Today we’re introducing some updates to our Magic AI assistant that make it more powerful, intuitive, and faster.
We built our Magic AI tools to help humans do their data work— writing queries, futzing with Python code, digging up answers— faster and with more confidence. Our goal is to augment, not replace, human insight. This is not an "AI data analyst", because honestly, we don’t really think that should be a thing.
The first version of these tools has been available for almost a year now, and we’re excited about how useful it's been for our users. Thousands of people each week are using Magic to get more done with data: editing complex SQL queries, debugging confusing Python code, and speeding up the most tedious parts of their jobs (cough pandas cough regex).
Early users are telling us that it “shaves hours off their analytical workflows”, they “don’t have to circle back after meetings anymore, because Hex can help get answers before the end of a call”, and that their “keyboard life is being extended by years from the saved keystrokes.”
Today, we’re introducing a big upgrade that makes Hex’s AI tools even more helpful, updates and polishes the UX, and introduces some powerful new functionality.
If you'd rather watch a video than read a blog post, here's a summary of the new features. Otherwise, read on.
What’s new
Magic is getting three major upgrades. Of course, we’ve also been making lots of speed and accuracy improvements along the way, so that your AI generations are both faster and smarter! But here's the big stuff:
Generate mode
Magic can now generate multiple cells at a time in your notebook, chaining together SQL, Python, and Chart cells to answer complex questions or kickstart a new analysis.
It’s easy to generate new cells from anywhere in a project: just hit Cmd+G, use ↓ from an edit mode prompt bar, or click the “Add with Magic” button between cells.
For example, you can get started quickly in a new project without having to forage for initial data:
Or generate new cells in the middle of an existing project to help with complex questions or new directions:
Hex’s cell-based UI really shines here! This upgrade lets Magic work in your notebook just like you would, using multiple cells to run queries, build charts, and iteratively construct complex logic. New cells are all created as drafts, so you can inspect, edit, and validate AI generated code before accepting it. Once accepted, the new cells seamlessly integrate into the notebook flow.
More powerful, streamlined prompting
To support new cell generation and more complex instructions, we’ve rebuilt the prompt bar, unifying all Magic actions into one simple interface. You can activate it on a cell with Cmd+Shift+M, and seamlessly switch between editing existing cells or generating new cells with the ↑ and ↓ keys.
The new design also supports more complex multi-line instructions and gives more power to auto-fixes. You can choose to auto-fix a problem with one click, or use the prompt bar and provide specific direction to guide a fix. I like to use this to both fix and edit a query in the same step!
Mentioning data
The prompt bar now supports "@ mentioning" datasets to specify what resources it should use. This is a simple addition, with a huge impact on prompting efficiency and accuracy.
You can use this to point at a specific database table (handy if you're off-roading and using some unusual, undocumented resources) or if you're in a complex project, to direct Hex towards a particular dataframe.
One you get into the routine of @'ing datasets, you'll wonder how you lived without it. And it will have a huge impact on the accuracy and quality of your generated code, so you're well incentivized to remember.
Improved SQL generation
Over the last few months, we’ve put in a ton of work on our state-of-the-art metadata retrieval pipelines, and seen steady improvement in Hex’s ability to accurately generate accurate SQL queries.
One important note is that no matter how good the models, or our prompt engineering around them, context is critical! The more metadata you can provide about your tables and columns, the better Hex can effectively prompt the model and maximize completion quality.
You can do this in three main ways:
dbt Docs: if you’re using dbt Cloud, all you have to do is update the metadata in your models, and it’ll automatically flow into Hex and be used to help inform query generation.
Support for warehouse metadata: ditto if you update information about your columns and tables in Snowflake, BigQuery, or Redshift.
via Hex’s built-in Data Manager, which lets you edit additional metadata as well as promote or exclude schemas and tables.
Limitations of AI
Amongst all the hype about what AI can do, it’s worth taking a moment to acknowledge the current state of LLMs and tools built around them.
Working with AI tools can be funny – they blow your mind one moment, and then the next you're shaking your head, wondering how they could possibly be so silly.
So, as we build and iterate on Hex's AI features, our north star has always been "are people finding this useful?" Our big target isn't 100% code perfection (though we're getting better all the time), and it's certainly not the construction of some “AI analyst” that replaces people. It's about being simple, useful, and solving the real problems data practitioners face every day.
But make no mistake – it’s not perfect, and definitely not a replacement for human judgement.
Take it for a spin
Magic AI is available for all Hex customers, including those on our community plan or in active trials. Community users are limited to 100 monthly requests per user.
Got some privacy and security questions? The important stuff: neither Hex or our model partners train models on customer data, and all metadata is stored in a secure vector database running inside Hex’s architecture. If you have other questions, you can learn more in our documentation.
Ready to try it out? Click here to jump directly into a project and see it in action.