Denodo’s Predictions for 2025
Reading Time: 4 minutes

In 2024, generative AI (GenAI) became top-of-mind, as companies began to leverage it for increased productivity. Additionally, storage continued to grow in capacity, epitomized by an optical disk designed to store a petabyte of data, and the global Internet population climbed up to 5.5 billion. But what do these trends tell us about 2025? Every year at Denodo, we gather predictions from across the company. For 2025, our top 5 predictions centered around the increased demands of AI and distributed data:

  1. For GenAI initiatives to be successful, organizations will need to embrace distributed data. – From Alberto Pan, Chief Technical Officer (CTO) for Denodo 
  2. GenAI initiatives will require a wider, more organizational focus. – From Terry Dorsey, Data Architect/Evangelist at Denodo
  3. Autonomous agents and agentic workflows will set the stage for GenAI development. – From Terry Dorsey
  4. Companies will leverage hybrid infrastructures, even moving key datasets to on-premises systems. – From Alberto Pan
  5. Denodo will draw from multiple platforms. – From Alberto Pan 

In this post, I’ll cover each of these predictions in turn.  

Prediction 1: Organizations Will Need to Embrace Distributed Data for GenAI

Alberto Pan predicts that by 2026, more than 50% of companies will identify data system distribution and heterogeneity as their primary challenge in developing GenAI-ready data products. 

GenAI applications need “AI-Ready” data, or data drawn from across an organization, in a secure, governed manner, in real time. However, even the most advanced approaches to providing GenAI applications with access to enterprise data sources, such as retrieval-augmented generation (RAG), cannot address the complexity of providing AI applications with AI-Ready data, which is inherently distributed.

Logical data management platforms can provide AI applications with the kind of AI-ready data described above. Such platforms enable real-time access to distributed data sources, while offering the ability to enforce security and data governance policies across the entire data estate from a single point of control. This provides AI applications with trusted, authoritative data, in real time.

Prediction 2: GenAI Initiatives Will Require a Wider, More Organizational Focus

In embracing AI, organizations are creating new roles, like Chief AI Officer (CAIO) and revisiting existing roles, like CIO and CDO, to accommodate AI initiatives. However, AI research is progressing extremely quickly, and this can sometimes outstrip adoption. 

Terry Dorsey says, “Many companies are attempting to integrate AI within the same organizational structures and using traditional methods, which may not be adequate. I predict that in 2025, we will see new focus areas for AI-driven transformation.” 

Dorsey predicts that such focus areas will include: 

    • Data Security and Privacy. This could involve the development of data governance frameworks that prioritize centralized yet flexible security models, to balance data protection with accessibility.
    • Enhanced Change Management and Change Control. Best practices may include cross-functional AI task forces, clear communication protocols, and training programs to facilitate smooth transitions during critical changes.
  • Alignment with Business Outcomes. Cross-functional collaborations, led by both technical and business leaders, would help AI projects align with core organizational objectives, to deliver tangible value.
  • Business Process Optimization and Integration of Emerging Technologies. This could include developing adaptive workflows that enable AI-driven insights to be seamlessly incorporated into business operations, for virtuous value-cycles. 
  • Restructuring Corporate IT for Agility and Collaboration. Companies could establish hybrid roles or dedicated AI integration teams within IT, blending technical expertise with domain-specific knowledge, to more effectively support AI and data initiatives.

Prediction 3: Autonomous Agents and Agentic Workflows Will Be the “New Normal”

Large language models (LLMs) can perform some incredible feats, and here at Denodo, we  especially appreciate the text-to-SQL and summarization capabilities in the Denodo Platform. 

Dorsey predicts that autonomous agents and agentic workflows will be a regular feature of the AI landscape in 2025. “Since LLMs are very good at assessing/reviewing information and have no ego when it comes to self-assessment,” says Dorsey, “we are seeing much research and many frameworks that seek to exploit this capability. They are also very good at making decisions about tasks and structuring information from natural language. These capabilities are the basis for autonomous agents and agentic workflows.”

Major players like Amazon, Google, and Microsoft have developed powerful frameworks that make it easier for companies to build AI-driven agents and integrate them into operations. With tools like Amazon Bedrock Agents and Google’s Vertex AI, businesses can now create agents that pull in data, respond to customer questions, and even perform actions, all without much human oversight. Organizations can start slowly, implementing and observing autonomous agents and agentic workflows before enlisting them for production-grade work. 

Prediction 4: The Future will be Hybrid

By 2026, Pan predicts, over 80% of organizations building centralized cloud data warehouse or data lakehouse infrastructures will decide to migrate certain workloads to other environments, including alternative data processing systems within the same cloud provider, systems in other clouds, or even on-premises environments, in a process known as data repatriation.

He made this prediction after reading in IDC’s June 2024 report, “Assessing the Scale of Workload Repatriation,” that around 80% of respondents anticipated some level of data repatriation in the coming 12 months. 

Repatriation is complex and costly, so organizations will also optimize costs by choosing the cloud environment and system that offers the best balance of efficiency and cost-effectiveness for each use case. Organizations should invest in technologies that simplify the migration of use cases to the most appropriate environment, as technology and business needs evolve. Open Table formats enable data representation compatible with multiple processing engines. Additionally, logical data management platforms shield data consumers from the nuances of individual processing engines, including SQL dialects, authentication protocols, and access control mechanisms.

Prediction 5: Data Products Will Draw from Multiple Platforms

Pan predicts that by 2026, more than 80% of organizations pursuing a data products strategy will create powerful data products that leverage multiple data platforms. This shift will pose challenges for enterprise-wide data democratization initiatives in organizations that initially envisioned a single-vendor approach.

Data products draw on an increasing number of distributed data sources, as no single platform can optimize functionality, performance, and cost across all data products. Consider that fewer than 5% of the people who leverage both Snowflake and Databricks plan to decommission one of these platforms, and the majority of these customers also plan to leverage additional cloud and on-premises systems. New data platforms will continue to emerge, so data product development should account for distributed data and a multiplicity of platforms.

Logical data management platforms support the development of data products by establishing a unified, governed infrastructure across diverse platforms. This approach enables data product developers to leverage diverse data sources while enabling the interoperability, reusability, and discoverability of all data products.

Looking ahead in 2025 

Clearly, two interconnected trends will set the tone for this year: The increasing demands of AI, and the increasingly distributed nature of data. But for both, a logical approach to data management promises to provide organizations with the agility and flexibility to thrive.

Saptarshi Sengupta
Latest posts by Saptarshi Sengupta (see all)