Google Embeds Ai Agents Deep Into Its Data Stack - Here's What They Can Do For You

Joan Cros/NurPhoto via Getty Images

ZDNET's cardinal takeaways

• Google is introducing powerful tech for agents and data.
• They are besides introducing a bid of data-centric agents.
• A caller command-line AI coding instrumentality is now available.

I americium nary alien to hyperbolic claims from tech companies. Anyone who's connected nan receiving extremity of a firehose of property announcements related to AI understands. Everything is game-changing, world-changing, nan most, nan best, yada, yada, yada.

And past there's Google. Google is nary alien to hyperbole. But erstwhile a institution truthful steeped successful information guidance arsenic portion of its halfway DNA talks astir "fundamental transformation," and says that nan world is changing because, "It's being re-engineered successful real-time by information and AI," we tin see those claims arsenic reasonably credible.

Also: Got 6 hours? This free AI training from Google tin boost your resume today

Just successful clip for Google Cloud Next Tokyo 2025, Google is making a bid of announcements that herald a awesome alteration successful really enterprises negociate data.

Yasmeen Ahmad, Google's managing head of Data Cloud, says successful a blog post, "The measurement we interact pinch information is undergoing a basal transformation, moving beyond human-led study to a collaborative business pinch intelligent agents."

She calls this agentic shift, which she describes as, "A caller era wherever specialized AI agents activity autonomously and cooperatively to unlock insights astatine a standard and velocity that was antecedently unimaginable."

Also: 5 ways to successfully merge AI agents into your workplace

From almost immoderate different company, claims for illustration this would look for illustration conscionable truthful overmuch basking air. But Google is dropping a bid of announcements astir caller offerings that supply real-world capabilities to information scientists and engineers successful beautiful tangible support of nan claims.

The agentic displacement

There's a reasonably good statement betwixt AI chatbots and AI agents. Chatbots are conversational, while agents are devices that execute autonomous tasks. Some users employment chatbots to execute tasks, arsenic I did when I utilized ChatGPT to analyse immoderate business data. Agents, like ChatGPT Agent, usage a conversational interface to person instructions.

A bully measurement to deliberation of agents is arsenic surrogate squad members. Perhaps 1 supplier does information normalization (cleaning up data), while different does migration. Each supplier does 1 aliases much defined tasks utilizing AI capabilities.

Also: Want AI agents to activity together? The Linux Foundation has a plan

In this context, Google is looking astatine agents that tin automate and simplify tasks for information workers, tin pass pinch each other, and tin free professionals from tedious activity truthful they tin attraction connected "higher-value tasks." Google is besides trying to get agents to activity together successful virtual teams.

There are, of course, questions astir whether agents aren't really freeing up nan clip of elder professionals, but are alternatively taking activity distant from much inferior employees. On nan different hand, I don't person anyone to do nan grunt activity erstwhile I'm afloat immersed successful a project. So thing I tin manus disconnected to an supplier is much clip for projects and writing.

Cognitive instauration

With each these agents moving around, accepted databases conscionable aren't up to nan task of keeping them fed. Agents do their reasoning aliases automation processes crossed silos. They request entree to some humanities and unrecorded data.

Classic information guidance methods for illustration real-time OLTP (online transaction processing) and deep-dive OLAP (online analytical processing) isolate information excessively overmuch for AIs to summation insights from trends and existent activities.

One measurement to thief unify each of these capabilities is by enhancing their database offerings. A fewer years ago, Google added a columnar motor for AlloyDB. AlloyDB is nan company's afloat managed database work connected Google Cloud Platform that focuses connected PostgreSQL users, which is perfect for those who require a PostgreSQL-specific solution.

Also: How AI agents tin make $450 cardinal by 2028 - and what stands successful nan way

A columnar motor is 1 wherever workloads query circumstantial columns of data, reference only nan fields needed for analysis. This leads to faster queries and allows for vectorized execution, wherever operations are applied to an full file of information astatine once.

Now, Google is adding a columnar motor to Spanner, its globally distributed, powerfully accordant database work that offers precocious readiness and scalability, designed for enterprises needing world scope and precocious transactional integrity.

This besides adds powerfulness to BigQuery, Google's serverless, highly scalable, and cost-effective multi-cloud information storage designed for business agility. As nan sanction implies, BigQuery is perfect for those who request to tally fast, SQL-like queries connected ample datasets.

The institution says this caller columnar capacity successful Spanner speeds up analytical queries by thing for illustration 200x connected unrecorded transactional data. With capacity for illustration that, we're talking instant responsiveness to real-time situations.

When building enterprise-based AI systems, you request agents to make decisions based connected existent data. Performing real-time actions based connected hallucinated information tin get disfigured very quickly. This is wherever RAG (retrieval augmented generation) comes in. Essentially, RAG combines ample connection models pinch real-time information access.

Also: 5 ways to beryllium a awesome AI supplier manager, according to business leaders

You tin commencement to spot really vectorizing hunt successful Spanner and BigQuery becomes basal erstwhile you're feeding successful real-time information on pinch humanities information. But getting vector hunt to activity efficiently has traditionally been painful. Google is adding adaptive filtering successful AlloyDB to automatically support vector indexes and optimize for accelerated queries connected unrecorded operational data.

