Sundial Overview

Sundial is an agentic analytics platform that gives every decision-maker a world-class AI analyst, and every data team the accuracy and observability to trust the answers. It runs on top of your existing warehouse and LLM tooling, combining the speed of AI with the rigor data teams expect from production systems.

Why Sundial

Getting to self-serve analytics traditionally meant months of modeling data and building dashboards in locked-in BI tools: slow, but reliable. Point an LLM at your warehouse and you get answers in seconds instead, but with a new problem: you can't tell how accurate any of them are.

Sundial closes that gap. It delivers fast, conversational analytics and tells you how confident it is in every answer, so you always know when a result is guaranteed versus directional, and the data team keeps a governed view of how the platform is used.

What Sundial does

Sundial acts as a data analyst and engineer for your whole organization. It:

• Audits your data quality and recommends what to capture next.
• Cleans, models, and builds a semantic layer out of your data using natural language.
• Runs deep, context-aware analysis against your warehouse: complex analysis, not just metric look-ups.
• Surfaces what matters, telling you what you need to know even when you don't ask.
• Reports its confidence, so you know when something is guaranteed or directional.
• Meets you where you work: Slack, Teams, and other interfaces, and evolves its context at the pace of your business.

The Sundial AI platform

Sundial is organized into three layers that work together in a continuous feedback loop:

• Context layer: captures business context, domain knowledge, semantics, and your company's analytical model so agents reason the way your team does.
• Agents: a set of specialized agents that operate on top of that context:
  • Analysis Agent runs context-aware analysis against your warehouse using expert playbooks. This is the default agent, available to everyone.
  • Data Quality Agent validates your data and recommends what to capture next.
  • Data Modelling Agent transforms raw tables into clean, analysis-ready pipelines.
  • Storytelling Agent distills analysis into narratives and recommendations.
• Interface layer: the surfaces where you interact with Sundial and where insights are delivered.

The Analysis Agent is available in every workspace; the Data Quality and Data Modelling agents roll out progressively, so ask your Sundial contact if you don't see them yet.

Built for trust

Sundial is designed to be trustworthy in production:

• Confidence on every answer: a trust signal (high, medium, low, or unverified) tells you whether a result came from governed definitions or ad-hoc SQL, so you know how far to lean on it.
• Expert playbooks encode methodological, repeatable analysis written by domain experts, not naive, one-off LLM output.
• Observability and governance keep every step transparent and auditable.
• Evals catch quality regressions before they reach business users.
• Data stays in your environment: Sundial is self-hostable, so your data never has to leave your premises.

How it fits together

The pieces form a feedback loop the data team can steer:

1. Business users ask the agents questions and explore dashboards.
2. Observability shows where answers are weak, usually missing or incomplete context.
3. The team improves the Context Engine: semantic definitions, AI context, or playbooks.
4. Evals confirm the change improved quality and didn't regress anything else.

Over time this turns reactive firefighting into a deliberate loop between real usage and better context.

Security and support

• SOC 2 Type II certified, HIPAA compliant, and self-hostable.
• An applied AI team of former data leaders helps set up your metrics, playbooks, and context in the first 30 days, so you're never left to figure it out alone.

Explore the docs

Start with the area that matches what you're trying to do:

Setting Sundial up for the first time?

• Connecting your data: connect your warehouse and data sources.
• Semantic Reference: model the dimensions and measures the agents and dashboards use.
• Self-Hosted Deployment: deploy Sundial inside your own cloud account.