Governed intelligence,
designed for decisions.
A product UX framework for natural-language analytics, safe query execution, explainable insight and accountable action.
Product & UX Overview
DataMind is a governed data-intelligence assistant. Natural language is the entry point; validation, permissions, provenance and user control shape every outcome.
Governed intelligence pipeline
Personas
The experience adapts to analytical maturity and responsibility without fragmenting the product into separate tools.
Business User
Low–medium- Primary goals
- Ask without SQL; understand charts; save dashboards
- Important UX needs
- Trust, plain language, Persian/English
Data Analyst
High- Primary goals
- Generate and debug SQL; reuse metrics
- Important UX needs
- Review controls and technical detail
BI Admin
High- Primary goals
- Permissions; approvals; usage governance
- Important UX needs
- Auditability and explicit confirmation
Data Engineer
High- Primary goals
- Schema, glossary, metrics and approved joins
- Important UX needs
- Precise definitions and change ownership
System Admin
High- Primary goals
- Deployment, models, security and monitoring
- Important UX needs
- Environment visibility without secret exposure
Accountability beyond end users
Owns product scope, delivery and technical quality.
مالک محدوده محصول، تحویل و کیفیت فنی
Primary internal sponsor and delivery stakeholder.
حامی داخلی و ذینفع اصلی تحویل محصول
Owns data definitions, Metabase governance and rollout.
مالک تعاریف داده، حاکمیت Metabase و استقرار
Defines PII, external inference and direct-access policy.
تعریفکننده سیاست PII، پردازش بیرونی و دسترسی مستقیم
Validate usefulness, clarity, latency and correction behavior.
اعتبارسنجی کاربردپذیری، وضوح، سرعت و اصلاح نتایج
Business user - Ask and understand
User Goals & Jobs
Design around the decision a person is trying to make - not around the model or connector doing the work.
Business User“I want to ask a business question and understand the answer without knowing SQL.”
Data Analyst“I want DataMind to generate and validate a query so I can focus on analysis.”
BI Admin“I want generated dashboards confirmed before they become shared assets.”
Data Engineer“I want approved terminology and relationships to be used correctly.”
System Admin“I want to control where models run and how data leaves the environment.”
Bilingual user-story catalogue
Ask in Persian and preview a validated Metabase question without writing SQL.
پرسش فارسی و مشاهده پیشنمایش سؤال اعتبارسنجیشده Metabase بدون نوشتن SQL.
Use English or mixed Persian-English business terminology.
استفاده از واژگان انگلیسی یا ترکیبی فارسی–انگلیسی.
Inspect and edit SQL before saving a Metabase question.
بررسی و ویرایش SQL پیش از ذخیره سؤال در Metabase.
Save generated content into a selected Metabase collection.
ذخیره محتوای تولیدشده در collection انتخابی Metabase.
Create dashboards from a high-level request after confirmation.
ساخت داشبورد از درخواست سطحبالا پس از تأیید.
Register metrics and synonyms used in provenance.
ثبت معیارها و مترادفهایی که در منشأ نتیجه نمایش داده میشوند.
Choose local CPU, local GPU or external API inference.
انتخاب استنتاج CPU محلی، GPU محلی یا API بیرونی.
Audit prompts, SQL, models, latency and created objects.
ممیزی prompt، SQL، مدل، latency و objectهای ساختهشده.
Receive clarification when a request is ambiguous.
دریافت سؤال شفافساز هنگام ابهام درخواست.
Review bounded correction when SQL fails.
بررسی اصلاح محدود هنگام شکست SQL.
Use DataMind Suite identity beyond the Metabase connector.
استفاده از هویت DataMind Suite فراتر از اتصال Metabase.
Receive a direct insight through an approved read-only connection.
دریافت بینش مستقیم از اتصال فقطخواندنی تأییدشده.
Publish a reviewed direct insight to Metabase.
انتشار بینش مستقیم بازبینیشده در Metabase.
