.
Modeling domain semantics — including the relationships among relevant concepts — further enhances the comprehensibility of information and, above all, the versatility with which it can be used.
The addition of algorithms provides the intelligence needed to translate the insights derived from information enriched with domain knowledge into business decisions or actions.

What is the Enterprise Knowledge Graph

An elegant way to represent the information architecture is through the construction of a virtual Enterprise Knowledge Graph, which links the domain’s semantic knowledge elements to the data available on the platform and to the metadata needed to enrich their context.

The information architecture in the form of a Knowledge Graph provides a shared context for multiple AI agents. [img. 3]

The following are the elements that make up the knowledge graph.

Enterprise Ontology

Enterprise Ontology: a conceptual model represented in graph form and interpretable by both humans and applications. It expresses domain knowledge in terms of:
• Concepts: the entities relevant to business processes (e.g., Customer, Product, Order)
• Attributes: the relevant properties of the concepts (e.g., Customer Status, Product Description, Order Date)
• Relationships: the semantic links between concepts (or attributes), expressed through predicates (e.g., Product is purchased by Customer).

The ontology is a more expressive form of conceptual model than the traditional Business Glossary, which is widely known in the field of Data Governance.

An example of business ontology [img. 4]

Data Product

Data Product: these are the units of modularization, ownership, and deployment within the data architecture, designed to expose one or more domain data assets in a way that is easily consumable by users, applications, or other data products.
Each data product is managed under the responsibility of a Data Product Owner, has a defined scope, delivers reproducible value to its consumers over time, features an independent lifecycle, and provides explicit interfaces governed by data contracts.
For more details on the anatomy of a Data Product, see the Data Mesh page.
The data products included in the architecture are registered in the Data Product Catalog, and access to them is managed through the Data Product Marketplace.
A data product exposes:
→ Physical data assets, both structured and unstructured, through output ports
→ Metadata, which provide context to the data and effectively transform them into usable information, through other types of ports

Physical Data Assets and Semantic Linking

The physical data assets exposed by data products can take various forms — relational tables and views, topics on an event-streaming platform, folders in a storage bucket, or shared content repositories.
Through metadata, a semantic link (also known as vertical lineage in traditional Data Governance) is established between data assets and the concepts defined in the enterprise ontology. Similarly, this link can be traced down to the individual physical field within a dataset, indicating which conceptual attribute the data in that field refers to. This mapping connects the physical layer with the domain model, enabling disambiguation of the relationships among assets.
A common example demonstrating the value of semantic linking is when a single table contains two foreign keys that both reference the same target table but represent different relationships. Without a semantic link, it is not immediately clear how to correctly interpret and use those keys in join operations.
Another practical example is linking both structured data tables and instances of data extracted from unstructured content (via AI applications) to the same ontology concepts and attributes. This makes it possible to analyze a domain concept — for instance, contracts — by combining tabular data generated by systems with related documents stored in a content management system.
The physical data assets, along with their metadata, are registered in the Data Catalog, which is one of the key tools used in Data Governance.
In this way, the virtual Enterprise Knowledge Graph implements the data, information, and knowledge layers of the information architecture, establishing the metadata management foundation that has long been one of the cornerstones of Data Governance.

The Enterprise Knowledge Graph as Context for AI Agents

The Knowledge Graph serves as a common context that can be accessed through retrieval processes by different artificial intelligence agents to enhance search, data querying, and user support in insight extraction.

Moreover, these agents can contribute to extending the underlying information architecture through learning processes, by persisting into the Knowledge Graph new elements of knowledge, information, and data acquired from conversational interactions with domain experts or derived from the processing of unstructured content already available within the organization.

The complete process

Quantyca supports clients throughout the journey required to successfully establish an information architecture, following these key phases:

1 AS-IS Assessment

Mapping the current state in terms of data culture adoption, maturity of existing metadata management processes, and tools in use.

2 Strategy Definition

Developing an integrated Data & AI strategy aligned with business needs.

3 Operating Model Implementation

Launching programs to develop the information architecture based on identified use cases.

4 Platform Implementation

Designing and deploying the technological platform required to build the information architecture

5 Pilot Projects

Developing data products (including metadata management) and supporting business stakeholders in modeling the enterprise ontology.

6 Enabling

Promoting principles, best practices, and both methodological and technological skills across corporate teams to enable large-scale expansion of the information architecture.

Benefits

Building an integrated information architecture in the form of a virtual Knowledge Graph enables organizations to:

Promote data reuse

A modular design based on Data Products, combined with semantic linking to the enterprise ontology, enhances the ability to compose Data Products flexibly to serve multiple use cases and reduce costs for the organization over the medium to long term.

Share domain semantics

Formalizing corporate knowledge helps identify domain concepts with shared meaning across the organization, reducing ambiguity and friction in data sharing among different business units.

Bridge the gap between business and IT

Collaboration among cross-functional teams in developing the information architecture, along with a focus on semantics, brings business language closer to IT terminology — improving operational efficiency and fostering the spread of a shared data culture within the organization.

Use Case

Data & AI Strategy