GenAI & Agents

Your AI agents have never met. Introduce them.

May 18, 2026

3 min read / Stephen Wagner

From Shadow AI to Shadow Agents
The Hidden Cost of Autonomy
Why Traditional Governance Breaks Down
Visibility Is the Missing Primitive
Human-in-the-Loop as a Design Choice
The Real Shift: From Limiting AI to Seeing It

It's Monday morning, and your supply chain lead drops a question into the team channel: "Severe weather hit Florida last week — what's it going to do to our Southeast forecast this quarter?"

The answer to that question doesn't live in one place. The sales history is in Snowflake, where Cortex Analyst can already translate plain English into trustworthy SQL. The weather impact is on the open web, where a search agent can pull and cite sources. The forecast model lives in Databricks, where Genie can interrogate it directly. Three best-in-class agents, three different vendors, three governance regimes — and one question that needs all of them to answer it together.

A year ago, stitching that together meant bespoke Python, brittle API glue, and a security review that outlasted the question itself. Dataiku external agents change the math. With external agents, teams can orchestrate best-in-class AI and data services, from Snowflake Cortex to AWS Bedrock, natively within the governed, collaborative environment of Dataiku, the Platform for AI Success.

Introducing AI agents thumbnail Dataiku blog banner

Breaking down silos: what are external agents?

Modern data systems are distributed: data in Snowflake, ML on Databricks or AWS, productivity tools via Google. Traditionally, using AI agents meant tab-switching through multiple proprietary interfaces, leading to fragmented workflows and inconsistent governance.

Dataiku breaks new ground as the first governed orchestration layer for external agents. By uniquely treating third-party agents as first-class citizens, we empower teams to build and command complex agentic systems with unmatched enterprise oversight.

Extending capabilities: the power of a connected ecosystem

Integrating third-party agents within Dataiku is more than API access; it's an expansion of your AI's operational footprint. With our latest integrations, external agents act as force multipliers across the data workflow:

Data specialists (Snowflake & Databricks):

Snowflake Cortex Agents & Tools: Connect securely to Snowflake-managed agents. Leverage Cortex Analyst for accurate Text-to-SQL conversion using enterprise semantic models, or use Cortex Search to surface documents from Snowflake stages.
Databricks Agents & Genie Tool: Use Genie AI to deliver instant, SQL-powered answers to structured business queries, harnessing data stored in your Databricks Lakehouse.

Reasoning powerhouses (AWS & Google):

AWS Bedrock Agents: Tap real-time streaming and citation extraction from Knowledge Bases, ideal for retrieval-augmented generation (RAG) with audit trails.
Google Vertex AI: Access Google's Agent Development Kit (ADK) to connect Dataiku to secured, enterprise-grade agents.

Practical-example-orchestrating multi-agent-systems

Consider our supply chain manager from earlier who needs to analyze: "How did severe weather in Florida last week impact this quarter's Southeast sales forecast?" Traditionally, this complex correlation required manual research, SQL expertise, and bespoke ETL work.

With Dataiku external agents, this workflow is automated:

1. Instantiate agents: Use Dataiku's agents and GenAI models to create a parent agent.

2. Add tools: Add relevant tools, such as Google Search and Cortex Analyst.

3. Configure strategy: Instruct the parent agent to reason across both search and structured data.

4. Orchestrate: Combine Cortex Analyst to query sales databases and Google Search tools to extract weather impacts.

In seconds, your Dataiku agent executes the multi-step logic: extracting relevant dates, querying live sales data, correlating findings, and presenting a synthesized, source-cited answer.

Why it matters: efficiency, scalability, governance

Moving beyond isolated AI tools to a connected, multi-agent architecture is a strategic imperative for modern enterprises. To achieve this, organizations need the Platform for AI Success, an environment that brings people together to seamlessly orchestrate complex workflows while enforcing enterprise-grade governance.

By centralizing these powerful third-party services within Dataiku, teams transform fragmented workflows into a cohesive, secure ecosystem built to scale.

External agents unlock:

No-code orchestration: Remove the need for bespoke Python scripts. Dataiku manages authentication, API handling, and audit trails natively.
Unified governance: Regardless of where execution occurs, from Bedrock to Databricks, every agent is registered in Dataiku's GenAI Registry, integrating seamlessly with Dataiku Govern for oversight, audit, and access control.
Flexible distribution: Deploy external agents into batch workflows, chat applications, or APIs — wherever your AI needs to operate.

The collaborative future is here

Standalone AI agents are relics of the past. Tomorrow's value lies in interoperable, agentic systems that can reason, act, and integrate seamlessly with your data landscape. Dataiku external agents empower teams to avoid lock-in, plug in the best specialized AI services, and maintain enterprise-grade governance and collaboration, always.