The data analyst assistant built for Persian
Ask your data.Get exact numbers, with sources.
ASKx understands questions in everyday Persian or English, runs the calculation on your organisation's own data, and writes an answer in which every number cites its source. No report building, no formulas, and no data leaving your organisation.
- Every number sourced
- Runs on your servers
- Local AI models
Which region grew sales the most last quarter?
- Identify the metric and period
- Compute on the sales data
- Check every claim against evidence
The North region grew most, up 18% on the previous quarter, followed by the West at 11% and Central at 4%.
Questions you can ask every day
- How did our market share change over the last three months?
- Which product's sales fell the most, and why?
- How does our price compare with competitors in each channel?
- Has ad spend moved with sales?
- Forecast next quarter's sales
- Which customers stayed loyal and which ones left?
- In this PDF, what is the total cost of phase two?
- Where is our biggest growth opportunity?
What is ASKx?
ASKx is an AI data analyst assistant that managers and analysts question in Persian (and English) about their business data. The language model never produces the numbers: governed tools compute them from the organisation's own data, and a validator matches every number in the answer to its source before it is shown. ASKx runs entirely on the organisation's servers with local open-source models.
For every business that decides with data
ASKx is not tied to one industry. Wherever you have tabular data and recurring questions about it, ASKx answers them, in Persian.
Retail and consumer goods
Market share, distribution, price and brand performance across channels.
“Which brand gained share in chain stores?”
Banking, insurance and finance
Deposits, loans, premiums and customer behaviour by branch and region.
“Which branches grew deposits the most?”
Telecom and technology
Subscribers, churn, revenue per user and service usage.
“In which province did churn rise?”
Pharma and healthcare
Drug sales, share by therapeutic class and pharmacy coverage.
“What is our share of the antibiotics class?”
Manufacturing and industry
Output, inventory, dealer sales and raw material prices.
“Which product's inventory is above average?”
E-commerce
Orders, baskets, campaigns and repeat purchase rate.
“What did last month's campaign do to orders?”
Energy, oil and petrochemicals
Production, exports, product prices and unit performance over time.
“Which product's exports fell year on year?”
Market research and advertising
Ad monitoring, media spend, share of voice and competitor benchmarks for clients.
“What was this brand's share of voice on TV?”
Why you can trust ASKx's answers
General chat assistants guess numbers. ASKx is built for real decisions and keeps every answer checkable.
Questions in everyday language
Ask the way you would ask an analyst. ASKx recognises names, periods and intent, and asks a clarifying question instead of guessing when a question is ambiguous.
Every number has a source
The language model never makes up, rounds or adds numbers. Governed tools compute them from your data, and a validator matches each number in the answer to its evidence.
Evidence kept apart from hypothesis
When explaining why something changed, each driver is labelled as in the data or a hypothesis. ASKx never turns two trends moving together into a causal claim.
Forecasts that pass their test
A forecast is published only when there is enough history and the model beats a naive baseline in rolling backtests; otherwise ASKx says why.
Ask about documents and images
Attach a PDF, Word, Excel or PowerPoint file, or an image. ASKx reads the text, reads images with Persian and English OCR, and quotes answers with page references.
Conversations that keep context
Follow-up questions keep earlier context, you can reply to a specific message, and every conversation stays in your history.
Tables, evidence and limitations
Answers come with downloadable tables, an evidence panel and their assumptions and limitations; when the data cannot answer, ASKx says so plainly.
Ready-made analyses
Standard methods such as competitor benchmarks, pricing and promotion, advertising effect and opportunity gaps are ready-made tools for the assistant.
Analyses ASKx is ready to run
Each analysis is a documented, tested method, not text the model makes up.
- Competitive benchmark
- Brand position
- Head-to-head comparison
- Change over time
- Pricing and promotion
- Advertising versus sales
- Advertising overview
- Customer loyalty
- Opportunity gap
- New product detection
- ROI scenario
- Backtested forecast
From raw data to a trustworthy answer in four steps
Setting ASKx up is a job for your technical lead; using it needs no technical knowledge.
Bring in your data
Upload Excel files from the admin dashboard. ASKx cleans them, runs quality checks and builds the analytical warehouse.
Define metrics once
A metric dictionary says how each metric is computed and at which levels it can be added up, so everyone sees one definition.
Ask in Persian
The assistant understands the question, calls the right analytical tool and asks for clarification when needed.
Check the answer
Answers come with cited evidence, tables, assumptions and limitations. You can trace every number back to its source.
Your data stays inside your organisation
ASKx is built for organisations that do not hand their data to outside services.
An admin dashboard that needs no technical skills
Admins configure everything from the browser, without editing files on the server.
- Users and roles: admin, analyst, marketer
- Data uploads and warehouse rebuilds
- AI model choice per task
- Connections to local model servers with a key and certificate
- Dictionary of brands, companies and synonyms
- Activity and audit log
Runs on your own servers
The app, the data and the AI model all run on your organisation's infrastructure.
Local open-source models
Local models are the default. External services can be used only when an admin explicitly changes the data egress policy.
Access control in the data layer
Each user's role and organisation come from the authenticated session and are enforced in the data layer, never in instructions to the model.
Audit and encryption
API keys are stored encrypted and important changes are recorded in an audit log.
ASKx next to the tools you know
ASKx does not replace dashboards; it answers the questions no dashboard was built for.
| ASKx | General chatbot | Traditional dashboard | |
|---|---|---|---|
| Free-form questions in Persian | Yes | Yes | No |
| Numbers from your own data, with sources | Yes | No | Yes |
| No report building in advance | Yes | Yes | No |
| Evidence kept apart from hypothesis | Yes | No | Partly |
| Backtested forecasts | Yes | No | Partly |
| Runs on your own servers | Yes | Partly | Yes |
Frequently asked questions
What does ASKx do?
ASKx is a data analyst assistant: it takes your question in Persian or English, computes the answer on your organisation's data, and writes a reply in which every number cites its source, with tables, assumptions and limitations.
Can the AI make up a wrong number?
In ASKx the language model produces no numbers. Numbers come only from computation tools or exact quotes from an attached document, and a validator checks every number against evidence before the answer is shown. An answer with an unsourced number is rejected.
Does our data leave the organisation?
No. ASKx is installed on your servers and by default works only with local AI models. An external service can be used only if an admin explicitly changes the data egress policy.
Is ASKx only for one industry?
No. Metrics are defined in a configurable dictionary, so ASKx works for retail, banking and insurance, telecom, pharma, manufacturing, e-commerce, energy and market research.
Which data sources does it support?
Today data comes from Excel files that admins upload in the dashboard. ASKx builds a cleaned analytical warehouse from them. Direct connections to databases such as SQL Server and PostgreSQL are on the roadmap.
Which AI models does it work with?
Any OpenAI-compatible model server such as vLLM, Ollama, LM Studio or llama.cpp, with open-source models such as Qwen. If an admin allows it, OpenAI, OpenRouter and Anthropic are supported too.
Can I ask about a PDF or an image?
Yes. Attach PDF, Word, Excel, PowerPoint files or images in the chat. ASKx reads text and tables, extracts text from images with Persian and English OCR, and answers with page references.
What if the data cannot answer my question?
ASKx says plainly that the data is not enough, that you lack access, or that the history is too short to forecast, instead of guessing.
How reliable are ASKx forecasts?
Every forecast is measured in rolling backtests before it is published, and is shown only when it beats a simple baseline. Otherwise the baseline value and the reason are shown.
Ask your first question today
See ASKx on your own data: every number sourced, on your own servers.