Accelerate Data Team Onboarding and Productivity

Introduction to Keboola AI Copilot

Keboola's AI Copilot, known as Kai (Kaboola AI), is designed to elevate the productivity of data teams. By leveraging foundation models and purpose-built tools, Kai assists, augments, and amplifies team capabilities within the Keboola platform. Whether onboarding new team members, automating repetitive tasks, or enabling seamless collaboration, Keboola AI Copilot is your intelligent partner for data projects.

Three Stages of AI Integration

• Assist: AI supports users by automating routine tasks, accelerating workflows, and ensuring teams stay focused on high-value activities.
• Augment: AI enhances human effort, enabling computers to independently handle more complex operations—such as generating onboarding guides or summarizing project architectures.
• Amplify: AI expands what’s possible, helping teams achieve more than before, facilitating cross-functional collaboration, and unlocking new insights from data.

From LLMs to Foundation Models

The AI landscape is evolving beyond text-based large language models (LLMs). Modern foundation models can process images, videos, sensor data, and more, making them adaptable for a wide range of data-centric applications within Keboola. These models are accessible via APIs or can be deployed and fine-tuned in custom infrastructure, democratizing access to advanced AI capabilities.

The Power of Tools and Agents

Tools integrated with foundation models enhance AI's capabilities. For example, while LLMs may struggle with complex calculations, integrating external tools (like Wolfram Alpha for math or Keboola's own metadata APIs) bridges these gaps. These tools enable specialized agents to act programmatically, automating tasks such as querying metadata, generating SQL, or providing contextual recommendations for data engineers and analysts.

Frameworks like LangChain and LlamaIndex have made it easier to deploy autonomous agents that can handle multi-step tasks without constant human oversight. However, the adoption of agents in production software is still emerging due to their non-deterministic outputs and the need for reliable, predictable software behavior.

Use Case: Streamlining Onboarding

Onboarding a new data analyst or engineer can be challenging and time-consuming. Keboola's AI Copilot reduces the friction by:

• Generating tailored onboarding guides based on project metadata, data sources, and business context.
• Summarizing project architecture—identifying key data sources (e.g., MySQL databases, social media, Shopify, business applications) and the flow of data through pipelines.
• Providing actionable examples—like how geocoding APIs or RFM segmentation are used within a project.
• Helping new users navigate complex environments by surfacing the most relevant data objects and guiding them to the resources they need.

Example: Real-World Customer Implementation

Consider Ascend Climbing, a Keboola customer operating multiple gyms. When onboarding a junior analyst, Keboola AI Copilot quickly compiles an overview of all integrated data sources—ranging from sales in Shopify to check-in systems in MySQL databases. It details data processing steps (SQL, Python, geocoding, segmentation) and distinguishes between raw and modeled data, helping new team members focus on the most relevant information for analytics and reporting.

Enhanced Collaboration Across Roles

Keboola’s AI Copilot bridges communication gaps between engineers, analysts, and business users. By generating shared documentation and aligning terminology, it reduces the need for repetitive meetings and manual explanations. Teams can operate with greater autonomy, clarity, and speed—unlocking time for strategic, high-impact work.

Metadata-Driven Intelligence

The foundation of AI-powered productivity in Keboola is robust metadata. By indexing every data object (tables, pipelines, transformations) and enriching metadata with AI-generated descriptions, Keboola enables powerful search, discovery, and lineage features. This cumulative metadata enrichment means more accurate responses from the AI Copilot and faster time-to-insight for users.

Automating Data Lineage and Query Generation

Keboola’s AI Copilot can:

• Visualize data lineage down to the column level, spanning SQL and Python transformations—crucial for understanding dependencies and impact analysis during system migrations or audits.
• Generate and explain SQL queries based on natural language questions. For example, business users can ask for "all unique summer camp names and signups", and Kai generates the appropriate SQL and even explains the logic behind it.
• Highlight key data tables and transformations involved in specific business processes, such as payroll reporting or revenue KPIs, making it easier to update or migrate systems.

Example: Answering Business Questions with AI

Suppose a youth director at a climbing gym wants to know the number of signups for each summer camp. Traditionally, this would involve manual lookups or complex SQL. With Keboola AI Copilot, the director simply asks the question, receives an accurate SQL query, and gets the results in a readable format—no technical expertise required.

Ensuring Reliability and Trust

AI responses are only as useful as their accuracy and clarity. Keboola AI Copilot is designed to avoid misleading or irrelevant answers. If a question is too simple or ambiguous, the AI may choose not to answer rather than provide incorrect information. This conservative approach ensures trust and reliability, especially for mission-critical data tasks.

Continuous Improvement and Experimentation

Keboola’s AI Copilot is under active development, with features like advanced data lineage, improved passive prompts, and richer metadata integration being continually enhanced. Feedback from real users—like those at Ascend Climbing—drives ongoing R&D, ensuring the platform evolves to meet the needs of modern data teams.

Key Benefits of Keboola AI Copilot

• Accelerated onboarding: Reduce new hire ramp-up time by weeks.
• Streamlined collaboration: Break down silos between technical and business users.
• Automated documentation: Keep data projects transparent and auditable.
• Metadata-driven insights: Leverage enriched metadata for smarter, faster answers.
• Actionable lineage: Visualize and understand complex data flows across SQL, Python, and beyond.
• Secure and compliant: Keep sensitive operations under control, even as AI augments productivity.

Getting Started

1. Sign up for Keboola or request a demo to see AI Copilot in action.
2. Connect your data sources—from databases and cloud warehouses to business apps.
3. Leverage AI-powered onboarding to get new team members productive fast.
4. Empower your team to ask natural language questions, automate documentation, and visualize data lineage—all within Keboola.

Conclusion

Keboola’s AI Copilot is redefining what’s possible for data teams. By integrating advanced foundation models, leveraging rich metadata, and focusing on real-world workflows, Keboola enables teams to move faster, collaborate better, and achieve more. Whether you’re onboarding new analysts, migrating systems, or answering business questions, Keboola AI Copilot is the intelligent partner you need to stay ahead in today’s data-driven landscape.