By Rakesh Reddy
Introduction
This marks our fifth Snowflake Summit. Each year we leave more excited than the last. The pace keeps accelerating, the innovation keeps getting bolder, and the momentum behind this platform shows no sign of slowing down. Each Summit raises the bar on what we can build for our customers, and 2026 cleared it by a wide margin.
This year Snowflake said it best: “models shift, data stays, govern both”. That line captures the whole moment. It was the year the agentic enterprise stopped being a vision slide and started shipping as a product, with the platform now reaching from how data lands, to how it is governed, all the way to the agents that act on it.
Here is our read on what was announced, grouped the way we think about it for customers, with our take on why each piece matters.
New names, more powerful AI
The headline is two agents with familiar DNA and new names.
- Snowflake CoWork (formerly Snowflake Intelligence) is the personal agent for knowledge workers. It is the place you go to ask a complex question and get an answer you can act on.
- Snowflake CoCo (formerly Cortex Code) is the coding agent for builders. It is where you build faster, natively against Snowflake data. Same context, same governance, different jobs. That pairing is the spine of everything below.
The knowledge worker gets a new front door
For years the answer to “I need to understand this data” was a dashboard someone else built. CoWork changes the default.
CoWork is a personal work agent that reasons through data, delivers briefs, and recommends actions. Cortex Sense lets it auto-learn the schemas, metrics, and business context across your estate, so it understands your data without a long setup project. It takes action without making you switch tools: draft an email, update a ticket, post to Slack, all from one governed conversation. And it works proactively, with automations that surface risks and insights before anyone asks.
Reusable Skills and governed agents are what let this replace static BI rather than sit beside it. A skill built once becomes something the whole organization can run, under the same policies as the underlying data. For the line-of-business leaders, CDOs, and CIOs who own AI productivity, the real shift is from static reporting to taking action on your data with agents.
The agents reach the tools teams already live in
An agent that can only act inside Snowflake is interesting. An agent that can act across the rest of your stack is useful.
The new hosted MCP connectors close that gap. Snowflake now offers a hosted MCP server for any Snowflake agent and a hosted MCP client, so Cortex agents can reach external tools securely. Governance comes with it: progressive OAuth that limits exposure, Snowflake’s native role-based access controls applied to external actions, and a full audit log of every action the agent takes.
This is the detail that makes agentic automation safe to put in front of an enterprise. The agent does real work in the tools people use, and security stays in one place.
A faster, AI-native path for builders
If CoWork is the front door, CoCo is the workshop, and it grew up at Summit.
CoCo Desktop brings full agentic development to a native IDE on Windows and macOS, with an always-on agent on your local machine and scheduled tasks for repeatable work. It ships with a VS Code extension, a Claude Code plugin, a CoCo Slackbot, and a Skill Catalog for sharing what you build. MCP integrations and shareable skills mean a good pattern spreads instead of getting rebuilt.
The promise for engineering teams is concrete: ship production-ready pipelines, apps, and agents faster, modernize Snowpark, dbt, and Streamlit work at a fraction of the old effort, and scale your best engineers’ expertise to every team. AI-native is now the standard for how builders work, not a side experiment.
The AI-ready foundation got stronger underneath all of it
None of the agent story holds up without the platform underneath, and Snowflake announced new features that make that platform even more powerful:
- Datastream brings Kafka-compatible streaming with Snowflake simplicity. It lands data continuously into Snowflake native or Iceberg tables, secures it on arrival with existing Horizon policies, and connects an existing Kafka ecosystem with zero code changes. You can stand up a production pipeline in minutes using CoCo, and retire a separate streaming system in the process.
- Horizon Context is a governed semantic foundation for AI, BI, and apps. It collects context from systems inside and outside Snowflake, including PostgreSQL, SQL Server, Tableau, Power BI, and dbt, enriches it with automated metadata while keeping humans in the loop, and activates it through CoCo. The payoff is trusted answers built on shared semantic definitions, query logs, schemas, and lineage rather than guesswork.
- Interoperable Lakehouse lets any engine work on a single governed copy of your data. With Iceberg V3, the Horizon Catalog, and Snowflake storage for Apache Iceberg, you read and write Iceberg tables from any engine while policies persist across them. That means no lock-in, no data movement, and one source of truth for AI to work from.
- Adaptive Compute is fully managed, workload-aware compute that tunes itself. It removes the manual sizing decisions that quietly drive up cost, and the performance numbers are real: 1.6x faster analytics, 2.2x more queries per hour, and 3.5x faster DML operations compared to the first generation.
- Trust Center: AI Security makes protection built-in rather than bolted on. Secure guardrails block prompt injection and govern agent actions, a single control pane manages your security posture, and automated policies defend against ransomware and data exfiltration. Compliance work that used to take days can move to seconds, with threat detection running around the clock.
What we are doing with it
Let’s say the quiet part out loud: when the platform gets this much better at building and governing, the easy version of systems integration gets commoditized. Generating code is no longer the hard part. The hard part is making the result trustworthy and tying it to a business outcome the customer will fund again. That is where we have leaned in.
It is why we built AuditIQ, our automated validation and governance layer that sits on top of what CoCo produces. CoCo accelerates the build. AuditIQ is how we prove the build is production-grade rather than a throwaway. Together they let us hand a customer something they can trust, not just something that runs.
This is also the thinking behind coThrive, our evolved investment program. Instead of selling an open-ended assessment, we invest alongside a customer to prove one high-value use case on their own Snowflake data, built with CoCo and validated with AuditIQ, with a clear line of sight to the larger initiative it unlocks. Prove one real outcome, then scale it into the bigger play. Snowflake’s improvements launched at Summit 2026 is what makes that possible. .
About kipi.ai
Kipi.ai, part of Capgemini, is a global leader in data modernization and democratization focused on the Snowflake platform. Headquartered in Houston, Texas, Kipi.ai enables enterprises to unlock the full value of their data through strategy, implementation and managed services across data engineering, AI-powered analytics and data science.
As a Snowflake Elite Partner, Kipi.ai has one of the world’s largest pools of Snowflake-certified talent—over 600 SnowPro certifications—and a portfolio of 250+ proprietary accelerators, applications and AI-driven solutions. These tools enable secure, scalable and actionable data insights across every level of the enterprise. Serving clients across banking and financial services, insurance, healthcare and life sciences, manufacturing, retail and CPG, and hi-tech and professional services, Kipi.ai combines deep domain excellence with AI innovation and human ingenuity to co-create smarter businesses. As a part of Capgemini, Kipi.ai brings global scale and execution strength to accelerate Snowflake-powered transformation world-wide.
For more information, visit www.kipi.ai.
