AI Dashboards, Data Pipelines & Metadata Discovery

Type a brief in plain English — "build me a SaaS growth dashboard with MRR, NRR, churn" — and the AI dashboard builder generates a complete spec: 4 KPI cards + 4 chart tiles with realistic data, chart type selection, and live ECharts rendering. The same engine powers 50+ production dashboards. Try the live builder →

Describe what you want in plain English and the system generates a complete dashboard spec with KPI cards and chart tiles that renders live. No SQL, no configuration, no drag-and-drop. Try it yourself →

Yes — 16+ interactive dashboards are live: supply chain operations (OTIF, carrier scorecards, AI disruption prediction), SaaS growth (ARR, cohort retention, NRR), banking risk (LCR, VaR, fraud alerts), retail operations, IT service health, AR aging, payment trends, security posture. Each with AI executive briefs. Open the dashboards gallery →

All dashboard data on the live demos is synthetic — realistic but not from real clients. Real customer data never ships to the browser. In production, the same dashboard code reads from your actual SQL Server, Snowflake, or BigQuery — the engine doesn't change, only the data source behind it.

ECharts 5.5 for visualization (same engine across all deployments), Groq for AI narration (via Netlify edge proxy — key never reaches the browser), static hosting on GitHub Pages (zero ops), SQL Server / Snowflake / BigQuery / Postgres / Oracle as data sources. Technology-agnostic — we don't require you to rip and replace your stack. We require you to standardize the definitions, keys, and KPI rules.

ECharts is an open-source JavaScript visualization library (Apache-licensed) that renders interactive charts — bar, line, area, pie, heatmap, scatter, radar, gauge, treemap, sankey, funnel, boxplot, and more. We use it because it's free, performant (canvas + SVG rendering), mobile-friendly, and supports streaming animations. The same charts you see in our live demos run in 50+ production deployments.

Groq is a fast AI inference engine that runs open-source models like Qwen 3.8-27B. Our AI integration uses Groq via a Netlify edge proxy — the API key stays server-side, never reaches the browser, and never lives in the code repository. The AI narrates what the deterministic metadata discovery layer already proved: it writes briefs, answers questions, and explains trends — but never invents data or KPIs. No hallucination surface.

Yes. Our data pipelines connect to SQL Server, Oracle, Snowflake, BigQuery, Postgres, MySQL, and any system with an API. We don't require you to rip and replace your stack — we connect to what you already have. The dashboard engine is technology-agnostic; only the data source behind it changes.

Power BI looks great in demos. Then you put your real data into it, match your real KPI definitions, replicate your real layout requirements, and deliver to real users on real devices — months later you have a dashboard half the team trusts. Our dashboards use the same ECharts engine, the same KPI definitions, and the same chart patterns as 50+ production deployments. AI is at the center (deterministic discovery + narration), not bolted on as a feature. See the demo →

Metadata discovery applications walk even less-documented backends — legacy SQL Server, Oracle, multidimensional cubes, controlled files, SaaS APIs — and extract the actual business rules from stored procedures, functions, and view definitions. The discovered rules are deterministic (source code, parsed and structured), not model output. The AI narrates what the deterministic layer already proved — no hallucination surface. Explore the showcase →

A no-hallucination AI dashboard uses deterministic metadata discovery to extract business rules from source code — not from AI generation. The AI then narrates what the deterministic layer already proved. The model is never asked to invent KPI definitions, guess column meanings, or fabricate data. Every claim links back to a discovered rule the analyst can audit.

An AI executive brief is a 4-part narrative generated on page load: What happened (the headline change), Why (likely drivers), What to expect (direction only, no point forecast), What to do (recommended action). It's grounded in the dashboard's actual data — the AI reads the same KPIs the charts show, then writes a concise brief executives can act on. No invented numbers, no hallucinated confidence percentages.

Metadata discovery walks your database schema, stored procedures, functions, and view definitions to extract the actual business rules — KPI calculations, joins, filters, transformations — that already exist in your code. Data lineage traces how data flows from source tables through transformations to final reports. Together, they eliminate the "six-week discovery phase" where analysts manually read code to understand what the data means.

