‘Agent is the atomic unit for AI’: Snowflake’s Baris Gultekin
In a conversation with YourStory, Snowflake's Baris Gultekin detailed how the company’s latest AI offerings are enhancing business intelligence for enterprises, and how the firm is bullish on India as a key market.
After generative AI, it’s undeniable that agentic systems are the next big trend for enterprises in 2025.
Known for its ability to perform complex, multi-step tasks in minutes, the effectiveness of an AI agent depends heavily on accurate ‘data retrieval’—the process of accessing structured and unstructured data from sources like databases, PDFs, and text documents.
As data sources grow, managing access securely and retrieving accurate information becomes challenging.
Addressing this AI-data mismatch is Snowflake, an American AI data-cloud firm which recently launched Cortex Agents—AI tools designed to efficiently retrieve and orchestrate both structured and unstructured data for business insights.
Since 2012, the San Mateo, California-based firm has been reshaping data management with its cloud-based data warehousing platform, enabling organisations to store, manage, and analyse large volumes of data.
With over 10,000 global customers, including many Fortune 500 companies, the firm is now expanding its capabilities into AI and advanced analytics, to help enterprises with secure data governance and intelligent insights.
In an interview with YourStory, Baris Gultekin, Head of AI at , detailed how the company’s latest AI offerings are enhancing business intelligence for enterprises, with strategic partnerships with Anthropic and DeepSeek driving its Cortex Agents, and how the firm is bullish on India as a key AI market.
YourStory: What is Snowflake's core vision for its latest AI announcement? Could you give us an overview of how customers are adopting these offerings?
Baris Gultekin: Snowflake is an AI data cloud. Our customers primarily use Snowflake as a data platform to secure, govern, and analyse their most valuable data assets. As AI becomes available, they increasingly seek to run AI directly alongside their data, eliminating the need for extensive data pipelines and improving efficiency.
Our strategy is to offer AI in an easy, efficient, and trusted manner, building on a strong data foundation. We have built a premier data platform to be able to analyse large amounts of data and govern it. Now that we have AI really easily accessible, our customers can do a lot more with that data.
We are seeing a lot of adoption across the board. Everyone’s building super exciting chatbots, where documentation that used to be hard to analyse is now suddenly available.
Secondly, there is a strong interest in using AI for business intelligence. Today, companies have to go through a long waiting time to unlock data that's already sitting in business accounts. For instance, a lot of our customers have dashboards that they look at to understand product trends, revenue. However, these dashboards tend to be relatively rigid. Once they want to dig into the next level of details, they need to go to an analyst and the analyst needs to write reports — that takes time and effort. Our product Cortex Analyst makes it very easy for businesses to just use natural language to query such business intelligence data.
YS: How does it enhance enterprise AI adoption, particularly given the challenges in convincing customers?
BG: We have an advantage because many customers initially need to determine which AI vendor to approve and procure AI from. A primary concern is ensuring that data remains in a secure, governed environment during transfers, as data is the asset that differentiates one business from another. As a trusted data platform already approved by over 10,000 customers, Snowflake provides an easier on-ramp to AI without requiring data movement.
Secondly, customers ask how to prevent AI hallucinations. We address this with state-of-the-art products that majorly focus on quality. In addition to providing access to the best models, we enable customers to ground these models with their own data.
Cost is, of course, the third pillar that matters to our customers. With ROI at the top of every customer’s mind, Snowflake delivers a strong ROI by reducing the total cost of ownership through AI running directly next to the data.
Initially, many customers begin with a proof of concept, where the focus is on achieving the desired quality rather than cost. Once the POC phase is completed, cost considerations become more prominent. At that point, customers optimise prompts and fine-tune models to ensure high ROI and throughput while maintaining quality.
YS: What makes Cortex Agents different from other agentic offerings in the market?
BG: There's a lot of excitement about agents and their capabilities. We started thinking of them as co-workers as capabilities increase. At Snowflake, we focus on a specialised type of agent known as ‘data agents’.
