Following up on our first recap, we’re diving into more of the transformative announcements from Google Cloud Next 2025. From AI-powered geospatial insights to multi-agent developer frameworks, and from conversational analytics to role-specific data assistants, these innovations represent powerful ways that marketers and businesses can unlock value from their data through the power of AI.
1. Geospatial Intelligence Gets a Generative AI Boost
Google shared new research leveraging generative AI and multiple foundational models, like its Population Dynamics Foundation Model (PDFM), to analyze geospatial data.
This innovative approach enables the extraction of richer, more nuanced insights from location-based information like satellite imagery and maps. As the advertising industry launch partner for PDFM, we’re integrating PDFM with our proprietary media performance data to unlock a new dimension of audience insights - grounded in location, events, and real-world behavior.
With AI-powered geospatial reasoning, clients can now gain a significant advantage by uncovering previously hidden patterns and trends within geographic data. This deeper understanding allows for more strategic decision-making in areas such as localized targeted advertising, geo expansion, supply chain optimization, and personalized customer engagement, ultimately driving greater efficiency and ROI across operations.
For example, a CPG company or retailer could use this technology to optimize product distribution and local advertising by analyzing neighborhood-level consumer behavior, foot traffic patterns, and regional demand trends—ensuring the right products are stocked in the right places, and promotions are launched where they’ll have the highest impact. Similarly, a film studio that produces and distributes movies could leverage geospatial reasoning to identify which regions have the highest potential audience engagement, allowing them to strategically target theatrical releases, tailor regional marketing campaigns, and allocate resources to cinemas in areas with the greatest box office potential.
This research demonstrates how generative AI can automate and enhance the interpretation of complex spatial datasets, enabling marketers to leverage location intelligence at scale for more targeted and effective campaigns. You can read more about this announcement here.
2. ADK + MCP: A Foundation for Multi-Agent AI Applications
The Agent Development Kit (ADK) is an open source framework designed to simplify the creation of complex applications that use multiple AI agents working together. It allows developers precise control over how these agents behave and interact, enabling standardized toolkits for building advanced multi-agent AI solutions, potentially accelerating development cycles, improving code maintainability, and enabling more ambitious AI projects. It also enables the development of more powerful and sophisticated AI applications capable of handling complex tasks by coordinating multiple specialized agents, leading to richer features and automation.
For example, a travel planning application built with ADK could use one agent to find flights, another for hotel bookings, and a third for local activities, all collaborating to create a complete itinerary based on user preferences.
On a related note, ADK also incorporates support for the Model Context Protocol (MCP). MCP offers a standardized way for AI models to connect with and use various data sources and tools. This eliminates the need to build custom integrations for each data source or tool that the AI needs to interact with.
Together, ADK and MCP promise faster development, better maintainability, and more scalable AI systems — though real-world adoption will depend on community contributions and thoughtful architecture.
3. Specialized AI Agents for Data Teams
Google introduced assistive agents purpose-built for roles like engineers, scientists, and analysts. Embedded directly into tools like BigQuery and Looker, these agents leverage Gemini models and are grounded in the user's trusted data.
Although these agents are built for data teams they also provide important value for media and marketing. For example, these agents can help agencies and advertisers automate the heavy lifting across the data and AI lifecycle—making workflows more efficient, campaigns more insightful, and decisions more data-driven. By automating routine operations, offering intelligent recommendations, and enabling scalable modeling, these agents empower marketing teams to focus on strategic, high-impact work like campaign optimization, audience modeling, and creative effectiveness analysis.
Specifically, the Data Engineering Agent can help advertisers and agencies streamline marketing data operations by:
- Building automated data pipelines that ingest and transform campaign data from platforms like Google Ads, Meta, YouTube, and CRM tools
- Performing data preparation at scale, including transformation and enrichment —for example, enriching customer segments with geo, device, and behavioral attributes
- Maintaining data quality (currently in Preview), with built-in anomaly detection to flag unusual shifts in ad performance metrics, spend patterns, or broken feeds
- Automating metadata generation, making it easier to manage multi-client datasets, track data lineage, and comply with privacy and audit requirements
In practice, this means a media agency could use the Data Engineering Agent to automatically flag anomalies in campaign performance, auto-generate metadata for new advertiser datasets, and reduce manual prep time for reporting.
