Skip to content
Agents tracked: 258 Downloads (7d): 219M up 6.1% GitHub stars: 5.5M VS Code installs: 148M Releases (7d): 293 Agent status: 2 with issues Updated Oct 7, 2026
Case study Modal

How Modal gave non-data teams self-serve analytics with Hex agents

Serverless compute company Modal connected its warehouses to Hex and let staff ask data questions through Hex's agents, including from Slack. Hex reports faster analysis and wider use beyond the data team.

The problem

Staff had the technical skills to query data. But setting up notebooks and validating outputs discouraged quick ad hoc questions, which mattered more as the company grew.

How they did it

  1. Connected two warehouses (Snowflake and ClickHouse) in one Hex workspace.
  2. Configured the agent with less than 50 lines of context, using endorsed dbt models and workspace rules.
  3. Enabled the Slack integration so people can ask @Hex questions directly.
  4. Used Context Studio to monitor agent usage and keep answers accurate.

Results

  • Hex reports multi-day analysis compressed to under 24 hours (75% faster).
  • 1,300 AI agent messages across Notebook Agent and Threads. 53% of Threads conversations started from Slack.
  • 6 cross-functional teams use Hex independently.
  • Hex says Modal unlocked $100M+ in pipeline by using the agent to build targeted outreach to 15 AI labs. These had Modal accounts but no usage.

As reported by the source (Hex customer story); AgentGid did not measure these figures.

Takeaway. Keep the agent's context short and based on endorsed models, and put it where questions already happen, such as Slack.
AgentGid's take

This suits a team that already has a warehouse and dbt models, plus someone who can curate the endorsed models and rules the agent relies on. Watch the cost: full agent access starts on Professional at $36 per editor a month, AI usage draws on per-seat credits, and the speedup and $100M+ pipeline figures are Hex's own reports. PandasAI (Free, open source) is a cheaper alternative if you can build the Slack layer yourself.

The agent used here

Similar use cases