Ask about your data
Trends, comparisons, counts and likely contributors.
Bold BI's AI Agent is a read-only chat in the Dashboard Viewer. Ask a question about the dashboard you are viewing, with no SQL, and it answers with a chart or a written explanation. It works on existing dashboards, considers your last 4 conversations and enforces row-level security.
30 days. No credit card required.

Row-level security is enforced before anything is retrieved. OAuth and SAML sessions carry over in Cloud and On-Premises.
Bold BI holds a SOC 2 Type 2 report. Its products are considered a Business Associate under HIPAA and independently audited for HIPAA compliance.
Customer data is not used to train or fine-tune Bold BI's AI models.
Read the SOC 2 overview and the HIPAA overview.
Bold BI's AI Agent skips the ticket: ask inside the dashboard you are already viewing, with no SQL or query language. "Which segments contributed most to the churn increase in March?" shouldn't wait for a report.
Bold BI's AI Agent works in the same 3 steps behind every question, whether it is a quick count or a written explanation.
"Show me top 5 products by margin this quarter." Plain English, the way you'd ask a coworker.
Understands business terms, not only field namesThe Agent maps your words to your data using schema context and enforces row-level security before it retrieves anything. SQL runs only inside the secure Bold BI environment.
Role-based access and row-level security on every requestA chart or a written answer. Written answers need "Enable Widget Insights and Dashboard Summary" switched on.
Multi-turn: considers your last 4 conversationsRead-only analysis
Bold BI's AI Agent covers 4 areas of read-only analysis and never edits widgets or the layout. On-premises users on v10.1 or later switch it on with the global "Enable AI Agent" option.
Trends, comparisons, counts and likely contributors.
Clear table and column names and exact column references improve accuracy.
Where Widget Insights and Dashboard Summary are enabled, answers can include a written summary.
Refine and drill with follow-ups instead of starting over.
The Agent only reads. Use AI Copilot to build visuals, or Prompt to Dashboard to generate a dashboard from one prompt.
AI-generated output should be reviewed before it's used for business decisions.
Bold BI has 3 AI surfaces on one AI engine and security model: AI Copilot, AI Agent and the AI Assistant hub, which embeds from v14.1.
A Bold BI AI Agent session needs no SQL and runs in 3 stages on a sample Sales Analysis dashboard: the question, the chart and a follow-up that refines the answer.



Embedding
Yes. The in-dashboard AI Agent embeds with the Dashboard Viewer, and the standalone AI Assistant embeds with the JavaScript SDK from v14.1. The embed token is generated on your server.
OpenAI, Azure OpenAI and Anthropic (Claude); Anthropic needs v16.1.90 or later. On-premises only.
An embedded AI Assistant carries no third-party AI branding.
Row-level security carries into embedded sessions, and so does SSO in Cloud and On-Premises.
Admins review AI credit use by user, site and time period.
Bold BI positions the AI Agent for 4 audiences: executives, BI and data teams, SaaS product teams, and IT and security.
Ask a question, read the explanation, skip the analyst meeting.
Every question the Agent answers is one less ticket in your queue.
Embed it, white-label it, ship conversational analytics as your feature.
Roll it out on role-based access, row-level security and OAuth and SAML sessions.
Bold BI's AI Agent is not the right fit for building or editing dashboards, or for calculations such as median and variance.
It is read-only: it never edits widgets or the layout.
Written answers need "Enable Widget Insights and Dashboard Summary" switched on, and its chat is not saved as a named conversation.
The AI Assistant features, including the AI Agent, support 6 aggregations (COUNT, COUNTD, SUM, MIN, MAX, AVG), not MEDIAN or VARIANCE.
A tenant can send up to 100 messages per minute, and accuracy varies by language.
Your own AI provider is on-premises only.
Questions
Articles