Google Cloud unveils new GenAI-fueled data, analytics tools

Progress toward GA

Making tools that had been in the preview and development stages generally available may seem relatively insignificant because it’s the natural progression of product development. However, generative AI capabilities going into general availability is a significant step, according to David Menninger, an analyst at ISG’s Ventana Research.

In the 15 months since OpenAI launched ChatGPT, generative AI has become the top product development focus for not only Google Cloud’s analytics and data management portfolios but also those of fellow tech giants AWS and Microsoft. In addition, data platform vendors including Databricks and Snowflake have prioritized generative AI, as have more specialized analytics and data management vendors including Alteryx, Informatica, Tableau and ThoughtSpot.

The reason is that generative AI has the potential to expand analytics use within enterprises as well as improve the efficiency of those employees already working with data.

Generative AI large language models (LLMs) have extensive vocabularies that enable the true natural language processing (NLP) that NLP tools could not because of their limited vocabularies. True NLP, meanwhile, virtually eliminates the need to write code to query and analyze data, broadening the potential audience for otherwise complex analytics platforms.

In addition, LLMs can be trained to generate code and automate processes, which makes existing data experts more productive by eliminating time-consuming repetitive tasks.

However, despite what the addition of generative AI capabilities to data management and analytics platforms could mean, most of the tools unveiled by data management and analytics vendors remain in some stage of development.

Tableau made its first generative AI tool generally available Feb. 22, and a handful of other vendors have also made some features generally available. But they are the exceptions, to date.

We are in the middle of a race to get generative AI capabilities into production in enterprises. We’re still early in that race, so it is significant that we are making progress.

Therefore, it’s meaningful that Google Cloud is making some of its generative AI-fueled data management and analytics capabilities available for use now.

“We are in the middle of a race to get generative AI capabilities into production in enterprises,” Menninger said. “We’re still early in that race, so it is significant that we are making progress. But no individual vendor can be declared the winner yet.”

Now that AlloyDB AI is generally available, it will enable developers to build AI applications by using vectors to feed retrieval-augmented generation (RAG) pipelines that continuously update and train AI, including generative AI models.

Meanwhile, Gemini models in BigQuery infuse BigQuery with generative AI, essentially enabling customers to connect their BI to their AI.

Following the general availability of both AlloyDB AI and generative AI capabilities in BigQuery, Google Cloud plans to make more data management and analytics capabilities generally available in April, according to Gerrit Kazmaier, vice president and general manager of data and analytics at Google Cloud, and Andi Gutmans, Google Cloud’s general manager and vice president for databases.

Both executives spoke during a press conference Tuesday.

Gutmans said the tech giant could have held back the general availability of AlloyDB AI and BigQuery’s generative AI integrations until Google Cloud Next ’24, a user conference scheduled for April 9-11. However, the features are ready now and there will be many other product developments to unveil during the conference.

“Customers are showing a lot of interest and excitement in a lot of these capabilities — they want them yesterday,” Gutmans said. “We didn’t want to artificially hold back capabilities we have ready to go, so we decided to get them out [before the conference]. Rest assured, we have plenty of announcements lined up for April.”

The next wave

Beyond making AlloyDB AI and BigQuery’s integrations with Gemini models generally available, Google Cloud introduced a new wave of generative AI capabilities planned for its data management and analytics portfolios.

Prominent among them is an emphasis on vector search and storage.

Business intelligence has historically been based on structured data, such as financial records and point-of-sale transactions. However, it’s estimated that only about 20% of the world’s data is structured data, while 80% of the world’s data is unstructured data, such as text, audio files, videos and photographs.

Vectors are numerical representations of data that when assigned to text, audio files and other unstructured data give it structure. Once given structure with a vector, previously unstructured data can then be searched, discovered and operationalized.

In particular, vector search and storage are emerging as a means of feeding and training generative AI models with an enterprise’s proprietary data so the model can help inform business-specific decisions.

AlloyDB AI automatically generates vector embeddings using SQL, turning AlloyDB into a vector database.

Next, Google Cloud will add support for vector search across its entire suite of databases with vector search in the open source Redis database, and Google databases Cloud SQL, Spanner, Firestore and Bigtable now in preview.

“Our belief is that any database, anyplace you’re storing operational data that you may use for a generative AI use case, should have vector capabilities,” Gutmans said. “Good vectors should just be a foundational capability of a database.”

Like the vector search capabilities targeted for Google Cloud’s databases, the tech giant unveiled new capabilities in BigQuery aimed at enabling customers to operationalize unstructured data.

Google unveiled vector search in BigQuery in preview Feb. 14. On Thursday, Google Cloud added an integration — also in preview — with Vertex AI aimed at enabling customers to analyze text and speech for potential insights.

Such data could not previously be used in a meaningful way, according to Kazmaier. But using it in combination with structured data will enable customers to develop new applications for their data such as customer sentiment from audio files.

“This data is usually not used in enterprise data analytics,” Kazmaier said. “We are entering an age where we have a whole new set of data and analytical capabilities.”

Menninger likewise noted that Google Cloud’s integrations between Gemini, BigQuery and Vertex AI to derive analytics value from unstructured data files is notable.

“Gemini’s multimodal capabilities to enable text, image and video data analysis are new and are at the forefront of what’s happening with GenAI,” he said. “So much of the data enterprises process is unstructured, making it critical to include this information in any analysis to have a complete picture.”

In addition to its emphasis on making unstructured data more accessible, Google Cloud unveiled RAG support in BigQuery and support for LangChain, a developer framework for building generative AI models, across its entire suite of databases.

Next steps

Google Cloud will make more generative AI-fueled data management and analytics capabilities generally available in the near future, Gutmans and Kazmaier said.

And as the tech giant introduces other new tools, one of its guiding principles will be enabling access to an ever-expanding swath of data signals, according to Kazmaier.

He noted that big data was essentially a collection of a lot of similar data. But freeing unstructured data for analysis actually enriches decisions by providing a wide range of data.

“We tend to think about GenAI as a mostly assistive functionality,” he said. “But when you really think about data and analytics and the challenges we have left, they’re really about going from what was called big data to what we call wide data.”

Menninger, meanwhile, said that amid the proper focus on generative AI, it’s important for not only Google Cloud but also all data management and analytics vendors to continue adding traditional AI capabilities.

Traditional AI is an important aspect of such applications as fraud detection and predictive maintenance, he noted. Therefore, it should not be ignored.

“GenAI investment is sucking all the air out of the room,” Menninger said. “It’s important to recognize that traditional AI is still an important element of certain use cases. So, I am anxious to see how Google and other vendors are going to bring together the worlds of GenAI and traditional AI to meet all the requirements of enterprises today.”

Eric Avidon is a senior news writer for TechTarget Editorial and a journalist with more than 25 years of experience. He covers analytics and data management.