MotherDuck Corp., the maker of a cloud-native data warehouse based on the open-source DuckDB analytical engine, is betting that artificial intelligence agents will reshape how data pipelines are built and managed.

Today, it unveiled a new capability called Flights that enables users to create and operate data ingestion workflows through natural-language interactions with AI assistants. The new service extends MotherDuck’s cloud-based analytics platform by allowing AI agents such as ChatGPT, Claude and Gemini to create, schedule and manage data pipelines from within a Python runtime environment. The company said Flights enables organizations to move from raw data ingestion to analytics and visualization within a single AI-driven workflow.

“AI has come for the data world,” said Jordan Tigani, MotherDuck’s co-founder and chief executive. “There’s a big opportunity to take things that were traditionally manual tasks and use prompting and the English language instead of having to hand-code everything.”

DuckDB is an analytical database designed for fast processing of structured data without the complexity of managing a traditional database server. Sometimes described as “SQLite for analytics,” it is known for its speed at querying large datasets directly from files such as CSV, Parquet and JSON, making it popular for data science, machine learning and business intelligence workloads.

Its small footprint enables DuckDB to be embedded within applications and to run on laptops or as cloud services, rather than on a separate server. It’s particularly popular in small and midsized organizations.

Flights is designed to address the longstanding analytics challenge of moving data from source systems into a form that can be analyzed. Organizations often rely on specialized tools such as Fivetran, Estuary or dbt, combined with a lot of manual configuration and custom code. MotherDuck says AI agents can automate much of that work.

For example, users can instruct an AI assistant to connect to sources such as customer relationship management applications, databases and application programming interfaces with the agent generating the code needed to ingest, transform and schedule the data pipeline. If a job fails, the agent can inspect logs and attempt repairs.

Room to experiment

Tigani said agent-created pipelines differ from traditional data engineering workflows because agents are more likely to experiment, investigate multiple options simultaneously and create temporary pipelines for one-off tasks.

“A human will build pipelines that tend to be heavyweight, static things,” he said. “An agent may decide that for a specific task it wants to spin up a pipeline, import some data and investigate a problem automatically.”

Flights is integrated with the Model Context Protocol emerging standard for connecting AI agents with external systems. Through MotherDuck’s MCP server, agents can access tools to create, schedule, update and monitor data pipelines.

Tigani said MCP could be as important to AI as application programming interfaces were to cloud development because the protocol allows non-technical users to interact with enterprise systems through conversational interfaces.

Flights also advances MotherDuck’s broader strategy of collapsing multiple layers of the analytics stack into a single platform. Earlier this year, the company launched Dives, a feature that uses AI to create dashboards and applications. The combination of Flights and Dives creates the potential for workflows in which agents ingest data, analyze it and generate visualizations without requiring separate integration, orchestration and business intelligence products.

“We’re trying to collapse them into one,” Tigani said. “People don’t really care so much about how they get the data in. They just need to get it.”

Tigani said he envisions fleets of AI agents managing data environments, monitoring changes and generating insights with minimal human intervention. Data engineers will still play a role, he said, but increasingly as supervisors of automated systems rather than builders of pipelines by hand.

Flights is priced based on runtime consumption, starting at about 60 cents per hour, although Tigani said the primary goal is not revenue generation but making it easier for customers to bring data into the platform.

MotherDuck’s cloud service currently has about 850 paying customers after 18 months of commercial operation. The company has raised $100 million in financing.

Photo: Flickr CC