What would make you trust AI to answer questions about your business database?

I’m building dig data, a self-hosted analytics workspace for PostgreSQL, MySQL, MariaDB, and MongoDB.
Teams can ask business questions in plain language, build dashboards, and share consistent definitions while running Dig and database connections on their own infrastructure.
The beta uses read-only database access, bounded query results, source-level permissions, and an audit history. For AI, admins can choose Anthropic or OpenAI, or connect any OpenAI-compatible Chat Completions server with tool calling. Ollama and LM Studio are examples of compatible servers. Dig does not download or manage models; company context is sent to the configured endpoint, so privacy depends on its location and configuration.
I’d love to learn from people who already use BI tools or AI database assistants:
• What would make you trust an AI-generated database answer?
• Which safeguards are essential?
• Do you prefer reviewing the generated query before seeing the result?
• What prevents your team from adopting tools like this today?
• Which dashboard or collaboration features matter most?
The project is available here: https://digdata.mhesham099.chatg...