@max_savonin1 All great questions, let me answer them one by one > Can DataLab's AI assistant handle complex data transformations and preprocessing tasks within the chat interface? If you have a good idea of what transformations and preprocessing you want to do, I suggest you switch to DataLab's "Code View", which is a fully-featured notebook editor. It has an AI code generator built in to write code from scratch or extend some code you already wrote. > How does DataLab ensure the explainability and trustworthiness of the insights generated by the AI assistant? All the code that the AI Assistant in the chat generates, is being added to a corresponding notebook. Users can seamlessly switch to this code and review, tweak, run and share it. Hope this answers your question! Any shares in your network are greatly appreciated!

@jgani Thanks for your message! We're at this journey for quite a while already. We started out with a code-first data notebook, but were always looking for ways to make it more accessible and easier to use, so that anyone, regardless of technical skill, could get insights from their data. The rise of generative AI has provided a new way of going about this challenge, allowing users to describe in plain English what insight they want. The quality of large language models (LLMs) is not (yet!) at the level where you can just throw any data warehouse at it and expect it to give accurate answers in all cases, but it still unlocks a great analytics experience: > Before, without AI: Analyst spends hours building a dashboard that can answer every possible question colleagues might possibly have. Colleague with a question needs to dig through a bunch of filters to get the insight they need. > With DataLab: Analyst on the team prepares a clean dataset with all the required properties and data slices, and makes this available through DataLab. Colleague starts a DataLab chat, ask what they need, e.g. ‘how many subscriptions did we sell in the past 30 days in France’. DataLab generates and runs code in the background, and answers the question on the spot. The generated to answer the question can be easily inspected, adjusted and re-run in DataLab’s fully-featured data notebook. LLMs will get more powerful, and DataLab will become increasingly smart about how to get the best results (leveraging past interaction data, previous data queries, company documentation, etc), making life easier for the analyst and getting insights significantly less frustrating for less technical folks. This answer turned out longer than expected, but I'm pretty passionate about all this, hope you don't mind!

@filipsch I'm all too familiar with the 'before AI' world. We had to build a custom solution for non technical people, literally like you described it. DataLab is very well positioned to take advantage of the AI revolution. Congrats FIlip!

@aman_wen Great question! In the "AI Chat" mode the user is not exposed to any code at all. Instead they just pop in a dataset and start asking questions. In the background, the AI is writing code to answer the question, but the user doesn't see this unless they want to. If the user decides to go to the "Code" mode, they're met with a clean, slick, yet powerful notebook editor with all the necessary features to delve deeper into the code, should they want to. Regardless of the mode that users are in, we're obsessed with keeping the UI and UX super simple and accessible, also for not super tech-savvy folks. Hope this answers your question!

@laocool Yes you can! https://workspace-docs.datacamp....

Hi @mashy! Excellent question! You can see an overview of security measures we take here: https://www.datacamp.com/datalab...

@yuvraj_dagur Imagine a startup building an offering online AI courses; to power their website and their product, they'll have a database (e.g. PostgreSQL), with different data models (courses, exercises, AI prompts, user feedback, etc). With DataLab, you can hook up to that PostgreSQL database in a matter of minutes, and ask the AI questions like "how many courses were started in week 20 of 2024". The AI knows about the database schema and will write and run SQL code (and maybe Python code after that) to answer your question and present it to you. All this to say: yes, startups and small businesses can definitely use this!

Hey @alex_dulub sorry for the late reply. We don't treat complex data queries differently from simpler data queries, but it's true that quality can suffer at some point (if the question is poorly defined, if the connected data source has a massive schema of 100s of tables, etc). LLMs will get more powerful, however, and we're constantly working to make DataLab increasingly smart about how to get the best results (leveraging past interaction data, previous data queries, company documentation, etc), making life easier for the analyst and getting insights significantly less frustrating for less technical folks.