Architecting for AI at enterprise scale

What modern data architects must prepare for

Forrester outlines three core challenges shaping the next generation of data architecture:

• Support hybrid and multi-cloud environments. Interoperability is now a baseline requirement. Hybrid and multicloud strategies enable flexibility, regulatory alignment, and performance optimization across hyperscaler ecosystems like AWS, Azure, and Google Cloud.
• Embrace AI innovation while keeping data at the core. Generative and agentic AI demand scalable, connected, and trusted data foundations that can deliver accurate context across the enterprise.
• Rearchitect for AI readiness. To handle real-time, high-volume workloads, enterprises must adopt architectures such as data fabric, data mesh, and vector databases that unify access and enable intelligent automation.

As Forrester Research notes,

“a modern, scalable data architecture is essential to realize data’s full potential… With the right foundation, organizations can streamline operations, unlock new innovations, and empower users to interact with AI through intuitive, natural language tools.”

The Forrester six-stage model

Forrester’s six-stage model provides a blueprint for designing, deploying, and operationalizing data architectures that can support advanced AI use cases.

1. Ingest / Stream

The ingest layer captures data from a wide range of sources: IoT sensors, applications, transactional systems, and streaming platforms.

To enable real-time analytics and AI, this layer must deliver low-latency, high-throughput ingestion with built-in fault tolerance.

Automation plays a critical role: event-driven architectures and AI-powered pipelines reduce manual intervention and ensure continuous, reliable data flow. As Forrester notes, “to unlock AI and real-time analytics, data must enter the platform with minimal latency and intact integrity.”

2. Prepare / Transform

At this stage, raw data is cleansed, enriched, and integrated into a model-ready state.

AI-driven transformation pipelines handle validation, normalization, and feature engineering at scale.

Forrester Research highlights the role of knowledge graphs, which “create rich semantic links between entities, enhancing context and insights.” Embedding these relationships improves model accuracy, supports personalization, and simplifies data discovery.

Organizations that invest in intelligent transformation accelerate time to insight and reduce the human overhead associated with manual data prep, critical for maintaining agility in fast-evolving AI environments.

3. Define / Model

Here, teams shape data into meaningful structures, applying semantics, ontologies, and metadata management to create a common language across systems.

This layer is also where AI models are developed, trained, and governed.

“A well-architected platform supports seamless collaboration, enabling scalable, repeatable model development, and governance across the AI lifecycle,” Forrester Research writes.

By standardizing metadata, lineage, and quality metrics, organizations establish traceability and explainability, essential for both compliance and trust in AI outcomes.

4. Store / Persist

A modern data platform must store structured, semi-structured, and unstructured data securely, and make it accessible across cloud, on-premises, and hybrid environments.

Forrester emphasizes cloud-native storage as the most scalable and flexible option, offering elastic capacity, built-in redundancy, and automation for replication and scaling.

The goal is to optimize for both batch and real-time access, balancing performance and cost while enabling continuous model retraining and low-latency inference. Continuous monitoring within this layer enforces governance and privacy compliance automatically.

5. Integrate / Orchestrate

Integration and orchestration connect data, models, and business systems into unified, intelligent workflows.

Modern architectures achieve this through decentralized data management, reducing bottlenecks while maintaining enterprise-wide oversight.

Automation again is key: orchestration tools coordinate data processing, model training, and deployment across hybrid and multi-cloud environments. By standardizing APIs and connectors, enterprises can scale AI pipelines efficiently while ensuring consistency, reliability, and governance end to end.

6. Deliver / Share

The final stage operationalizes insights, ensuring data products, predictions, and AI-driven outcomes reach business users securely and in real time.

Personalized dashboards, APIs, and embedded analytics bring decision intelligence into every workflow.

Continuous feedback loops refine AI performance, while low-latency infrastructure sustains adaptive, context-aware systems.

As Forrester Research highlights, delivery isn’t just about access: it’s about closing the loop between data, model, and decision to drive measurable business outcomes.

From infrastructure to intelligence

The model outlined by Forrester (you can read a full description of the model in the report) positions architecture as the enabler of enterprise intelligence, connecting people, process, and technology in one adaptive system.

When each stage is built with scalability, interoperability, and governance in mind, AI can move confidently from experimentation to production, delivering insights that are trusted, explainable, and actionable

Why a Universal Semantic Layer is essential

Here at Strategy, we believe a Universal Semantic Layer is the most effective way to unify fragmented enterprise data and future-proof your analytics and AI strategy. Enterprises today operate across multiple data warehouses, clouds, and BI tools, creating inconsistent definitions, duplicated logic, and governance gaps that undermine trust in analytics.

A Universal Semantic Layer solves this by decoupling business definitions and security from the underlying data sources. It establishes a single, governed layer where metrics, hierarchies, and relationships are defined once and applied everywhere, ensuring that “revenue” means the same thing across every dashboard, query, or AI model.

This shared semantic foundation powers agility, not complexity. It delivers consistent performance across multicloud environments, centralizes security and governance, and enables cross-platform analytics without data duplication. It also provides semantically rich, AI-ready data that accelerates modeling, reduces hallucination risk, and ensures business-specific context for large language models.

By implementing a Universal Semantic Layer, organizations can unify governance, accelerate insights, and achieve true architectural independence—turning their data ecosystem into a trusted, scalable foundation for AI-driven decision-making.

Access the Forrester Research report, The Forrester Data, AI, and Analytics Architecture Model, courtesy of Strategy, to learn how to build a platform designed for scalability, trust, and innovation.