SQL Server
Query SQL Server in plain English. No T-SQL required.
Insiro Datasight
Ask your data anything. Get the dashboard, not a ticket.
Datasight turns plain-English questions into validated SQL, charts and narrative across 30+ data dialects, then assembles BI dashboards from a one-line brief. No semantic-layer project first: it discovers your ontology for you.
NL → SQLAuto ontologyFederated queriesBI StudioPower BI import
Which regions grew revenue fastest this quarter, and why?
10 steps · 412 ms
QoQ revenue growth by regionbar · auto
The story
APAC led with +31%, almost all from enterprise renewals in Japan (+$2.1M).
LATAM slipped −4% after the Brazil price change in August; volume held, discounting rose.
Next: compare Brazil discount rates by rep
What is Insiro Datasight?
Insiro Datasight is a natural-language analytics product. Business users ask questions in plain English and get validated SQL, charts, KPIs and a written explanation, then turn answers into multi-page BI dashboards. It supports 30+ database and warehouse dialects and automatically builds a semantic layer from your schema.
- 30+SQL and warehouse dialects
- 10validated steps per question
- 27chart types chosen for you
- 0write access to your data
01 · Coverage
30+
SQL and warehouse dialects
Relational, enterprise, cloud warehouse, analytical and NoSQL, each with its own SQL dialect handled for you.
02 · Rigor
10
validation steps per question
From ambiguity detection to confidence scoring, with the SQL and trace one click away.
03 · Expression
27
chart types, picked for you
KPI, waterfall, sankey, funnel, treemap, geo map and more, chosen by the shape of your answer.
Ask in English
A conversation with your warehouse.
Ask a question, then a follow-up. Datasight keeps the context, shows how confident it is, and lets analysts open the SQL, edit it and run it again.
- Multi-turn sessions that remember filters, entities and time ranges.
- Confidence badges (green, amber, red) and an Ambiguous query flag when a question could mean two things.
- Drill down by clicking any chart point; Datasight writes the next question for you.
- Context on demand: index your wiki or business-logic docs so answers use your definitions.
Why Insiro, not the alternativesGeneric chat-with-your-data tools send the whole schema to a model and hope. Datasight grounds every query in a discovered semantic layer, validates the SQL before it runs and shows its confidence, so a wrong answer looks wrong.
Auto ontology discovery
Your knowledge graph, discovered for you.
Connect a database, tell Datasight which ERP or CRM it belongs to, and it maps tables, relationships, metrics, dimensions and PII. What used to be a quarter-long modelling project becomes a background job.
- Dependency analysis recommends the related tables you forgot to select.
- Business context from SAP, Salesforce, NetSuite and other ERP/CRM catalogues.
- Editable YAML with versioning, reset, download and scheduled refresh.
- PII flags carried into every agent and dashboard that uses the layer.
Why Insiro, not the alternativesBI tools expect you to model data before anyone can ask a question. Datasight inverts that: it discovers the model, you correct it, and the same layer then serves agents, AI teammates and dashboards.
Join paths inferred from keys and values Metrics and dimensions labelled PII flagged and masked Business names from NetSuite
Under the hood
What happens between your question and the answer?
Ten checked steps: the semantic layer is loaded, ambiguity is resolved, the query is planned, written, translated for each database, validated (and repaired if needed), limited, run read-only, scored for confidence and explained. Click any step to see what it did on a real question.
Step 1 of 10 · 12 ms
Load schema
Loaded 8 tables, 37 joins and 12 metrics from the YAML layer for sales_dw and billing.
Step 2 of 10 · 31 ms
Detect ambiguity
Two readings found. The semantic layer defines revenue as net_amount, so net is used and the choice is shown to you.
Step 3 of 10 · 44 ms
Plan query
Plan: orders ⋈ regions, filter this quarter vs last, group by region, compute QoQ growth.
Step 4 of 10 · 120 ms
Generate SQL
SQL generated from the plan with LAG() over quarters. Model routed: reasoning tier.
Step 5 of 10 · 9 ms
Transpile dialect
Date functions rewritten for each engine: DATE_TRUNC for both, QUALIFY removed for Postgres.
Step 6 of 10 · 18 ms
Validate SQL
First pass failed: column region_nm does not exist. Auto-repaired to region_name and re-validated.
-- before
SELECT region_nm … ✗ unknown column
-- repaired
SELECT region_name … ✓ valid
Step 7 of 10 · 2 ms
Inject row limit
LIMIT 200 added. Estimated scan: 1.2 GB, inside the per-query budget.
Step 8 of 10 · 210 ms
Execute
Ran on Snowflake (184 ms) and Postgres (26 ms); results joined in the session.
Step 9 of 10 · 6 ms
Score confidence
High: unambiguous after step 2, all joins on declared keys, row counts match last refresh.
Step 10 of 10 · 40 ms
Interpret
Bar chart chosen for 6 categories. Story written, 3 follow-up questions suggested.
492 ms end to end · 1 self-repair · read-only
Insiro BI Studio
One English sentence. A full dashboard in seconds.
Insiro BI Studio is an alternative to Power BI, Qlik Sense and QlikView: describe the dashboard in one English sentence and get it in seconds, with no sprint planning, no data modelling backlog and no visual-by-visual build. Type "executive overview of Q3 sales" and pick a template. Datasight plans the pages, writes a question per visual, runs them two at a time and saves as it goes. An expert pass then fixes chart choices and writes the narrative.
- Templates for executive, sales, operations and finance reporting.
- Edit in words: "make this a line chart", "new page for churn", "undo".
- Publish and share a read-only link; refresh re-runs the saved questions.
- Power BI import: rebuild .pbix reports on your live warehouse with DAX translated to SQL.