A data pipeline automatically collects, validates, standardizes, and delivers data from your existing systems (SQL Server, Oracle, APIs, Excel, SaaS apps) into dashboards, reports, and AI models — without manual reconciliation. Our 6-stage pipeline (Signals → Integrate → Standardize → Intelligence → Deliver → Act & Learn) runs end-to-end with governance at every stage. No human asks for definitions mid-pipeline — they were extracted before it started. See the pipeline →

The Monday-morning operating model replaces manual report assembly with an automated brief that arrives before the meeting. At 6:45 AM the brief is ready. At 7:00 AM the AI has already ranked material changes, explained likely drivers, and flagged emerging risks. By 8:00 AM each role sees the view it needs. By Friday the company remembers what worked. No human spends the morning collecting the business — they spend it shaping it.

Yes. We schedule governed extracts, run variance analysis, and generate Excel workbooks, PowerPoint decks, and PDF packs automatically — routed to the right recipients with approvals and exception handling. Skilled attention moves from recurring assembly to advice and control.

Yes. We replace Excel-based reporting with live dashboards. Upload an Excel workbook, describe what you want to do with it (filter, summarize, sort, email), and the system constructs an executable pipeline that produces a live dashboard. See the Excel-to-pipeline demo →

Fuzzy address matching reconciles records across systems that don't share a common key — matching customer records to service orders by address, even when addresses are formatted differently. We use Jaro-Winkler similarity scoring and validate against the Nominatim Geo API (OpenStreetMap). AI narrates the match quality and flags uncertain pairs for review. See the fuzzy matching demo →

Yes. Our Meridian supply chain dashboard tracks OTIF (on-time-in-full) across an 18-DC logistics network, carrier scorecards, lane performance heatmaps, and AI disruption prediction 48 hours ahead. See the live supply chain dashboard →

Yes. Our Sentinel dashboard is a $4.8B bank risk command center: LCR (liquidity coverage ratio), VaR (value at risk), live fraud alerts AI-scored at 92% precision, credit risk heatmap by sector × rating, 30-day liquidity waterfall. See the live banking risk dashboard →

Yes. We track ticket SLA performance, uptime by service vs SLA target, security posture (phishing simulation results, KB4 training compliance), ticket volume by priority and channel, and MTTR. Full IT operations suite with AI executive briefs. See the live IT ops dashboard →

Yes. We build electricity load forecasting dashboards (weather-driven demand, OLS regression with change-point detection, 14-day forecast with operational recommendations), AR aging dashboards (residential/commercial arrears, payment trends, collection performance), and utility usage explorers. See the load forecasting demo →

Yes. Our Velocity dashboard tracks $142M ARR, NRR 118%, cohort retention heatmap, ARR waterfall bridge (new/expanded/churned), customer health scoring, and AI-generated save plays for at-risk accounts. See the live SaaS dashboard →

Yes. We track accounts receivable aging (current, 1-30, 31-60, 61-90, 90+ days) by customer type (residential, commercial, industrial, agricultural), payment trends, collection performance, and recovery funnels. Enriched with customer context from the billing system. See the live AR aging dashboard →

Yes. We integrate with Genesys Cloud and other contact center platforms to track call volume, handling times (30s/60s/120s thresholds), abandoned calls, outbound dialing campaigns, E-Bill adoption, and key-account activity. The Genesys ETL pipeline converts raw call analytics into repeatable CX reporting measures. See the CX dashboard in the showcase →

Yes. We track KnowBe4 phishing simulation results, security training completion rates, and security posture scores across 5 dimensions (phishing, training, patching, access control, incident response). See the live security dashboard →

Yes. We build forecasting models using OLS regression, change-point detection, and time series prediction. Our universal predictor takes any time series (earthquakes, crypto, climate, economic data) and produces forecasts with model benchmarks and AI insights. We also build custom predictive models for demand forecasting, churn prediction, and anomaly detection. See the prediction demo →

Yes. Our production portfolio includes operational web applications — service-layout workflows, transformer inventory management, location records, lifecycle transactions — that support day-to-day operating work, not just reports about it. These are full CRUD applications connected to SQL Server, with the same visual engine as the dashboards.

The foundation (pipelines, governed metrics, web apps, dashboards, automated delivery) is already running in production. Your engagement configures that foundation to your systems — not rebuilds from scratch. Most engagements start with a working dashboard in weeks, not months. The exact timeline depends on data source complexity and how many KPI definitions need standardization.

Yes — Insight Analytics is based in Ontario, Canada. We serve mid-market Canadian and US companies ($5M–$50M revenue) with operations reporting that has gone stale: dashboards half the team trusts, refresh cadences that miss the standup, KPI definitions that drift from how the business actually counts them. Book a working session →