Introduction
The End Goal: From BI to AI
BI has always been considered the End Goal of the Data Warehousing and Analytics ecosystem or as we fondly call it the ‘Modern Data Stack’. It has served a rightful place in the architecture serving the business users with visual descriptive insights that often leads to a diagnostic analysis.
But with AI (read GenAI) evolving at a breakneck pace, the end goal or rather the last mile of data consumption is also moving towards it, and for a very good reason. To put it simply, AI can do everything that BI does, and more!
From ‘Conversational Analytics to ‘Actionizing the Data’
It all started with the humble ‘Talk to the Data’ use case where end users could ask a question in natural language and LLMs would fetch the answer from the structured (using Text to SQL) or unstructured (using RAG) data.
But as LLMs became smarter, the next thing we knew was LLMs doing ‘Reasoning’, LLMs performing ‘Orchestration’ & LLMs taking ‘Actions’. All of this led to what we now know as ‘AI Agents’.
Introducing AI Agents in Snowflake and Snowflake Intelligence
Agentic AI refers to AI systems that can autonomously make decisions, plan, and act to achieve goals with minimal human intervention. Think of it like a persona-based or skill-based job that requires a bunch of tasks to be orchestrated & executed.
Snowflake as a platform has evolved very quickly in the Agentic AI space. With availability of top-of-the-line LLM models (like OpenAI GPT-5, Anthropic Claude Sonnet 4.5 and many others), it’s now easier than ever to build Agents that can reason and make plans swiftly. These AI Agents can also take actions like retrieving data from the structured tables using Cortex Analyst. Or extracting insights from unstructured data using Cortex Search. The actions or the tool calling can also be extended to sending emails, calling REST API’s or even executing scripts in the Snowflake environment. See the architecture below to understand how reasoning, planning and action come together natively inside Snowflake.
AI Agents can be exposed to the end users as an API (to be further integrated into their chat platform of choice, like Slack) or a Streamlit based Native App. And if the end users are looking for a ChatGPT like user interface, Snowflake Intelligence will fit the bill.
Snowflake Intelligence Architecture
All of this can now be built in your Secured and Governed Snowflake environment with all data privacy intact. Snowflake also provides a tight layer of Observability & Explainability (via TruLens) along with guardrails (via Cortex Guard) that brings trust and control in AI.
Serving the Last Mile (of Data Consumption)
Delivering the Last Mile is a common phrase in the logistics ecosystem, but it’s not too far from the data and analytics world. After years of building pipelines, warehouses, and dashboards, the last mile of data consumption remains the hardest. It’s getting actionable insights into the hands of every decision-maker.
There are many ways of serving good quality data, but it depends on the consumers on how they would like it to be served. Business Users may prefer KPI’s and dashboards. Field Ops may prefer voice conversations. Customer Service Representatives may prefer to chat with their data, and so on. The goal of the ‘Last Mile Delivery’ is to meet the consumers where they are, and that’s what Snowflake Intelligence, as a wrapper on top of persona-based AI Agents, does best!
Before LLMs became mainstream or AI Agents were a thing, the world still worked just fine (or at least we thought so). We just didn’t know what we were missing out on until we got to touch and feel it. Enterprises were using chatbots for general-purpose Q&A; Search Engines were helping with unstructured data lookup; BI tools were helping represent the visuals; and reasoning/planning was left to the humans. Never did we think of all of this under one roof, consolidated, standardized, and delivered with human-like context-specific, subjective responses.
Snowflake Intelligence brings it to life with a simplified user experience of a chat interface on top of persona-based AI Agents that can be extended to visualize business KPI’s, review tabular data, download relevant information, send emails, or take additional actions. All of this in one unified user interface and governed with Snowflake’s market-leading capabilities. Finding the relevant data, querying it, analysing the response & contextualizing it based on the question was never this easy!
By Industry Domain, Persona & Use Cases
A marketing operations manager may want to know which channels or campaigns they are spending more (Descriptive Analytics). Or why a specific channel or campaign is giving better ROAS (return on ad-spend) than the other (Diagnostic Analytics).
A manufacturing plant manager may want to know when the next predicted maintenance cycle for the critical equipment would be, so the downtime can be planned effectively well in advance (Predictive Analytics).
A sales representative may want to know the Next-Best-Action (as a recommendation) for a client for further cross-selling or up-selling opportunities (Prescriptive Analysis).
