Introduction
Last week, the Google Cloud Summit in London brought together industry leaders to discuss the future of databases and data interaction. The event highlighted a significant shift towards AI-driven databases, with a focus on inexact queries and the growing role of agents in data platforms. Executives Sailesh Krishnamurthy, VP of Engineering, and Yasmeen Ahmad, Product Executive for Agentic Data Cloud, shared insights into Google's vision for AI-native infrastructure and the evolution of data interaction.
The summit underscored the transformative impact of AI on database technology, emphasizing the need for more flexible and context-aware queries. This shift is not just about improving efficiency but also about redefining how businesses interact with data. As AI continues to evolve, the traditional methods of data retrieval and analysis are being replaced by more dynamic and intelligent systems.
AI-Driven Databases and Inexact Queries
One of the key themes at the summit was the emphasis on inexact queries in databases. Unlike traditional queries that focus on retrieving exact results, inexact queries aim to provide the best possible results based on context and intent. This shift is driven by the increasing use of AI and agentic workloads, which require more flexible and context-aware queries.
For example, natural language queries that take into account previous interactions are becoming more prevalent. This approach allows for a more intuitive and user-friendly experience, as users can now interact with databases in a more conversational manner. The focus is on understanding the intent behind the query rather than just matching keywords.
This evolution is particularly significant for businesses that rely on large-scale data analysis. By leveraging inexact queries, organizations can gain more relevant insights and make more informed decisions. However, this shift also introduces new challenges, such as ensuring the accuracy and reliability of the results.
AI-Native Infrastructure
Google is heavily investing in AI-native infrastructure, which includes advancements in vector indexing, text indexing, and graph technology. This infrastructure is designed to handle both structured and unstructured data, operating in terms of inexact results and data quality.
A notable development is the evolution of the Knowledge Catalog, formerly known as Dataplex. This platform aggregates organizational data from multiple sources, including structured and unstructured data, and serves as context for large language models (LLMs). By providing a comprehensive view of an organization's data, the Knowledge Catalog enables more accurate and context-aware AI applications.
This infrastructure is crucial for supporting the growing demand for AI-driven applications. As businesses increasingly rely on AI to drive innovation and efficiency, the need for robust and scalable AI-native infrastructure becomes more apparent. Google's investments in this area are aimed at providing the necessary tools and technologies to support this evolution.
AI Functions in Google SQL
Google SQL, used in Spanner, AlloyDB, and BigQuery, now includes AI functions such as AI.IF, which evaluates conditions described in natural language and returns true or false. These functions leverage Gemini LLMs to evaluate prompts, which can return errors or null values if the model fails or is unavailable.
Proxy models, which are tiny models trained on the fly based on a small sample of data, are being used to improve efficiency and reduce latency. These models provide a lightweight alternative to larger AI models, enabling faster and more efficient data processing.
The integration of AI functions into Google SQL represents a significant step forward in making databases more intelligent and responsive. By incorporating AI capabilities directly into the query language, Google is enabling developers to build more sophisticated and dynamic applications.
Future of Data Interaction
Yasmeen Ahmad predicts that humans will not interact directly with data platforms in the next three to five years. Instead, humans will orchestrate agents, and agents will perform the actual work. This shift is driven by the evolution of AI models from co-pilots or assistants to systems capable of multi-step execution, parallel execution, and handling complexity.
The focus for humans will be on defining intents, goals, and outcomes, while AI models handle the low-level data wrangling. This transformation promises to make data interaction more efficient and accessible, allowing businesses to leverage the full potential of their data assets.
However, this shift also raises important questions about the role of humans in the data interaction process. As AI models take on more responsibilities, it will be crucial to ensure that humans retain control and oversight over the decision-making process.
Business Intelligence and Conversational Analytics
Traditional dashboards are being replaced by conversational analytics for business users. This shift is driven by the need for more intuitive and user-friendly data interaction tools. Improved context, such as that provided by the Knowledge Catalog, mitigates issues like hallucinations and prompt injections in generative AI.
Customers have achieved over 90% accuracy with conversational analytics, a significant improvement from 18 months ago. This progress highlights the potential of AI-driven analytics to transform business intelligence, making it more accessible and effective.
The adoption of conversational analytics is expected to accelerate as businesses recognize the benefits of more intuitive and interactive data tools. This trend is part of a broader movement towards democratizing data access and empowering users to make more informed decisions.
Growth and Challenges
AI is driving growth in data storage and token usage, with Spanner now running 7.5 billion queries per second at peak, up from 5 billion queries per second a year ago. Spanner and BigTable each handle roughly 7 billion queries per second and store double-digit exabytes of data.
The inefficiency of AI functions and the cost of tokens are challenges that need to be addressed. However, higher productivity and reduced staff costs may offset these challenges. As AI continues to evolve, it will be crucial to find ways to optimize performance and reduce costs.
The rapid growth in data storage and processing capabilities is a testament to the transformative impact of AI on database technology. However, this growth also brings new challenges, such as ensuring the scalability and reliability of AI-driven systems.
Conclusion
The Google Cloud Summit London 2026 highlighted the significant shifts in database technology driven by AI, including the move towards inexact queries, AI-native infrastructure, and the increasing role of agents in data interaction. These advancements promise to transform how businesses interact with data, making it more accessible and efficient.
However, these changes also introduce new challenges that need to be managed. As AI continues to evolve, it will be crucial to find ways to optimize performance, reduce costs, and ensure the accuracy and reliability of AI-driven systems. The future of data interaction is undoubtedly exciting, but it will require careful planning and execution to realize its full potential.