Systems | Development | Analytics | API | Testing

Large Language Models (LLMs) are transforming how organizations interact with their data infrastructure, offering unprecedented capabilities for both technical and business users. However, this transformation brings unique opportunities and challenges that vary significantly based on user personas, security requirements, and implementation approaches. This writeup explores these dimensions through the lens of practical implementation using tools like Keboola MCP and various client interfaces.

Joe Reis, author of Fundamentals of Data Engineering, known for practical education and his YouTube content — formerly CEO/co-founder of Ternary Data — is reviewing our Keboola MCP (Model Context Protocol) Server with Claude to list Shopify his tables, run exploratory analysis, export data to BigQuery, and generate a star schema. And you can do the same in minutes! We will go through everything here from one-click setup to best propmts to use MCP with.

If you’re in procurement, finance, or operations at a mid-sized ecommerce or retail company, tariffs aren’t just a political talking point - they’re a silent force eating into your profitability. And the real problem isn’t always the tariffs themselves. It’s the lag between policy changes and when your data catches up. If any of it sounds familiar, we’d love to hear your point of view in the short survey below.

When we officially launched our Model Context Protocol (MCP) server integration on June 12, 2025, we weren't just adding another feature - we were fundamentally changing how data engineers interact with their tools. One month later, the transformation has exceeded our wildest expectations.

I work in marketing, but I spend more time in automation flows than in campaign briefs. And I love it. ‍ At Keboola, I’m part of the Marketing team, but behind the scenes, I’m the one wiring up lead capture, validation, routing, enrichment, and notifications. Every form submission, every hot lead, every triggered campaign has to land somewhere fast and clean. ‍ That’s why I’m genuinely excited about this: Keboola and Make have officially partnered!

Remember when employees went rogue with cloud apps and Shadow IT became IT's nightmare? Well, meet its chaotic sibling: Shadow AI. Employees everywhere are quietly dabbling with AI tools, and IT usually has no idea. It's fast, it's convenient, and it's also a ticking security bomb. Nearly 80% of companies already faced some AI-driven fiasco, from data leaks to embarrassing decisions - and IT leaders are seriously freaking out (TechRadar, 2025). But banning AI isn’t the answer.

Sometimes the biggest opportunities come disguised as unproven protocols released on a random Monday. Here’s why we bet on MCP before anyone asked us to. ‍ Two months before anyone knew what MCP was, we made a bet that it would fundamentally transform how people interact with their data infrastructure.

‍After 10 days since launching the Keboola MCP (Model Context Protocol) Server, we've gathered the most common questions from our data community. This article combines practical answers with best practices inspired by successful AI-assisted development patterns, helping you get the most out of your AI-powered data workflows.

Keboola Data Templates are AI‑enhanced, reusable pipelines you launch with one click or API call. They eliminate the need to rebuild data workflows from scratch—so every use case is a fast-track to insights and impact. No more reinventing the wheel: once a template exists, it can be shared and adapted across departments—marketing, finance, ops—saving weeks of build time and hundreds of engineer-hours.

The biggest issue with chatbot implementations powered by generative AI is the accuracy and reliability of the output. Models can give erroneous or inaccurate answers due to hallucinations or simply because they lack information specific to a given business case, as many of them don’t have access to new data outside of pretraining. Retrieval-Augmented Generation (RAG) is a technique designed to address this limitation by integrating an external retrieval mechanism with a generative model.