What 2025 Holds For Data And Analytics: 8 Predictions

1 year ago 74

Malcolm Hawker is the CDO at Profisee, and is an expert in the fields of Data Strategy, Master Data Management and Data Governance.

getty

It’s the beginning of another year, which means it’s annual prediction time.

2024 was a tumultuous time for many, with lingering impacts from high prices, global uncertainties and AI-driven disruptions. What will 2025 bring to the world of data and analytics?

The Decline Of CDOs

This year, we will see a decrease in the number of chief data officers (CDOs) in private-sector companies. This will result from an ongoing failure of many CDOs to operationalize AI, a trend recently noted in a Wavestone survey noting only 4.7% of companies had a GenAI-based solution in production.

As a result, many CDO functions will move under CTOs, and many will struggle to retain a grip over governance responsibilities. The struggles of CDOs will hinder corporate AI adoption but will not stop it. Off-the-shelf LLM adoption will increase regardless of whether CDOs can adapt their legacy data management approaches for AI use cases.

The Data Catalog Hype Machine

Data catalog hype will reach a fevered pitch in 2025, especially after Ole Olesen-Bagneux publishes his book on the "meta grid," his third on catalogs. I predict the meta grid will become the next "big thing" in data, particularly following Ole’s keynote speech at Data Day Texas.

One positive outcome I predict from this hype is there will be growing talk of the need to bridge the worlds of structured and unstructured data and the role a "next-gen" data catalog—or the "meta grid"—will play in filling the gap between traditional metadata and knowledge management solutions.

Semantics, RAGs And Knowledge Management

In 2025, we’ll see a growing awareness of knowledge management practices in the world of traditional data management. This focus will stem, in part, from the growth of complex retrieval-augmented generation (RAG) patterns that influence the behavior of LLMs, where insights stored as knowledge are more actionable for LLMs than insights stored as data.

RAG patterns are also growing in popularity because studies show drastic improvements in the accuracy of LLMs when using them. One study from Juan Sequeda (with Dean Allemang and Bryon Jacob) showed that Graph RAG increased GenAI accuracy of "text to SQL" queries from 16% to 54%. Performance improvements like this are causing 60% of LLM applications in enterprises to leverage RAG patterns.

The AI For Data Management Paradox

While I predict many CDOs will continue to struggle to operationalize GenAI, a growing number will concurrently awaken to the fact AI is increasingly a requirement for managing their data estates at scale. This growing use of AI to scale data management was something discussed in multiple presentations at the 2024 Gartner Data and Analytics Summit, and I suspect this trend will only accelerate in 2025.

More focus will also come to the issue of using AI to bring structure to unstructured data so it can be better managed and—more importantly—leveraged by LLMs. The paradox of using AI to automate data management so that data can be more easily consumed by AI will be embraced by an increasing number of companies.

Data Products

Data products continued their march to the top of the Gartner hype cycle in 2024, which means there’s only one direction they can go this year: down.

This decline will result from many data product initiatives being unable to provide material business value given their highly internal focus. I discussed these challenges for data products in a recent episode of my CDO Matters podcast and as a guest on Brian T. O’Neill’s Experiencing Data Podcast.

Data Fabric

The data fabric will continue to build momentum in 2025, buoyed by ongoing investments and a continued focus on data management from Microsoft. If history is any guide, then the fabric will also gain momentum on the heels of competing products that could potentially launch in 2025. (Databricks and Snowflake, for instance, both launched data catalogs very shortly after Microsoft launched Purview.)

I predict that Microsoft will expand fabric capabilities to connect to operational data stores that house unstructured data, where data can be profiled by a data quality tool and analyzed into a structured format. Meanwhile, the fabric will remain primarily an analytical tool, which means any system that integrates into the fabric that can also support operational uses of data (like an MDM tool) will have significant advantages.

AI Governance

AI governance will continue to be the thing that everyone is talking about at conferences but can’t fully explain how to implement.

Meanwhile, there will be no shortage of people talking about implementing AI governance frameworks or trying to sell you a framework, but when it comes to defining the specific data governance policies and rules that can be executed at scale for things like AI ethics or AI bias, there will little clarity to be had.

Sustainability In Data

The focus on sustainability in data will continue to grow in Europe, fueled partially by the launch of digital product passports and other evolving ESG policies in the region. In the U.S. and Canada, the increasing problem of massive energy consumption within data centers, being driven by rampant data hoarding and an increased focus on AI, will remain largely ignored.

Conclusion

If you are a data leader looking to stay ahead of these rapidly evolving trends, I recommend the following four actions in 2025:

• Commit to quantifying the business value of the services you provide and find a way to drive value using a GenAI-based solution.

• Get familiar with the concept of knowledge management and how it differs from traditional data management. Consider attending a knowledge management conference.

• Familiarize yourself with complex RAG patterns and how to implement them using widely available tools—like LangChain and Hugging Face—which could be applied to any generally available LLM.

• Work to understand how integrating GenAI to automate your data management processes could drive increased efficiency and lower costs and embark on some form of a POC to test its benefits.

2025 will be an interesting year with plenty of opportunities for data leaders to become their company’s data heroes.


Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?


Read Entire Article