If your Servoy solution has a database behind it, you have a backlog of report requests that never gets shorter. A user wants to see profit by region. Another wants top customers by margin. Another wants order trends over the last 18 months. Each one becomes a ticket, a form, a query, a layout, and a release cycle. Meanwhile the user who asked has already moved on to the next question.
The Dotzlaw team has shipped two versions of an AI system that lets users type a question in plain English and get a complete six-card dashboard back in under 30 seconds, no SQL written, no analyst in the loop. Both versions are designed to bolt right onto a Servoy solution.
Type “top 10 customers by profit margin in the last year” and get a sortable data table, a ranked bar chart, a trend line, a distribution donut, a scatter or combo chart, and a geographic pin map. We tested both versions against Microsoft’s WideWorldImporters database (90.5 million rows, 48 tables, 561 columns, 13 years of history) and hit a 100% SQL success rate across 10+ query categories.
Two versions, same AI pipeline, different front-of-house:
Version 1 — Metabase Edition: A full BI platform with natural language on top. Drill-through, scheduled reports, row-level security, and a visual query builder analysts already know, with the AI dashboard layer over top. Best for clients who want a real BI platform alongside the AI layer.
Version 2 — Native React Edition: Zero BI server, zero Java runtime. Charts render natively with Apache ECharts, AG Grid, and Leaflet. Three npm packages replace a 493 MB Java install. Best for embedding AI dashboards directly into a product you already ship.
What you get either way:
- Six coordinated visualizations per question, picked intelligently from 12 chart types based on the shape of the question. The system enforces that all six are different and only shows a map when the question is actually geographic.
- 100% SQL success rate via three reliability layers (14 T-SQL prompt rules, 7 post-processing fixes, retry safety net) plus a seven-point security validation layer that blocks any query that is not a pure SELECT.
- Saved dashboard library, reloadable in under 5 seconds with fresh data. A regenerate button re-runs the entire AI pipeline so saved dashboards automatically pick up improvements as the prompts and SQL logic evolve.
- Full SQL transparency. Every query the AI wrote is visible in an expandable panel, so you can verify the work, debug unexpected results, or hand the query off to a developer.
- Read-only database connection. Your existing Servoy application logic is untouched. The system does not create tables, run migrations, or modify your schema.
Why this matters for Servoy teams specifically:
Servoy is excellent for building structured, transactional applications. Forms, foundsets, workflows, business logic. It is less excellent at giving end users an open-ended way to explore their own data. Building a new report in Servoy means a developer touches the solution. Every “can you add a chart that shows X by Y” request becomes a release.
The natural language dashboard layer sits beside that. Users get self-service exploration over the same database, with no Servoy work required to add a new dashboard. Your developers stay focused on the application. Your users stop waiting in line for charts.
The system queries the database through a layer of views (prefixed `_v_` in our reference implementation) so you control exactly what the AI can see and query. If you already use views to simplify your data layer for Servoy forms, you are most of the way there. Works against MS SQL Server, PostgreSQL, and MySQL.
Happy to help. If you want a live demo, want to talk architecture, or want to see what either version looks like running against your own data, reach out. The Dotzlaw team can help get either version integrated into your Servoy solution quickly, or stand it up as its own application against the database you already have.