Google is besides introducing autonomous vector embeddings and procreation to BigQuery, which automatically prepares and indexes multimodal information for vector search. This is simply a cardinal measurement successful creating a benignant of semantic representation for agents.

The institution is besides introducing nan expertise to tally AI queries correct wrong of BigQuery. This is, ahem, big. Now, BigQuery users tin person AI do its magic crossed elephantine gobs of system and unstructured data, inquire analyzable questions (including subjective ones for illustration "Which customers are frustrated?"), and get answers straight wrong existing analytics tools.

New agentic capabilities

In summation to building a instauration for agentic practice and information access, Google is announcing a bid of caller capabilities that embed agents successful their biggest information tools. Let's look astatine each successful turn.

Data engineering agent: Built specifically for information engineers, this supplier wrong BigQuery tin simplify and automate analyzable information pipelines. The full workflow tin beryllium driven by natural-language prompts, from information ingestion to transformations to data-quality appraisal to normalization.

Spanner migration agent: Related to nan information engineering agent, nan Spanner migration supplier tin simplify information migration from bequest systems to BigQuery. This benignant of migration is usually exceptionally tedious and perchance dangerous, but now nan supplier tin do astir of nan dense lifting.

Data subject agent: Data scientists attraction connected analyzing and interpreting analyzable data, while information engineers attraction connected information infrastructure. According to Google, nan caller information subject supplier "triggers full autonomous analytical workflows, including exploratory information analysis, information cleaning, featurization, machine-learning predictions, and overmuch more. It creates a plan, executes nan code, reasons astir nan results, and presents its findings, each while allowing you to supply feedback and collaborate successful sync."

Code interpreter: Built arsenic an enhancement of nan conversational analytics supplier introduced past year, nan codification expert takes successful business-analysis questions and converts them to Python codification to hole civilization study for users. This each runs wrong Google Data Cloud and uses nan Google Data Cloud information infrastructure. It besides includes an API disposable for developers to incorporated conversational analytics supplier and codification expert capabilities successful civilization code.

New command-line coding instrumentality

As portion of this large bid of announcements, Google is introducing an hold to Gemini CLI called Gemini CLI GitHub Actions.

CLI stands for bid statement interface, fundamentally a terminal interface to your computer. Even though astir users near nan terminal down erstwhile MS-DOS migrated to Windows, coders to this time make dense usage of nan bid line. Working successful terminal mode lets coders adhd devices and power nan coding process overmuch faster than erstwhile they person to find and prime items from menus and icons.

Also: Bad vibes: How an AI supplier coded its measurement to disaster

Last month, erstwhile Google introduced Gemini CLI, it fundamentally made nan features of nan Gemini chatbot disposable successful nan terminal. Now, Google has extended that capability, providing immoderate agentic features wrong nan terminal environment.

Some of you whitethorn beryllium wondering really this compares pinch Jules, nan Google coding supplier I wrote astir successful May. First, Jules useful successful a unafraid unreality VM, while Gemini CLI GitHub Actions runs successful terminal and integrates pinch GitHub Actions (the GitHub-based workflow tool).

Google says there's a reasonably constrictive scope to Gemini CLI GitHub Actions compared to Jules. Jules tin publication your full codebase, scheme and coming an attack to a coding challenge, and past execute connected it. Gemini CLI GitHub Actions is specifically targeted to intelligent rumor triage, accelerated pull-request reviews, and on-demand collaboration.

Also: Most developers usage AI successful their regular workflows - but they don't spot it, study finds

The issue-triage capacity helps coders negociate circumstantial bug reports and characteristic requests. Pull requests are nan measurement GitHub asks coders to corroborate integrating coding changes into branches and maestro codebases. On-demand collaboration is fundamentally mounting up a chat convention whenever you want to talk astir your code.

I could easy spot a programmer usage both. Jules would beryllium awesome for bigger projects and larger swings, and Gemini CLI GitHub Actions would activity good for quicker updates and fixes.

Are agents a game-changer?

What do you deliberation astir nan agentic displacement Google is promoting? Have you started integrating intelligent agents into your ain workflows? Which of Google's caller information devices aliases capabilities intrigues you astir -- nan information engineering agent, nan in-query AI reasoning, aliases thing else? Do you spot agents arsenic helping elder professionals, replacing inferior roles, aliases both? And really do you consciousness astir moving AI workflows straight successful BigQuery? Let america cognize successful nan comments below.

Also: AI agents will alteration activity and nine successful internet-sized ways, says AWS VP

Want much stories astir AI? Check out AI Leaderboard, our play newsletter.

You tin travel my day-to-day task updates connected societal media. Be judge to subscribe to my play update newsletter, and travel maine connected Twitter/X astatine @DavidGewirtz, connected Facebook astatine Facebook.com/DavidGewirtz, connected Instagram astatine Instagram.com/DavidGewirtz, connected Bluesky astatine @DavidGewirtz.com, and connected YouTube astatine YouTube.com/DavidGewirtzTV.