Launched this week
Dex
Ask your data. Get answers and next steps.
476 followers
Ask your data. Get answers and next steps.
476 followers
Dex is an AI data analyst for founders. Connect your databases, spreadsheets, or BI tools, ask questions in plain English, and get instant answers with recommended next steps based on your data.
The multi-source approach is what makes this interesting to me — being able to query across Postgres, Sheets, and Stripe from one place is way more useful than yet another SQL copilot. The fact that you can inspect the actual query Dex ran is a nice trust-builder too.
Dex
Dex
@_anasansari
Thanks for asking, Anas
Custom integrations like an MCP server or similar setups aren’t supported yet, but this is something we plan to add in future updates as we expand DexData’s workflow capabilities.
The 'next steps' feature in Dex sounds like a huge time-saver for data-driven decisions. I have a question for the makers: does the AI suggest these steps based purely on current data, or does it learn from historical patterns as well?
Also, I’m wondering how others in the community are planning to use this for their content or product strategy. Any creative use cases yet?
Dex
@faisal_ahmed_rony Hi Faisal, great question! The next steps feature in Dex is primarily based on the current data, but it also takes historical patterns into account based on user queries when they’re available to suggest more informed actions.
@topealabi_ Cool product! How do you measure the quality or confidence of the structured data outputs?
Dex
@damian_quiroga
Appreciate that Damian!
We don’t just let the AI wing it every query goes through a pipeline that ensures it only pulls from real fields and tables in your database. Low-confidence or ambiguous outputs are flagged for follow-up, and the system can ask for clarification when needed. On top of that, we benchmark performance against public datasets (bridsql -minidev) with predefined queries to objectively measure accuracy. This approach ensures our outputs are both reliable and correct.
Good job! Love that you're solving this problem.
Curious whether & how it handles messy/incomplete data?
Dex
@zeebs Thanks for the feedback.
Our data ingestion layer already handles this.