How It Works Agents Solutions Security Pricing Sign In

LumenQube Agents

Your AI Analytics Team.
Hired. On-call. Governed.

Six specialist AI Agents that own the full analytics lifecycle — from raw connector to dashboard to automation. Each one is grounded in your semantic layer. Together, they replace the brittle chatbot with something that actually works.

Start Free See how they collaborate
Ask
Dashboard
Schema
Data Prep
Automation
Insight
Not a chatbot

A team, not a text box.

Generic AI copilots hallucinate columns, invent metrics, and produce SQL no one can trust. LumenQube Agents are different: each one is bound to your governed semantic layer, each action is audited, and every output is validated — automatically retried if it's wrong.

GroundedEvery query runs against your certified metrics and dimensions. No hallucination.
Self-correctingEvery Agent validates its own output and retries before you see anything.
Governed & auditedRow/column security is enforced. Every action is logged with full lineage.
CollaborativeAgents hand off work to each other — exactly like a real analytics team.
Agent 01 • Conversational Analyst

Ask Agent

Your conversational analyst. Ask a question in plain English — the Ask Agent picks the right metrics, runs governed SQL against DuckDB, cites every source, and explains how it got to the answer.

Grounded in the semantic layerUses only your certified metrics — never invents columns.
Root-cause analysisRanks contributing drivers when a KPI moves.
Transparent reasoningShows the query, metric, and lineage for every response.
Hand-off to peersEscalates to Schema, Dashboard, or Automation Agents as needed.
Natural languageGoverned SQLRLS/CLS enforcedCitations
Live Why did MRR drop last week? Checking metric “MRR” • semantic layer MRR dropped $18K (−4.2%) vs last week Three contributing drivers ranked by impact: Payment failures (APAC) −$11K Churn uptick (Enterprise) −$5K Plan downgrades −$2K Source: metric mrr_total • RLS: revenue_team • 2.1s
Agent 02 • Dashboard Designer

Dashboard Agent

Designs complete dashboards from a single prompt. Picks the right widget types, lays out the grid, binds every panel to governed metrics, and validates the final output — retrying automatically if anything fails.

21 widget typesLine, bar, pivot, geo, sankey, heatmap, gauge, funnel, and more.
Smart auto-layoutGrid-aware placement with cross-filters and drill-down wired in.
Built-in validatorEvery widget is simulated before render — no broken dashboards.
Template libraryStarts from battle-tested templates for SaaS, e-commerce, ops.
Auto-layout21 widgetsSelf-validatingTemplates
Building PROMPT “Executive SaaS dashboard with MRR, churn, NPS” MRR$428K+6% Churn2.1%+0.3% NPS64+2 ARPU$142 MRR trend — 12 months Churn by plan
Agent 03 • Semantic Architect

Schema Agent

Authors and maintains your semantic layer. Proposes dimensions, measures, joins, and certified metrics — then simulates them against real rows before you commit. The foundation every other Agent depends on.

Star schema authoringFact + dimensions, with join paths inferred from data.
Simulate endpointEvery proposed metric is test-run against live data before apply.
Impact analysisShows which dashboards, metrics, and Agents depend on a change.
Glossary & lineageBusiness terms, synonyms, and data lineage generated automatically.
Star schemaSimulateImpact graphGlossarySelf-correcting
Proposing fct_orders fact • 2.4M rows order_id, amount, ... dim_customer 128K rows dim_product 1,240 rows dim_date date spine dim_geo country/region CERTIFIED METRICS revenue✓ arr✓ mrr✓ nrr✓ churn_ratesim ltvsim cac_paybacknew magic_nbrnew AGENT SAYS Simulating ltv on 128K rows… 3.2s
Agent 04 • Data Engineer

Data Prep Agent

Turns messy CSVs, raw tables, and broken extracts into clean, governed datasets. Describe the shape you want — the Data Prep Agent generates DuckDB SQL, enriches with ML (sentiment, entities, language), and streams the result.

Natural language → SQL pipelines“Clean emails, parse dates, drop dupes, fill nulls” → full DuckDB.
Streaming ingestionMemory-efficient imports from 33+ connectors — any size.
ML enrichmentSentiment, entity extraction, classification, language detection.
PII auto-detectionFlags sensitive columns and proposes masking rules for the Schema Agent.
NL → SQLDuckDBPII scanML enrichmentStreaming
Transforming PROMPT “Normalize emails, parse dates, dedupe, classify sentiment” RAW INPUT John@ACME.COM ,, 2024/03/01 -null- " USA " $12 great product!! #duplicate row #duplicate row j.smith@x..com 01-APR-25 £9 awful service ,,, DE 3 meh DuckDB SQL CLEAN OUTPUT john@acme.com • pos 2024-03-01 US • $12.00 “great product” row_unique ✓ dedup→1 j.smith@x.com 2025-04-01 • $11.34 “awful service” neg null DE • $3.00 “meh” • neutral
Agent 05 • Workflow Builder

Automation Agent

Translates a sentence into a full workflow. Triggers, branching conditions, chained actions, cooldowns, guardrails — assembled, validated, and ready to run in seconds.

