You may have more answers to your marketing questions sitting in your data warehouse than you realize. But how do you get the data that's already there into more people's hands, in the moment they need it, without requiring everyone to become a SQL expert first?
Google has reframed BigQuery as the foundation for agentic AI, addressing this data access challenge head-on. Instead of treating natural language and unstructured data as an add-on to a traditional warehouse, BigQuery is building those capabilities into itself. That way, BigQuery becomes the foundation that gets more people to an answer, in the format they actually work in, without asking them to learn new syntax or wait on someone else's schedule.
Here's what's changed inside BigQuery itself, and what it means for how your team gets from question to answer.
Conversational analytics lets both business and technical users ask questions about their data in plain language and get back answers, visualizations, and even multi-step analysis, without writing a line of SQL.
Analysts have always been able to query the data warehouse directly, but marketers used to have to wait in a queue, file requests, and hope for a dashboard update. Now with conversational analytics, someone on a marketing or operations team can ask a direct business question, such as which properties are underperforming or how a metric has shifted since last quarter, and get a grounded answer pulled straight from BigQuery.
BigQuery conversational agents rely on user-selected BigQuery sources (tables, views, and UDFs) and utilize provided guidelines in the form of instructions, verified queries, and glossary.
Agent configuration is what makes the difference between an agent you can trust and one that just sounds confident. The agent isn't guessing at what your terms mean, because your team has already told it. That upfront work is what turns a conversational agent into something closer to a well-trained analyst than a chatbot that occasionally gets close.
Grounding an agent in your tables and glossary is one thing when that data is neatly structured. But a lot of what teams want to ask about doesn't live in rows and columns. Reviews, photos, call recordings, and support tickets all carry answers, too, and until recently, getting an agent to reach them meant building a separate pipeline just to translate that content into something BigQuery could use.
BigQuery has always been synonymous with rows and columns, but that's not the full picture. Object tables let you reference files stored in Cloud Storage, including audio, video, and images, directly from within BigQuery. The files stay where they are, while BigQuery provides a structured way to access, query, and govern them alongside the rest of your data.
Analysis that used to require a separate pipeline built specifically for image or audio processing is now a query away from the rest of your warehouse. And because conversational analytics agents can pull from both structured and unstructured sources in the same conversation, a single question can span a reservation table and a folder of guest photos without the person asking it needing to know the difference.
In practice
Twiddy, a vacation rental company in North Carolina, had thousands of property photos and no efficient way to know which ones actually helped a listing get booked. Working with Adswerve, Twiddy combined their structured booking data with property photos inside BigQuery, using BigQuery ML to predict how well a given image would perform. That gave their team a data-driven way to make photo editorial decisions, and they've been able to do it using the SQL skills they already had in house rather than standing up a separate machine learning function.
Structured or not, none of this data does much good sitting still. Let’s dive into how to put that same data to work directly inside the queries your team already writes.
Somewhere between simple SQL and a full conversational agent sits a set of AI functions now built directly into BigQuery. Traditional SQL is great at counting, averaging, and grouping. It has never been great at understanding the semantics. Now you can run a standard GROUP BY on a table of customer reviews and layer in a function that generates a summary of what those reviews say, alongside the usual counts and averages.
Your existing SQL has new capabilities like being able to generate text, detect sentiment, and filter results with natural-language conditions, all within a query you already know how to write. Teams with deep SQL expertise get access to analysis that would have previously required a separate machine learning pipeline, while those without that depth get the same lower barrier to entry that makes conversational agents so useful, without ever leaving familiar syntax.
Getting data into BigQuery has traditionally meant either building custom pipelines or paying a third-party service by the row to move data from platforms like Facebook, LinkedIn, or Salesforce. But now, Google is building native connectors for exactly these kinds of sources, including Adobe, HubSpot, and Stripe, which means teams can set up a direct connection without adding another vendor to the stack.
Native connectors don't just simplify the technical setup. Paid data-transfer tools often charge based on volume, and costs can climb quickly for teams pulling from several platforms at once. With more native connector options, you can get a more complete picture of your marketing performance without incurring additional data transfer costs.
The last capability that rounds out this picture is Gemini's role in writing SQL itself. Think of it as an intelligent autocomplete. You describe what you want in plain language, and Gemini generates a suggested query you can run as is or refine further, with the underlying SQL fully visible.
This benefits people at both ends of the skill spectrum. Someone newer to SQL gets a working starting point instead of a blank query editor. An experienced data analyst gets a faster way to draft complex queries, then spends their time refining logic instead of writing boilerplate from scratch.
Taken individually, these updates each solve a specific problem. Taken together, they explain why Google is positioning BigQuery as the foundation for agentic AI rather than just a place to store data. Conversational agents need grounded, well-structured data to work from. Object tables and AI functions extend that grounding to formats SQL was never built to handle. Native connectors and Gemini's SQL generation make it easier for more people to get value out of what's already there.
If your team is exploring what this looks like in practice, from setting up a conversational analytics agent to auditing what's currently getting in the way of clean, connected data, we'd love to talk through what's possible for your BigQuery environment.