Snowflake has been actively expanding its artificial intelligence offering for two years now. The AI Data Cloud, which already supports numerous workloads, can also serve as a valuable base for this type of application. We spoke with Head of AI Baris Gultekin about Snowflake’s goal of delivering comprehensive yet simple AI solutions for businesses.
Gultekin sees the AI Data Cloud as the platform that companies cannot ignore if they are seeking the core values of simplicity, efficiency, and trust. “For everything we have built, we are essentially building a single, uniform platform. When AI runs alongside data, it is much easier to build these systems. And when it comes to efficiency, it’s all about quality. We differentiate ourselves by having the world’s best retrieval engine for both structured and unstructured data. Finally, trust is essential in AI,” says Gultekin. Snowflake offers governance options to ensure trust. He cites that Snowflake is the only platform that can run OpenAI and Anthropic together within the same secure environment. It is possible to apply access controls to all types of data.
By focusing on these three characteristics, Snowflake has won the hearts of companies during its active involvement in AI workloads. The number of companies working with AI in the AI Data Cloud every week has now passed 5,000. According to Gultekin, this demonstrates the rapid progress of AI adoption within the platform. Often, it’s about getting a few large companies excited about being the first to use an AI feature, gathering feedback, and then using that optimization and enthusiasm to drive further adoption.
From building blocks to mass adoption
The Snowflake Summit 2023 was the official kick-off for targeting AI workloads. That’s when Snowpark Container Services came out, making it possible to fine-tune and run LLMs. At the Summit a year later, the building blocks for AI emerged, including Cortex Analyst for structured data and Cortex Search for unstructured data.
These building blocks formed the basis for many customers to experiment with AI. “We went from getting the building blocks ready last year to a huge group of customers using AI,” said Gultekin. Companies that have invested in building their data platform in recent years are now reaping the benefits of that preparation.
Structured and unstructured data together
An important insight that customers share is that AI requires access to a significant amount of data, says Gultekin. This immediately raises the question of how data can be secured and managed. It’s not just about structured data, but also unstructured data that is now becoming accessible to AI applications.
According to Gultekin, many customers refer to AI as “the wild west” within their companies. Many proof-of-concepts are emerging, particularly around unstructured data that has become increasingly valuable. Organizations, therefore, need to consolidate all this data in a unified data platform. “Many of our customers who have invested in their data in recent years and prepared it have benefited directly from being able to build on it via that data platform easily,” says Gultekin about the transition to AI.
Tip: Snowflake makes AI Data Cloud the brain of any business
From reactive to proactive AI
The latest development is the shift to agentic AI. Unlike generative AI, which is primarily reactive, agentic AI is proactive. These agents can plan steps, make decisions, and intelligently use multiple data sources. “The amount of time that AI can automate in a single step increases linearly,” Gultekin observes. At the same time, agentic AI systems are often more complex to set up than other AI systems. Organizations need to monitor latency and costs, but also keep an eye on accuracy. If something goes wrong at the beginning of a process, it is difficult to recover later, warns Gultekin.
With Snowflake Intelligence, the company is also taking an important step toward agentic AI that is more widely applicable within organizations. Where traditional automation and chatbots are often limited to fixed tasks, Snowflake Intelligence enables companies to build full-fledged agents that autonomously generate insights, support decisions, and initiate actions. This all happens within the familiar Snowflake environment, where existing governance and security are automatically applied. Gultekin emphasizes that the platform breaks down data silos and enables non-technical users to work with their data using natural language. Thanks to integrations with Box, Google Drive, and Workday, data agents can extract insights from structured and unstructured sources. By also incorporating external knowledge via Cortex Knowledge Extensions, users get context in addition to answers, essential for acting faster and smarter based on complete business information.
New products
Two new products are set to accelerate AI adoption in the coming period further. Snowflake Intelligence, as described above, functions as a research assistant for analysts, providing agents with access to both structured and unstructured data. AISQL, in turn, integrates AI directly into the SQL engine, simplifying complex tasks.
Gultekin describes an example of something that is now possible as “AI join.” Normally, an analyst would have to link two data sources manually. With AI join, this can be done using natural language. For example, a portfolio manager might ask, “Show me all the news stories my portfolio should be concerned about,” and the system would automatically extract relevant information from news sources. This was once much more challenging.
This development confirms the assertion that companies are increasingly building agents and AI in general. According to Gultekin, Snowflake Intelligence will be the key product for the AI-filled future because it enables organizations to develop high-quality agents with access to all business data quickly.