“D&A is going from the domain of the few, to ubiquity,” said Gareth Herschel, VP Analyst at Gartner. “At the same time D&A leaders are under pressure not to do more with less, but to do a lot more with a lot more, and that can be even more challenging because the stakes are being raised. There are certain trends that will help D&A leaders meet the pressures, expectations and demands they are facing.”
Highly Consumable Data Products
To capitalise on highly consumable data products, D&A leaders should focus on business-critical use cases, correlating and scaling products to alleviate data delivery challenges. Prioritising the delivery of reusable and composable minimum viable data products is essential, allowing teams to enhance them over time. D&A leaders must also come to a consensus on key performance indicators between producing and consuming teams, which is vital for measuring data product success.
Metadata Management Solutions
Effective metadata management begins with technical metadata and then expanding to include business metadata for enhanced context. By incorporating various metadata types, organisations can enable data catalogues, data lineage, and AI-driven use cases. Selecting tools that facilitate automated discovery and analysis of metadata is imperative.
Multimodal Data Fabric
Building a robust metadata management practice involves capturing and analysing metadata across the entire data pipeline. Insights and automations from the data fabric support orchestration demands, improve operational excellence through DataOps, and enable data products.
Synthetic Data
Identifying areas where data is missing, incomplete, or costly to obtain is crucial for advancing AI initiatives. Synthetic data, either as variations of original data or replacements for sensitive data, ensures data privacy while facilitating AI development.
Agentic Analytics
Automating closed-loop business outcomes with AI agents for data analysis is transformative. Piloting use cases that connect insights to natural language interfaces and evaluating vendor roadmaps for digital workplace application integration are recommended. Establishing governance minimises errors and hallucinations, while assessing data readiness through AI-ready data principles is essential.
AI Agents
AI agents are valuable for ad hoc, flexible, or complex adaptive automation needs. Beyond relying solely on large language models (LLMs), other analytics and AI forms are necessary. D&A leaders should enable AI agents to access and share data across applications seamlessly.
Small Language Models
Consideration of small language models over large language models is advised for more accurate, contextually appropriate AI outputs within specific domains. Providing data for retrieval of augmented generation or fine-tuning custom domain models is recommended, especially for on-premises use to handle sensitive data and reduce compute resources and costs.
Composite AI
Leveraging multiple AI techniques enhances AI's impact and reliability. D&A teams should diversify beyond GenAI or LLMs, incorporating data science, machine learning, knowledge graphs, and optimisation for comprehensive AI solutions.
Decision Intelligence Platforms
Transitioning from a data-driven to a decision-centric vision is crucial. Prioritising urgent business decisions for modelling, aligning decision intelligence (DI) practices, and evaluating DI platforms are recommended steps. Rediscovering data science techniques and addressing ethics, legal, and compliance aspects of decision automation are essential for success.