Show me the trend of sales by product category between June and August (Descriptive Analysis)
Why did sales of Fitness Wear grow so much in July? (Diagnostic Analysis)
Imagine the vast variety of jobs that each persona needs to perform, and multiply that by all the Business Functions in the org. Now add the Industry Domain knowledge and the skillset expertise on top of it. The combinations seem to be endless!
With Snowflake Intelligence, you can build as many AI Agents as needed and expose them all to the relevant users, and let the users choose the right Agent for the right job in the runtime.
Getting Started with Snowflake Intelligence
Building & Deploying AI Agents on Snowflake Intelligence is super simple, just like everything else in Snowflake, “It just works!”. This quickstart demonstrates how, with a few quick steps, you can orchestrate Cortex Analyst, Cortex Search, and tool calling to empower users to ‘Talk & Reason with their Data’ and actionize the next steps.
Prerequisites
It all starts with good quality data. Make sure your data is accurate, complete, consistent, timely, unique, and relevant. Set up the right data & user security with Tokenization, Encryption, RBAC, etc. Lastly, gather some ‘ground truth data’ that you can use to test the Agent response.
High-Level Steps
- Set up Snowflake Intelligence
- Load the data and define the Semantic View/Model
- Create an Agent
- Add Custom Tools
- Start using it
Considerations
- First and foremost, work with the end users directly and learn how they would like to consume the data. Understand their functional and non-functional requirements
- Start with low effort & high impact items first
- Build observability & explainability to gain trust
- Leverage external knowledge (like Cortex Knowledge Extension) to enrich the insights
‘Cross the Chasm’ with Snowflake Intelligence
Even the best technology fails without adoption; that’s where most AI initiatives fall into the chasm. Make Snowflake Intelligence work for your business and deliver the right business impact by crossing the Adoption chasm.
Technology Adoption Lifecycle
Most technologies get stuck in the Chasm between the Early Adopters and the Early Majority, as the pragmatists don’t see the real ROI. To ensure you cross the chasm with ease, make sure you have your Pilot or Minimal Viable Product (MVP) up and running and get some prima facie feedback before you plan to scale up or scale out. At this point, you can think of adding more Agents to serve additional jobs for the same persona. Or add a new set of Agents to serve a completely new persona in the org. Or make the existing agents better with additional data for better insights.
This is also a good time to consider monitoring and optimizing for Cost, Performance, Security, and other non-functional aspects.
The power of Snowflake Intelligence can truly be unleashed only if it can drive the right business value for your business. With Reasoning, Planning, Orchestration & Tool Calling, these AI Agents can create a lasting impact for your organization in many ways. To keep it simple, you can generate revenue, save costs, or safeguard your org with Snowflake Intelligence.
Snowflake & Tredence: A Winning Partnership
Turning AI potential into enterprise outcomes requires deep domain expertise and strong platform partnerships; that’s where Tredence and Snowflake create real impact. Our winning combination of domain & technology expertise allows us to understand the business needs better and to be able to build AI Agents, keeping the persona in mind. We are a Snowflake ELITE Services partner and recognized as the 2025 Partner of the Year.
Having worked with numerous Fortune 500 clients and solved complex Data & AI needs for 8 of the top 10 retailers in North America, we deeply understand enterprise data. This helps us build Semantic Models, which act as a backbone for AI Agents. Our dedicated AIOps practice makes sure you can trust AI and scale it to your needs.
For a leading North American retailer, Tredence used Snowflake Intelligence to build & deploy AI Agents that reduced their Sales & Operations planning cycles by 40%, all within governed Snowflake data.
Tredence can help at every stage of the AI journey, helping clients discover & prioritize use cases, build & deploy MVP Agents, scale out the AI Agents, and govern the whole AI ecosystem.
Reach out to us for a free AI maturity assessment and connect with us to see how Snowflake Intelligence can redefine your enterprise AI journey.
Conclusion
Snowflake Intelligence and AI Agents are here to stay. The momentum is undeniable!
As MCP’s become generally available within Snowflake and LangGraph becomes mainstream, it’s not just enabling AI, it’s redefining how enterprises reason, plan, and act with data. We can perform more complex orchestrations and practically call any tool of choice to perform actions. All of this makes it possible for the end users to use Snowflake as the platform of choice and Snowflake Intelligence as the tool to perform the last-mile tasks & jobs without a swivel chair.
AUTHOR - FOLLOW
Sumit Bhatia
Snowflake Field CTO, Tredence
Next Topic
Telco Data Monetization with Data Clean Rooms: Secure Analytics & New Revenue Streams
Next Topic
