Attended Data & AI Summit '25 Fall

On November 27, 2025, we attended the "Data & AI Summit '25 Fall" hosted by Google.

Our company has now decided to take a serious approach to "data utilization," and a project quietly kicked off. At first, we were proceeding in a relaxed way, taking on what we could, step by step. Then one day, we came across information that Google was hosting the "Data & AI Summit '25 Fall."

We thought, "We need to keep up with the latest data utilization trends!" So we decided to attend with a casual mindset, and it turned out to be the right call. We heard far more valuable insights than we had expected.

What Is Data & AI Summit '25 Fall?

As you may have guessed from the event name, the main theme of the keynotes was "AI agents." The event was held in a hybrid format and drew a strong turnout of approximately 3,000 participants. I attended in person, and some sessions had lines forming outside or were so full that attendees had to stand. It made me feel firsthand how much attention data utilization and AI are receiving.

Here's the timetable for the day.

https://cloudonair.withgoogle.com/events/data-ai-summit-25f

In this article, we report on the keynote and the sessions that left the strongest impression on us!

Keynote: Entering the Era of Co-Creating with Data Agents - Next-Generation Data Analysis Powered by an AI-Ready Data Foundation

The keynote was structured in three parts: it opened with Mr. Hamada from Google Cloud discussing "Three Requirements for AI Agents," followed by case studies on data utilization from Lion Corporation and Mercari, Inc., and concluded with a demo of conversational analytics released by Google's technology division.

◼︎Three Requirements for AI Agents

Mr. Hamada presented the three essential requirements for deploying AI agents to take on real business tasks: "Action" (proactivity), "Memory" (retention), and "Context" (reliability).

Action: AI agents should act proactively rather than passively.

The key here is "transparency of thought process." The session explained that the concept of "Advanced Reasoning," which breaks a series of flows into stages, is central to improving the transparency of analysis.

Memory: "Short-term memory" and "long-term memory" that allow AI agents to function effectively.

In the discussion of "long-term memory," which gives AI agents real-time access to a company's data, a case was presented in which customers placed product orders by communicating with an AI agent via voice input. It was a reminder that, compared with traditional tap-based interfaces, this is far faster and is already becoming a familiar part of everyday life.

Context: Context is essential for ensuring the reliability of answers.

The session explained the importance of a "semantic layer" for guiding users to the correct database and calculation formulas.

I felt that the semantic layer came up often in the sessions that followed, and that its visibility is growing even further.

◼︎Case studies on data utilization

Lion Corporation presented on building a data foundation, and Mercari, Inc. introduced its efforts toward data democratization.

The topic of data democratization was especially memorable, as it felt close to home. With AI agents, the structure that enables anyone to access data, a role traditionally held by analysts, is something our company would like to learn from.

At an internal sharing session held afterward, people were already talking about wanting to adopt it right away.

◼︎Demo of conversational analytics

A demo of a conversational analytics agent was presented, using "Conversation analytics agent in BigQuery" as an example.

Although I had used BigQuery a little, I had no idea it was now possible to analyze data in natural language, and it was truly eye-opening. The support is also comprehensive, including data accuracy improvements through Gemini suggestions and automatic generation of golden queries, and I was very impressed that the setup is relatively simple.

Conversational Analytics API in Practice! Building AI Agents That Pull Data Hidden in BigQuery Through Natural Language

In a presentation by Eilet Co., Ltd., a case study was introduced that combined BigQuery and the Conversational Analytics API to enable data analysis through a chat-style interface.

The approach stored website data in BigQuery and fully delegated query execution and analysis tasks to the Conversational Analytics API. By allowing AI agents to be accessed from everyday communication tools, the setup made it easy for anyone to make use of data.

It embodied the "democratization of data" discussed in the keynote, and the structure was remarkably simple. We immediately decided to adopt it at our company.

We would love to implement it ourselves, but given our realistic skill level, it would take years, so we have decided to entrust this to our engineers!

Minimal Steps! Google Trends Integration AI Agent × Product Master Collaboration

In a session by AEON Smart Technology Co., Ltd., the speaker presented how AI is applied to product master data, an initiative unique to companies with physical stores.

The presentation covered two challenges: resolving fuzzy search and automatically matching trends. Both were built using Google Vertex AI Search.

— Fuzzy Search in Product Master Using AI

As a challenge unique to companies with physical stores, the speaker described vague customer questions that lack specific product names. Traditional systems require exact matches, so questions humans can answer easily become major barriers for the system. The talk presented an AI-based approach to address this issue.

The demo showcased a system that instantly delivers appropriate product information from big data, even in response to vague inputs like those in an internet search. I didn't fully understand how it worked, but I was simply amazed that a company could go this far.

— Trend tracking and automatic matching with product data

The talk covered a tool under consideration that instantly captures external trend information and automatically links it to the company's own product data, with the aim of shortening the lead time from the moment customer needs arise to the point where products actually appear in stores.

This was another session on cutting-edge technology, and it was really fascinating. I haven't yet figured out how it could apply to our services, but I personally would love to try out "Google Vertex AI Search."

Summary

I've only highlighted the sessions that left the strongest impression, so how did it land for you? I hope I was able to share, even a little, the excitement I felt about "the innovation brought by data agents."

To be honest, we had been taking a fairly relaxed approach to data utilization, but attending this summit was a very good opportunity that made me think, "We need to get serious now so we don't get left behind by this wave!"

First, I plan to start with "defining data and the semantic layer," which was emphasized repeatedly throughout the event.