Additionally, the new Data Science Agent serves as a powerful foundation for advertisers’ advanced analytics needs and is designed to be your core data science workbench, accelerating the delivery of models and insights by:
- Automating feature engineering to create powerful predictive variables from raw campaign and audience data
- Offering intelligent model selection, helping choose the best approach for tasks like churn prediction, LTV scoring, or optimal bid strategy
- Supporting scalable training, allowing teams to train models on large datasets (e.g., impressions, conversions, user behavior) without infrastructure bottlenecks
- Accelerating experimentation, enabling faster A/B test analysis and iterative model development
For example, an advertiser's data science team can use the agent in Colab to quickly generate features from campaign logs, receive model suggestions for conversion prediction, and scale model training across multiple audience segments.
4. Looker’s Conversational Analytics API: Data Talks Back
Looker’s new conversational analytics API lets users ask business questions in natural language and get instant, accurate answers — no SQL or dashboard wrangling required.
This is facilitated through a new conversational analytics API that data teams can embed into existing applications and workflows. It leverages AI to understand user intent based on business terms (like 'revenue') and accurately calculates metrics using the LookML semantic layer.
This democratizes data access by allowing non-technical users to ask questions about data in plain language and get answers quickly. It improves the ease of use for data exploration and analysis while also potentially improving the accuracy of LookML by understanding user queries better.
For example, a marketing manager could ask their embedded analytics tool, "What was our total revenue by region last quarter?" and receive a chart or table with the answer, without needing advanced technical skills.
Note that the conversational analytics API was announced as being in Preview, meaning its features, stability, and availability might change. Effectiveness depends on the quality and completeness of the underlying LookML model.
5. AlloyDB AI Evolves into an AI-Native, Natural Language-Ready Database
AlloyDB AI is a high-performance, AI-enhanced database platform purpose-built for enterprise-scale workloads—including those driving modern marketing, advertising, and media operations.
With its support for real-time analytics, intelligent automation, and seamless integration with advanced AI models, AlloyDB AI now empowers advertisers and their agencies to turn data into action faster—whether for campaign optimization, dynamic audience segmentation, or creative personalization.
At Google Cloud Next, several key enhancements were announced that make AlloyDB AI even more powerful for media and marketing workflows:
- Natural Language Querying: Users can query structured data using plain language, bypassing the need for SQL.
- Agentspace Integration: AI agents within Google Agentspace can now retrieve structured data directly from AlloyDB AI.
- Vector Search + SQL Optimization: AlloyDB AI now supports seamless blending of semantic vector search with traditional SQL filters and joins.
- Integrated AI Models: AlloyDB AI has three new three built-in models — a reranking model for search quality, a multimodal embeddings model for cross-format search, and a Gemini embedding model for deeper text understanding.
As a result of these enhancements, AlloyDB AI now enables more accessible, intelligent, and responsive data interactions:
- Non-technical users can query data with natural language.
- Applications can execute smarter, hybrid search experiences that combine semantic meaning with structured filters.
- Developers gain access to Google’s advanced AI models directly inside the database for real-time search enhancement and content understanding.
Final Thoughts
From agent-based development frameworks to intuitive analytics and deep spatial insights, these announcements reflect Google’s continued mission to embed AI deeply across the entire enterprise stack.
Whether you're a developer, data scientist, marketer, or strategist, the message is clear: AI is no longer something to explore — it’s something to build with, at scale. Google’s ecosystem is fast becoming a launchpad for smarter workflows, faster insights, and entirely new AI-powered applications.
.webp)
.webp)