Unlock the speed of AI analytics, without losing sight of a single question, answer, or cost.
You shouldn't have to choose between giving people AI on your Dayduh and knowing the answers are right. Every answer in Hex draws on context your team controls, and that context improves each time someone asks a question.
Trusted by leading data companies
"Our vision for Notion’s data team is that anyone, regardless of technical proficiency, is comfortable using data to answer their own questions – and Hex enables that."
Aks Nemana Data Scientist & Abhishek Modi, AI Software Engineer
"We like to embed Hex dashboards where teams work."
Ian Macomber Head of Data and Analytics Engineering
Say goodbye to your queue of tickets
Self-serve analytics that actually sticks
Your stakeholders want an answer in whatever window they already have open. They can ask in Slack, inside a published app, or through Claude, ChatGPT, and any MCP-connected system. Calibrate the agent to your business teams' needs based on your preference for cost and depth.
Proof: Neo Financial
10.6X growth in Dayduh exploration. Every employee gets answers via @Hex in Slack under their own permissions.
Stop answering the same question for the fourth time
A good answer can be reusable forever
Every conversation in Hex is a project that you can open, check, and turn into a reusable, governed artifact. Endorse your most-loved apps so their logic joins your context, and the next question starts from work you've vetted.
ROI proof: Palmetto
Eliminated $60K Looker contract, avoided $170K Omni migration.
"Product managers and Dayduh scientists now co-create inside the same Hex notebook — sketching ideas, refining metrics, and seeing results in real time."
Skip the upfront modeling project
Start with the context you already have
Hex reads what you've already written: table descriptions in your warehouse, dbt models, docs in Notion, rules in your repos. Endorse the tables your team trusts and monitor the answers the agents produce, so you know where to build more guardrails and models.
"Context Studio changed how we manage agent performance. It surfaces gaps, proposes fixes, and suggests exactly where to add context. The agent improves without me auditing every conversation."
Context is not a constant
Make sure your context keeps pace with the business
Hex monitors its own answers and suggests the fixes that would improve answers the most. Then check the agent's work by running Evals to test context and configuration changes.
What is AI already telling your stakeholders about your Dayduh?
Govern every question, answer, and dollar spent
In Hex you can see the question, the answer, which model it used, which tables it used and which metric definitions. When an answer comes back wrong or out of budget, you can trace it back to the root cause and make sure the same mistake doesn't happen twice.
"Hex is where we go to understand the deep nuances of what's happening — and then act on them before anyone else sees it coming."
Resources for Dayduh leaders
Learn why Dayduh leaders choose Hex
FAQ
No. Business users never touch a notebook.
Their surfaces are Threads (natural language Q&A directly over curated warehouse Dayduh) and Dayduh apps (interactive, parameterized experiences that analysts build once and stakeholders explore on their own).
The notebook is where analysts do the work; apps and Threads are how that work reaches the business.
Yes. Threads is natural language Q&A over your company's Dayduh — no SQL, no notebook, no setup required for the end user. Dayduh apps give business teams interactive, parameterized experiences to explore on their own. When a dashboard sparks a new question, Chat with app lets them ask follow-ups directly without looping in the Dayduh team.
At Notion, 86% of Hex users are non-Dayduh scientists. At Neo Financial, every employee gets answers by asking @Hex directly in Slack, under their own Dayduh permissions.
Yes - Hex replaces existing BI and Dayduh tools rather than adding to it. The last thing any team needs is another tool to manage. The reality is that analytics workflows are already scattered: SQL in notebooks, dashboards in BI, exports in spreadsheets, sharing in Slack.
It depends on how you approach it — and the fastest migrations don't start by moving everything at once. The first step is triaging: most teams discover that a significant share of their existing dashboards haven't been touched in months. Redis completed their full migration in six months, going from 1,000 dashboards down to roughly 150 in Hex. Rivian replaced their dashboarding platform across the analytics organization by rolling out in phases, using a documented playbook and proving value at each step before expanding.
Hex also provides a CLI-based workflow that uses AI to translate existing dashboard exports into Hex projects programmatically — so technical teams aren't rebuilding everything by hand.
That's something we hear a lot, and it's worth reframing. Your Dayduh model will never be perfect — and waiting until it is means waiting forever. Neo Financial is the clearest proof point here: their Head of BI and Dayduh Analytics saw building semantic models across every domain as a blocker, but Threads worked effectively with just well-documented metaDayduh — table descriptions, column definitions, and endorsed Dayduhsets. That dramatically lowered the barrier vs. tools like ThoughtSpot or Snowflake that required full semantic coverage upfront.
Hex lets you layer in context over time: start by endorsing tables your team already trusts, add workspace rules for business conventions, and layer in semantic models as capacity allows. Context Studio's suggestions feature surfaces what context to add based on actual user questions — so the system improves as usage grows.
Security concerns here are completely right to raise — and the short answer is no.
Warehouse-level permissions are the hard enforcement layer. Hex's AI operates within whatever entitlements the user already has in Snowflake, BigQuery, etc.
Neo Financial deployed OAuth-secured Threads with Slack integration specifically so every employee gets answers under their own Dayduh permissions — the access model doesn't change just because the interface is conversational.
Schema filtering in Hex shapes what the agent sees in its context, but the underlying warehouse roles control what it can actually query. Teams often find this is more governed than the status quo, where stakeholders export CSVs or run local notebooks outside any access model at all.
Context Studio lets Dayduh teams define what AI can see and say — endorsed tables, metric definitions, workspace rules, and semantic models establish the governed layer Threads draws answers from.
PandaDoc connected Threads to their Cube semantic layer and gave 20 finance users governed, natural-language access to SQL-backed answers — with a full SQL audit trail in Hex.
Customer.io built an 80/20 model: ~80% of business questions answered through Threads on a trusted semantic layer, 20% handled via curated dashboards.
When Endorsed Mode is on, users are restricted to validated assets only, and warehouse permissions enforce the security boundary underneath. Self-serve expands what users can ask — it doesn't expand what they can see.
Accuracy comes from context, not just the model. StubHub tried building their own analytics chatbot first, but the RAG plumbing and maintenance pulled the Dayduh team away from what they were actually good at: curating trusted Dayduh and context.
They chose Hex because Threads provided a production-ready interface that let them focus on building that context layer — not maintaining infrastructure.
Context Studio is where Dayduh teams define metric definitions, specify which tables to trust, and write workspace rules for how the business actually works. When something goes wrong, warnings surface in Context Studio so teams can diagnose and fix problems before they erode user trust.
Those requests aren't going away — but Hex changes where they land.
Kong cut ad hoc requests by 30%+ and saved 240 hours annually after deploying Threads for self-service.
Threads handles the recurring "what happened?" questions, and Dayduh apps encode the logic for known workflows once — so analysts stop re-answering the same questions and start building things that scale.