01
Ask in everyday language
Users can ask questions like 'Why did revenue slow down last week?' or 'Which stores missed their target?' — no need to learn reporting tools first.
Reporting made simple for everyone
Ask your business questions. Get answers instantly.
No SQL, no spreadsheets, no waiting for analysts. Just type your question in your own language, and Bilge fetches the answer from your database — with full transparency into how it got there.
bilge
>
|
01
Question reviewed — intent: margin analysis
02
SQL generated — SELECT with JOIN on orders, customers
03
Safety validated — read-only, row limit applied
04
Executed in 0.34s — 12 rows returned
05
Answer ready — 3 customers drove 68% of margin drop
Designed to simplify reporting
Ask a question, get an answer, share it.
That’s it.
No training, no dashboards to configure, no SQL to learn. Your team goes from a plain-language question to a shareable report in minutes.
02
Get a clear report
Bilge returns summaries that non-technical users can easily understand, with clean tables and chart-ready outputs.
03
Ask follow-up questions
Compare periods, drill into a region, or ask 'why' questions; no need to build a new report from scratch.
04
Share results effortlessly
Export, e-mail, favorite, and visualize — all on the same screen so useful reports don't stay with just one person.
For everyone on your team
Anyone with a question gets an answer.
Managers
Stop waiting for analyst reports. Ask questions like "How did Istanbul perform last week?" and get answers in seconds — in your own language.
Analysts
Skip the repetitive queries. Let Bilge draft the SQL while you focus on the insight. You can always review and adjust the generated query.
Business owners
Track revenue, spot trends, and make informed decisions without depending on technical staff. Your data, your language, your pace.
The 5-stage agentic workflow
Every answer passes through visible checkpoints.
Unlike black-box AI tools, Bilge makes every step inspectable. You always see what happened and why.
01
Question review
Detect language, assess intent, refine ambiguity, reject unsafe requests.
02
SQL generation
Translate business language into dialect-aware SQL with workspace context.
03
Safety validation
Apply row limits, timeout constraints, and schema rules before execution.
04
Execute & recover
Run the query, capture failures, analyze errors, retry when justified.
05
Final answer
Return inspectable results with optional narrative and chart-ready outputs.
Proactive insight discovery
Answers to questions you haven't asked yet.
Bilge quietly profiles your workspace data, proposes hypotheses, and verifies every single one against real numbers before you see it. No hallucinated findings — only insights that survive the math.
- Anomalies, trends, correlations, and distribution shifts — surfaced automatically
- Every insight is numerically verified against your database before it reaches you
- One click turns an insight into a chat question or a recurring report
- Opt-in per workspace — nothing is scanned without your consent
Before the first question
It reads your data critically, not credulously.
An answer is only as sound as the table underneath it. Bilge profiles every workspace it connects to, scores what it finds, and carries the caveats into the SQL it writes.
Nine quality checks, per column
Empty and constant columns, null rates, duplicate keys, negative measures, future and placeholder dates, tables that stopped updating, and orphaned references — each raised only above a materiality floor, so the list stays worth reading.
A score you can argue with
One health score per workspace, with the findings that cost it points listed underneath. No black box: every 2.5 points of severity halves the score, and fixing anything moves it.
Relationships found for you
Declared foreign keys are read from the database; the rest are inferred from naming conventions and graded by confidence. The join paths become prompt context, so questions that span tables are answered correctly without anyone drawing a diagram.
- Findings become caveats in the generated SQL, not just a report
- Profiling is read-only — nothing is written back to your database
- Re-run whenever the schema changes; resolved findings raise the score
Wherever you work
Ask from WhatsApp. Get reports in Slack.
Bilge now lives in the messaging tools your team already uses. Ask questions in natural language, get governed answers back — and have scheduled reports delivered right into the chat.
- Two-way Q&A over WhatsApp, Slack, Microsoft Teams, and Google Chat
- Scheduled reports fan out to e-mail and your linked channels — with Excel attachments
- Secure account linking: verified numbers, one-time link codes, per-user enablement
- Admins decide which channels are on — per company and per user
WhatsApp
Slack
Microsoft Teams
Google Chat
E-mail
No database required
Try Bilge on realistic data in one click.
Create an organization and start asking questions straight away. Every trial workspace opens on a ready-made dataset — no database connection, no API key, nothing to install.
Anadolu Perakende — Retail Chain
Two years of sales, returns and campaign data for a fictional retail chain with stores in 12 Turkish cities. Realistic seasonality, VAT rates and TRY amounts.
- What was total revenue last month?
- What are the top 10 best-selling products this year?
- Show total sales by city
Ege Danışmanlık — Invoice Ledger
Two years of e-invoice ledger data for a fictional consultancy: 900 clients, invoices, collections, aging and per-consultant billing.
- How did monthly invoiced totals trend this year?
- What is our overall collection rate?
- What is the total overdue outstanding balance?
Bulut Panel — SaaS Subscription Metrics
Three years of subscription economics for a fictional B2B SaaS: 3,600 accounts, MRR movements (new/expansion/contraction/churn), plan mix and monthly usage.
- What is our current total MRR?
- Show the monthly MRR trend for the last 12 months
- What is our monthly MRR churn rate?
- Sample data is read-only — nothing you do can change it
- Each dataset ships with a glossary, metric definitions and 30 example questions in Turkish, English and German
- Connect your own database whenever you are ready — the trial workspace stays separate
Enterprise-ready features
Not a chatbot. A governed analytics agent.
Ask in your own language
Ask in your own language — Turkish, English, German, and more. No SQL needed, but the SQL is always visible for those who want it.
Safety-first execution
Every query passes through validation gates — row limits, schema constraints, timeout rules — before touching data.
Domain-aware context
Define glossaries, business metrics, and sample questions so the AI speaks your domain from the start.
Visualize and share
Generate charts, export results, email answers, save favorites, and set up recurring reports.
Your AI, your data
Use OpenAI, Claude, Gemini, or Grok in the cloud. Or run Ollama on your own servers so no data ever leaves your network.
Cloud or on-premise
Deploy on your own infrastructure or use our managed cloud. Same product, same features, your choice of environment.
Open data API
Expose governed answers to partner systems through a secure REST API. SQL is generated once by the agent, then cached and replayed — no LLM cost per call.
Usage analytics & audit
Tenant dashboards track questions, engagement, and quotas day by day. Deleted chats stay auditable, and every action leaves a trail.
Cost-efficient by design
Prompt caching, token budgets, and reusable SQL templates keep AI spend predictable — per question and per company.
Broad database reach
One interface for your entire data estate.
Connect to the databases and AI providers your organization already uses. No migration needed.
Supported databases
AI providers
Ready to try?