Govern Direct DB independently from Metabase.
حاکمیت مستقل Direct DB از Metabase.
Information Architecture
Separate asking, reusable outcomes, organizational knowledge, governance and administration.
Information Architecture
Core User Flows
Every flow preserves intent, validates generated work and introduces confirmation before consequential creation.
Core natural-language query
Dashboard creation flow
Natural Language Query UX
Persian, English and mixed-language prompts resolve into visible intent, assumptions, query and next actions.
نمودار فروش ماهانه استان تهران در ۱۲ ماه گذشته را بساز
AI Interaction Model
Communicate progress and decision points without exposing chain of thought or infrastructure noise.
AI interaction state machine
Clarification & Ambiguity Handling
When multiple valid meanings exist, DataMind asks a targeted question instead of silently selecting one.
Clarification flow
دو تعریف معتبر در مدل داده پیدا شد. برای ادامه یکی را انتخاب کنید.
SQL Generation & Validation UX
SQL is collapsed by default, reviewable on demand and never executable when unsafe or unvalidated.
SQL generation and validation
Metabase Interaction UX
Metabase is a connector. DataMind owns intent, validation, confirmation and the cross-source experience.
Metabase interaction
Dashboard Creation UX
Dashboard creation moves from proposal to review to explicit confirmation before an asset is created.
Dashboard lifecycle
Executive sales overview
4 cards · Metabase · Private until confirmed
- 1Monthly RevenueValidated
- 2Revenue by ProvinceValidated
- 3Top ProductsValidated
- 4Monthly GrowthValidated
Direct Database Insight UX
An optional, approved and policy-bound mode for bounded read-only database analysis.
Direct Database Insight · Optional
Direct Database Insight
3 grounded insights Net revenue grew in 7 of 9 provinces. Tehran contributes the largest absolute increase.
Insight & Visualization UX
Computed facts and AI interpretation must remain visibly distinct.
Revenue increased compared with the previous period.
Computed from executed resultThe increase appears concentrated in Tehran.
Interpretation · Review source detailsProvenance & Trust UX
Every important result can explain where it came from, how it was validated and when it was produced.
Evidence and execution details for the current answer.
Feedback & Correction UX
Feedback should describe what was wrong and offer a direct recovery path.
States & System Behavior
Consistent state language prevents ambiguous loading, silent failure and premature success.
Assistant state machine
Dashboard lifecycle
Error Handling
Errors explain the outcome, preserve useful context and present the safest next action.
The generated query did not pass validation. No query was executed.
This data is unavailable under your current permissions.
The query completed successfully but returned no records.
Metabase is temporarily unavailable. Your request is preserved.
Permissions & Security UX
Security is visible enough to establish trust without overwhelming routine analysis.
Safe by default
Normal analysis is read-only, permission-aware and bounded before execution.
Audit & Governance UX
Administrators can trace requests, retrieved context, validation, execution and resulting actions.
Audit log
Governed trace of analytical activity
Localization & RTL
Persian is a first-class product language; technical content keeps the direction it requires.
Net revenue in Tehran province
Review net revenue using the approved metric definition.
SELECT net_revenue FROM fact_sales;Responsive Behavior
The experience preserves intent, review and confirmation across desktop, tablet and mobile.
Desktop
Persistent navigationSide-by-side reviewFull tables and chart contextTablet
Collapsible navigationStacked detail panelsScrollable data tablesMobile
Drawer navigationBottom-safe composerProgressive SQL and chart detailUX Principles
A concise contract for every new workflow and component in DataMind.
- Ask, don’t guess.
- Validate before execution.
- Explain important decisions.
- Keep the user in control.
- Make data provenance visible.
- Separate facts from interpretation.
- Be safe by default.
- Respect permissions.
- Progressively disclose technical detail.
- Treat Persian and English as first-class.
- Confirm before consequential actions.
- Make failures actionable.