Data agents require access to and reasoning about data, yet scaling them poses several challenges. Accuracy is crucial for trust and security when handling sensitive data, and it is essential to govern this access with robust built-in controls.
Ensuring these capabilities are enterprise-ready is a top priority for our customers. They often require agents that handle both structured data—such as Snowflake or other tabular sources—and unstructured data from PDFs, repositories like Google or SharePoint, and other document-based sources.
For example, a brand and product planning agent might need to answer “What are my top-selling products?” by accessing sales records (structured data) in Snowflake, then determining why customers purchase them by analysing relevant unstructured content.
YS: Could you name some of the prominent customers using your latest offerings?
BG: We have been working with many customers on building these AI systems. For instance, Bayer—a global healthcare and pharmaceuticals company with over 100,000 employees—uses our AI products to create more flexible business intelligence analytics, enabling users to query and dig deeper into structured data.
Meanwhile, Siemens Energy leverages half a million pages of research documents to power internal chatbots, boosting team productivity. S&P Global, a financial data provider, is analysing around 200,000 quarterly earnings call transcripts and making those insights accessible to the financial industry.
What we see again is a mixture of unstructured data, structured data, a lot of focus on quality, and a lot of focus on making sure that the data stays where it stays from a governance perspective.
YS: With models like Anthropic’s Claude and DeepSeek-R1becoming increasingly sophisticated and capable of advanced reasoning, how do you see AI agents evolving within enterprises?
BG: All model providers agree that 2025 will be the year of AI agents. At Snowflake, we collaborate closely with leading AI providers like Anthropic, whose models support Cortex Analyst. Anthropic’s models are highly capable in both code generation and orchestration, and we integrate them directly into Snowflake to run alongside enterprise data. Cortex Agents leverage these models for orchestration and reasoning, enhancing AI-driven workflows. Some of our customers are interested in DeepSeek as well.
Ultimately, we believe choice is essential for our customers. By offering a selection of premier models alongside a scalable, enterprise-grade AI platform, we ensure they have access to the best solutions to their needs.
YS: What are the next planned enhancements for Cortex Agents in terms of scalability and functionality? How do you view India as a strategic market to drive these solutions?
BG: Firstly, we are making Cortex Analysts—our leading text-to-SQL solution—generally available. We are also extending Cortex Search to all customers, with significant improvements in both quality and scalability. Cortex Search supports over 100 million documents, and we have reduced its price by 30% to make it more accessible. By optimising the system, we have lowered operational costs and are passing those savings on to our customers.
Secondly, India has long been at the forefront of technology adoption. We see India as a strategic market for Snowflake and already have many valued customers there. With its young and dynamic market, I believe India will adopt AI faster than average.
YS: Traditional startups and SaaS companies are increasingly integrating AI into their operations and product offerings. How do you envision the evolution of agents over time, and what does the future of AI look like to you?
BG: It's a very exciting phase. We have thousands of customers already using our AI solutions, and see no end in terms of demand. A lot of our customers are getting started. Many of them already have production use cases with our solutions. Agents are indeed the future.
Each technological paradigm had an atomic unit — from web pages for the internet, applications for mobile...for AI, it is agents. We believe agents are the future. They are capable of calling other tools. They're capable of orchestrating (tasks) amongst themselves. I believe, starting from this year, we're going to see a lot more focus on agents.
These agents are increasingly capable of orchestrating difficult tasks, with data serving as their essential fuel. Running agents next to the data makes a big difference for our customers.
What differentiates Snowflake in this space is its commitment to governance, which is critical for enterprise customers. We offer an end-to-end solution that ensures fully governed access, operating directly alongside the data. Additionally, we emphasise accuracy, delivering highly precise retrieval results for both structured and unstructured data.
Many current agents focus only on document analysis, but our approach supports both document-based insights and business intelligence—a key factor for the next generation of agents.