Why Insiro, not the alternativesPower BI, Qlik Sense and QlikView make you model data and build visuals one by one, often across a sprint. and AI add-ons summarise dashboards you already built. Datasight builds the dashboard from your intent, and every tile stays a question you can read and re-run.
ARR
$61M
▲ 9%
Net retention
118%
▲ 3pt
CAC payback
14 mo
▲ 1 mo
Churn
1.9%
▼ 0.4pt
Revenue by month
Pipeline by stage
Data sources
Which data sources does Datasight connect to?
Datasight connects to relational databases, cloud warehouses, lakehouses, NoSQL stores, search indexes, Excel workbooks and PDFs, and answers plain-English questions over each in its own dialect. Connections are read-only, credentials are encrypted at rest, and you can ask one question across several sources at once.
Oracle
Natural-language search across Oracle tables, views and schemas.
PostgreSQL
Ask questions of PostgreSQL data without writing a query.
MySQL
Turn a business question into a MySQL answer, with source attribution.
TiDB
Search distributed TiDB clusters in natural language, fully on-premises.
Azure SQL
Self-serve reporting over Azure SQL Database, no SQL expertise needed.
AWS Redshift
Query Redshift clusters in plain English, no warehouse tuning needed.
IBM Db2
Natural-language search across Db2 tables, on-premises or in the cloud.
Kusto
Ask questions of Kusto log and telemetry data. No KQL required.
Microsoft Fabric
Search your Fabric warehouse without hand-building a semantic model.
Power BI
Query the data behind your Power BI models in natural language.
Teradata
Self-serve analytics on Teradata without the BI request queue.
Snowflake
Ask your Snowflake data anything in plain English, with cited sources.
Databricks
Natural-language queries across your Databricks lakehouse tables.
BigQuery
Explore BigQuery datasets by asking, not by writing SQL.
Cassandra
Natural-language search over Cassandra. No CQL required.
DynamoDB
Ask questions of your DynamoDB tables without code.
Cosmos DB
Search Azure Cosmos DB collections in plain English.
Elastic
Turn Elasticsearch indexes into clear, sourced answers.
Excel
Search across Excel workbooks like a queryable database.
PDFs
Pull structured answers out of data trapped in PDF documents.
Federated sessions
Select several sources and ask one question across them. Each source keeps its own dialect and semantic layer.
Read-only by design
SELECT-only validation and injected row limits. Datasight cannot write, update or drop anything.
Full transparency
Every answer shows its SQL, the retrieved context, the policy decision and a numbered pipeline trace.
Why Insiro
Datasight vs. the alternatives.
| Capability | Insiro Datasight | Traditional BI tools | Chat-with-your-data add-ons |
|---|---|---|---|
| Semantic layer | Discovered automatically, editable | Modelled by hand first | Schema dump into a prompt |
| Question to answer | Plain English, multi-turn | Drag-and-drop or DAX | Plain English |
| SQL validation and confidence score | 10-step pipeline | Not applicable | Rarely shown |
| Dashboards from a brief | BI Studio, in seconds | Built visual by visual (Power BI, Qlik) | No |
| Federated queries across databases | Multi-source sessions | Via extracts or a data model | Usually one source |
| Database coverage | 30+ dialects | Broad | A few warehouses |
| Feeds agents and AI teammates | Same layer everywhere | Separate | Separate |
What is Insiro Datasight?
Insiro Datasight is a natural-language analytics product. Business users ask questions in plain English and get validated SQL, charts, KPIs and a written explanation, then turn answers into multi-page BI dashboards. It supports 30+ database and warehouse dialects and automatically builds a semantic layer from your schema.
How accurate is Datasight's natural-language-to-SQL?
Every question runs a 10-step pipeline that detects ambiguity, plans the query, generates and validates SQL in your database's dialect, and scores confidence. Answers show a confidence badge, the exact SQL and the context used, so analysts can verify, edit and re-run any query. Ambiguous questions are flagged instead of guessed.
Which databases does Datasight support?
Datasight supports 30+ SQL and warehouse dialects, including SQL Server, Oracle, PostgreSQL, MySQL, TiDB, Azure SQL, Amazon Redshift, IBM Db2, Teradata, Snowflake, Databricks, BigQuery and Microsoft Fabric, plus Kusto, Power BI models, Cassandra, DynamoDB, Cosmos DB, Elastic, MongoDB, Excel workbooks and PDFs.
What is auto ontology discovery?
When you connect a database, Datasight inspects the tables you select, recommends related tables, and writes a semantic layer in YAML: descriptions, metrics, dimensions, identifiers, relationships and PII flags. Telling it your ERP or CRM (for example SAP, Salesforce or NetSuite) improves the business descriptions. You can edit, version and schedule refreshes.
Can Datasight query across multiple databases?
Yes. You can add several connected sources to one session and ask questions that span them, for example orders in Snowflake and customers in PostgreSQL. Datasight uses each source's semantic layer to plan the federated query and returns one answer with the SQL for each source.
Can Datasight change or delete data?
No. Datasight runs read-only: generated SQL is restricted to SELECT, a row limit is injected, and connection credentials are encrypted at rest. Columns flagged as PII in the semantic layer can be masked, and every query is logged with the user, SQL and result size.
Can I import existing Power BI reports?
Yes. Upload a .pbix, .pbit or zipped .pbip file and Datasight rebuilds the report on your connected database, translating DAX measures to SQL and rating each visual as verified, exact, approximate or needs review. It never uses data stored inside the file, only your live warehouse.
Point Datasight at one database. Ask it something hard.
In a 30-minute session we connect a read-only replica, let Datasight discover the schema and answer your team's real questions.
- 30+SQL and warehouse dialects
- 10validation steps per question
- 0write access to your data