Many AI initiatives don’t stall because of the models – they stall because of the data behind them.

Teams can build promising prototypes, but moving those models into production is where progress slows down. Data is spread across systems, pipelines are difficult to maintain and governance often lags behind how data is actually used. What starts as a technical challenge quickly becomes an operational one.

This is the pattern many organizations are encountering as they move from experimentation to production AI: not a lack of innovation, but a gap in how data is managed, trusted and operationalized.

If AI is the goal, the real work starts with the data foundation.

The real constraint isn’t AI – it’s data

Most organizations are not short on data – they’re short on data they can actually use with confidence.

Data is distributed across clouds, on‑prem systems and SaaS applications. Pipelines are often built for specifics, making them difficult to scale or reuse. Governance is often applied after the fact rather than embedded into workflows.

These challenges show up across the entire lifecycle – from preparing data to building models to deploying them in production. Research from the Data and AI Impact report found that fragmented data environments and insufficient governance remain among the top barriers to AI progress. Simply put, AI initiatives fail not because of models, but because of the data foundation underneath them.

From data management to data intelligence

The term “data intelligence” has gained traction in recent years, but its meaning matters. It represents a clear shift beyond both traditional data management and modern data platforms.

Historically, data management has focused on infrastructure: moving, storing, and preparing data. Modern data platforms, such as lakehouses, have extended this by solving for storage and compute at scale. These capabilities are necessary – but not sufficient.

Data intelligence is about what organizations can actually do with their data – reliably, at scale and in production. It connects data access and preparation, governance and trust and analytics and AI execution into a single, continuous system.

In practice, that means turning raw data into trusted, reusable assets. Operationalizing analytics and AI on top of that data with confidence is also important. Without that connection, organizations can store data and build models, but still struggle to deploy them in real-world environments.

Put simply: data management makes data reliable, while data intelligence makes it usable and valuable in real-world decisions.

The path to achieving data intelligence

Traditional data management was built for a different era – batch processing, structured data and downstream governance. That model breaks down in an AI-driven environment.

To move forward, organizations are shifting toward a more integrated approach – one that addresses three persistent challenges: fragmentation, low trust and performance constraints.

In practice, that shift comes down to three priorities:

• Modernize your data for AI – Make data accessible, consistent and ready for use across environments.
• Operationalize trusted data and AI at scale – Embed governance, lineage and policy into how data and models are built and deployed.
• Reduce friction in the data ecosystem – Improve performance without increasing duplication or complexity.

Together, they define a repeatable path from fragmented data environments to achieving trusted and scalable AI.

But defining the path is only part of the challenge.

The data intelligence layer

Most modern platforms focus on storing and processing data.

But the harder problem sits above that layer: making data usable, trusted, and governable for analytics and AI.

That’s where many organizations are now focusing – on what’s often described as a data intelligence layer.

Rather than introducing another system this layer connects how data is accessed, governed and used – helping teams answer fundamental questions:

• Can I trust this data?
• Where did it come from?
• What policy applies?
• Can I reuse it for this model or decision?

Organizations that answer these questions effectively are better positioned to move beyond experimentation to AI that operates reliably in production.

A roadmap built for the AI era

As AI raises the stakes, the core challenge becomes clear: How do you accelerate AI without losing control?

At SAS Innovate 2026, one theme stood out clearly – data engineering, govenance and execution can no longer operate as separate layers. They need to function as a unified system.

That separation creates friction:

• Pipelines are duplicated.
• Governance is inconsistent.
• Performance becomes unpredictable.

SAS’ data management roadmap reflects this shift, bringing together three interconnected areas: data engineering, data governance and data acceleration.

You can read more about that in the SAS Communities article What’s New with Data Management in SAS Viya: Copilots, Data Acceleration, and What it Means for You.

1. Data engineering: From fragmented pipelines to shared workflows

Data engineering is where most data initatives begin, but in many organizations, it remains fragmented across tools, languages and teams.

This fragmentation leads to duplicated pipelines, inconsistent logic and workflows that operate outside governance controls.

This is addressed by creating more unified environments – where teams can work across different tools and languages while still operating against the same governed data.

AI is also reshaping this experience, not as a separate layer, but as a way to help teams move faster within existing processes.

Capabilities like SAS Viya Copilot for Code Assistance provide context-aware support directly within pipelines and code, grounded in the active environment and operating entirely within governance controls.

2. Data governance: From afterthought to embedded capability

As access expands, governance must move from an afterthought to a core capability embedded in how data is used.

In many environments, governance is still applied after pipelines are built, disconnected from how data is accessed and reconstructed during audits. This approach does not scale for AI.

SAS shifts governance upstream, embedding it directly into access, workflows and decision-making.

This allows users to answer a fundamental question: Can I access this data and can I trust it?

At the same time, governance becomes easier to use. Natural language capabilities help users understand context, lineage and quality without interrupting their workflow.

Importantly, governance doesn’t stop at access. Continuous monitoring helps identify drift, detect anomalies and surface downstream impact – ensuring trust persists over time.

3. Data acceleration: Bringing analytics to the data

Once data is prepared and governed, execution becomes the next constraint. Traditional architectures assume that data must be moved to analytics environments, creating duplication, latency and cost.

SAS takes a different approach: bringing analytics to the data.

This reduces movement, improves performance and simplifies execution across platforms. It also helps organizations to scale analytics and AI more efficiently – without reworking their existing data infrastructure.

From capabilities to outcomes

When engineering, governance, and acceleration operate together, the result is not incremental improvement – but a structural shift.

• Data engineering becomes faster and more consistent
• Governance becomes continuous and embedded
• Execution becomes scalable and efficient

Together, these changes enable organizations to move beyond experimentation to AI that is operational, repeatable and trusted.

The bottom line

AI doesn’t fail in the model. It fails in the data.

SAS is reinventing data management around a simple idea: Data engineering, governance and acceleration must operate as one system.

Because in the end, success with AI depends on whether your data is truly ready.

Watch the full Data Management Reinvented roadmap session above

Want to explore more?

• Explore how SAS Viya delivers a unified data intelligence layer
• Download the eBook: Transform your data into an AI engine