7 trigger typesSchedule, webhook, data change, alert, chain, refresh, manual.
20+ action typesEmail, Slack, Discord, webhook, refresh, export, snapshot, run SQL, AI brief, policy…
Conditions & chainingAND/OR logic, freshness windows, automation-chain triggers.
Guardrails built-inCooldowns, snooze, dry-run, per-tenant caps, automatic failure alerts.
7 triggers20+ actionsChainingCooldowns
Generating “Every Monday, email me if NPS drops below 50, and post to #cx-team” Trigger • Schedule: every Mon 09:00 PT Condition • metric nps_score < 50 • freshness < 24h Action • Email CX Lead Action • Slack #cx-team Action • Attach AI brief with drivers & recommendations Workflow deployed • cooldown: 24h • first run: next Mon
Agent 06 • Proactive Monitor

Insight Agent

Watches every metric, 24/7. Detects anomalies, scores severity, surfaces drivers, and delivers an executive-ready brief — before anyone else notices something changed.

Continuous anomaly detectionSeasonality-aware outlier detection on every certified metric.
Root-cause rankingDecomposes drivers by dimension — segment, plan, region, cohort.
Executive auto-briefPlain-English summary with impact, driver, and suggested action.
Freshness monitorCatches stale data and pipeline failures before they hit a dashboard.
24/7 watchRoot causeAuto-briefFreshness
Alerting Revenue — last 30 days Real-time • seasonality-adjusted ! Anomaly • Revenue −23% (critical) Root cause: APAC payment outage • 64% of drop • 4h duration Suggested action: trigger incident runbook, notify FinOps
Agent Orchestration

One question. Six Agents. Zero friction.

Ask anything — the Agents figure out who does what. Here's what a single prompt actually triggers.

YOU “Why did MRR drop?” Ask Agent Needs LTV metric… Schema Agent Authors & simulates Data Prep Cleans & enriches Dashboard Builds the view Insight Agent Watches & briefs Automation Pages Slack on drop Slack delivered
Example scenarios

Real prompts. Real coordination.

“Build me an executive SaaS dashboard.”
1
Dashboard Agent picks widgets (MRR, churn, NPS, cohorts).
2
Schema Agent confirms metrics are certified — simulates LTV.
3
Dashboard Agent lays out grid, binds every panel, validates.
4
Insight Agent starts watching for anomalies.
“Why is APAC revenue down this week?”
1
Ask Agent parses intent and runs governed query.
2
Insight Agent provides driver decomposition (payment outage, churn).
3
Ask Agent writes a narrative answer with citations.
4
Automation Agent offers to notify FinOps on Slack.
“Load this messy CSV and make a report.”
1
Data Prep Agent cleans, dedupes, parses dates, flags PII.
2
Schema Agent proposes dimensions, measures, and joins.
3
Dashboard Agent creates a one-pager with key cuts.
4
Automation Agent schedules a weekly email refresh.
“Alert me when churn hits 3% with context.”
1
Automation Agent builds the trigger + condition + actions.
2
Insight Agent computes the driver decomposition on fire.
3
Ask Agent writes the Slack message with the narrative.
4
Schema Agent logs the event to the metric's history.
Capabilities matrix

What each Agent is best at.

Each Agent is a specialist. Together, they cover the entire analytics lifecycle end-to-end.

AgentWhat it doesCore skillsHands off to
Ask Conversational analyst. Answers plain-English questions with governed queries, citations, and narrative. NL → SQL • Root cause • Citations Schema, Dashboard, Insight
Dashboard Designs entire dashboards from a prompt, then validates every panel before render. Widget selection • Auto-layout • Self-validating Schema, Insight
Schema Authors your semantic layer. Dimensions, measures, joins, certified metrics — simulated first. Star schema • Simulate • Impact analysis Data Prep, Ask
Data Prep Cleans, normalizes, enriches, and de-duplicates data with NL-driven DuckDB pipelines. NL → SQL • Streaming • ML enrichment • PII Schema
Automation Turns “if-this-then-that” sentences into real workflows with triggers, conditions, and guardrails. 7 triggers • 20+ actions • Chaining • Cooldowns Insight, Ask
Insight Watches 24/7, detects anomalies, ranks drivers, and delivers executive-ready briefs. Anomaly detection • Root cause • Auto-brief Automation, Ask

Ready to hire your AI Team?

Start free. All six Agents included. Connect your first dataset in minutes.

Start Free Today Talk to Sales

Free forever plan. No credit card needed.