A Note From Coco

Our office CoCo — four-legged, curly-haired, and endlessly enthusiastic — has always had a sixth sense for who’s worth trusting. When we introduced her to Snowflake’s Cortex Code (also nicknamed CoCo), she wagged. We’ll take that as a five-star review.

There’s a version of AI-assisted coding that saves you 10 minutes. Then there’s the version that saves you 10 weeks. Snowflake’s Cortex Code — affectionately nicknamed CoCo by its growing user base — is firmly in the second camp.

The Problem with Generic Coding Assistants

Most AI coding tools are like hiring a contractor who’s never seen your building. They’re technically skilled, but they don’t know which pipes are load-bearing, which data tables are business-critical, or why that one transformation costs three times what it should.

Every time you spin up a generic AI assistant on your Snowflake environment, you start from scratch — re-explaining schemas, re-introducing governance rules, re-contextualising your data model. The tool is capable. But it’s working blind.

Why it matters

Analysts estimate that 85% of AI initiatives fail to move beyond the pilot phase — not because the AI is bad, but because it lacks the governed, cross-platform context that production systems demand.

What Makes CoCo Different

Snowflake’s Cortex Code knows your stack the way your best engineer does — without the onboarding period. Launched in November 2025 and significantly expanded in February 2026, CoCo is a data-native AI coding agent built to understand your entire enterprise data context: schemas, governance rules, compute constraints, sensitivity labels, operational semantics, and production workflows.

Ask CoCo to build a pipeline. It doesn’t just generate SQL — it generates SQL that respects your specific tables, your governance policies, and the downstream dashboards that depend on the columns it’s touching.

What CoCo Can Actually Do

This isn’t a list of features — it’s a list of outcomes. Here’s what happens when your data team gets CoCo in their corner:

Build Pipelines in Plain English

• Translate natural language into production-ready data pipelines, ML workflows, and AI agents — no boilerplate, no re-explaining context.

Debug With Context

• CoCo understands why transformations fail and suggests fixes aligned to your actual schemas and operational patterns.

Admin at the Speed of Thought

• Manage permissions, create users, set policies, and optimise costs — all through conversational commands.

Works Where You Work

• Available in Snowsight and as a CLI in VS Code, Cursor, or your terminal. dbt and Apache Airflow® now supported too.

Governed by Design

• CoCo never leaves your Snowflake perimeter. RBAC controls govern every action. Your data doesn’t move, your policies stay intact.

Multi-Model Flexibility

• Choose your AI backbone — Claude Opus 4.6, Claude Sonnet 4.5, or GPT — balancing quality, latency, and cost per workload.

The CoCo Effect in the Real World

At a leading global marketing and communications group, CoCo aligned naturally with existing team workflows, letting engineers translate evolving data requirements into AI-powered solutions without disrupting what was already working.

At a company that delivers solutions centred around strategy, security, scalability and cloud-first technologies, a team used CoCo to migrate complex Talend XML jobs into Snowflake’s Bronze/Silver/Gold architecture — work that would normally require weeks of painstaking manual rewrite. With CoCo’s deep schema awareness, the job became an AI-assisted sprint. The result? Over 500 hours saved and roughly $100,000 in value — in just the first 20 days.

The shift that matters

CoCo doesn’t just accelerate tasks — it moves teams from reactive troubleshooting to intelligent remediation. From manual engineering cycles to AI-augmented delivery. That’s not incremental productivity. That’s delivery transformation.

Two CoCos. One Philosophy.

Our office CoCo, Chief Happiness Officer, professional lap-warmer, and the best context reader in any room — has always known something the rest of us are just catching up to: the best way to help someone is to truly understand their world first.

She doesn’t bark at every stranger. She reads the room, picks up on cues the rest of us miss, and shows up exactly where she’s needed. No re-introduction required.

Snowflake’s Cortex Code works the same way. It doesn’t parachute in with generic answers. It reads your data environment — your tables, your governance, your production pipelines — and shows up already knowing what matters. No re-explaining. No blank-slate context dumps. Just context-aware intelligence, right where you need it.

Two CoCos. Both curiously intelligent. Both excellent at understanding their environment before they act.

The Bottom Line

The data engineering bottleneck has never been about raw capability. Teams know how to write SQL. They know how to build pipelines. What slows them down is context — the constant re-explaining, the manual lookup, the gap between what an AI suggests and what your production environment will actually accept.

Cortex Code closes that gap. It’s the coding agent that reads your data like it’s been on your team for years — and ships production-ready code to prove it.

And if our office CoCo — paws up, tongue out, tail wagging — approves? That’s all the endorsement we need. 

Snowflake, Cortex Code, and related marks are trademarks of Snowflake Inc.