The genie of your business. It explains, recommends and acts.
Marut AI is native to the platform, not bolted on. Like a genie for the whole enterprise, it answers what happened, finds out why, forecasts what is next, recommends what to do and does it when you say so. On your own model, inside your own walls, with every answer cited.
Claims in the West rose 18%, and most of that came from Ahmedabad. Three large flood claims arrived in one week.
Opening Claims Register (Monthly) for 1 to 30 September 2026. I will run it as a password-protected PDF, like last time. Send it to the claims team as well?
At the current pace, 14 of 412 dealers will miss target. Nine of them are in the North, and the shortfall sits in two models.
From what happened to what to do
Most tools stop at the chart. Marut keeps going.
Why
Finds the root cause behind every change, not just the chart that moved.
Three flood claims from one surveyor drove 62% of the rise.How
Shows how it happened: the branches, products and people behind it, ranked.
Ahmedabad explains 81% of the increase.When
Forecasts when it will matter, with how sure it is.
At this pace, 14 dealers miss target by the 25th.What to do
Recommends the next step, in plain words.
Start a review. Set an alert at 75%.Do it
Runs the report, sends the brief, starts the approval. After you say yes.
“Claims report kholo.” Done.You stay in control. Marut asks before it changes anything, every answer shows its source, and it never sees more than you are allowed to.
Grounded in your semantic layer.
The copilot answers from your governed definitions, not raw tables: what each measure means, how it rolls up, what people call it and who may see it. That is the difference between a chatbot that sounds right and a copilot that is right.
- Accurate, cited answers
- Smaller prompts and lower AI cost
- Synonyms, so people ask naturally
- Permissions enforced in every answer
Five wishes, one genie
Ask the way you would ask a genie who has read every report and knows what helps the business grow.
“How did West do last month?”
Instant answers from your governed numbers, with the chart and the source.
Ask Marut →“Why did claims go up?”
Root cause in plain words: the branches, products or people behind the change, ranked by contribution.
Insights →“Will we hit target?”
Forecasts with confidence bands, anomalies flagged early, drivers explained.
Forecasts →“What would you do?”
Recommended next steps: start a review, set an alert, send a statement, move stock.
Recommendations →“Claims report kholo.”
Commands by voice or text, workflows drafted from a sentence, every action confirmed first.
Voice & Commands →Intelligence in every part of the platform
From conversational answers to machine learning, all permission-aware and cited.
Ask in plain language
Questions answered from your semantic layer, so the numbers match your dashboards. Ask for root cause, forecasts and next actions.
Voice in English and local languages
“Open the sales dashboard.” “Make it a line chart.” “Top 10.” “Filter APAC.” Speech is recognised on your own servers and audio is never stored.
Ask your documents
PDFs, scans, Word, Excel, PowerPoint, e-mail and images. Every sentence of the answer is cited to the page and the line is highlighted.
Hybrid answers
Marut joins a governed KPI with the contract clause or policy that explains it, in one answer.
Forecasts with confidence bands
See where a number is heading, and how sure the model is.
Anomaly detection
Unusual values flagged automatically, before they become problems.
Key drivers, explained
SHAP-based driver analysis shows what actually moves a metric.
What-if analysis
Change an assumption and see the impact before you commit.
Describe a process, Marut builds it
Explain an approval in plain words and Marut drafts the workflow for you to review.
Ask. Get the answer, the reason and the next step.
Ask Marut is grounded in your semantic layer, so it uses the same definitions as your reports. Ask what happened, why it happened, what will happen next and what to do about it.
- Root-cause analysis in plain language
- Forecasts and suggested next actions
- Briefs on mobile and WhatsApp
Ask Marut
WorkspaceCommands onRun Claims Register (Monthly) · Period: 1 Mar to 31 Mar 2026 · Format: Excel
Every document, answerable. Every answer, cited.
Upload, read, index, search and answer, inside your own environment. Keyword and meaning search with re-ranking finds the right passage, and every sentence links to its page.
- PDF, scans, Word, Excel, PowerPoint, e-mail, HTML, images
- Document answers as a dashboard widget
- Full Q&A log for audit
Motor_Policy_Wording_2025.pdf
OCR · 42 pagesPage 14| Add-on | Premium | Covers |
|---|---|---|
| Engine Protect | 1,250 | Water ingress |
| Zero Depreciation | 2,980 | Parts |
| Roadside Assist | 399 | Towing |
Document Intelligence, measured
Results from our offline evaluation on 42 questions, CPU only (October 2026).
Sovereign AI you can trust
Your own model, on your servers
Run Ask Marut on your own LLM on-premises, or with no cloud AI at all.
Grounded, cited, honest
Answers come from governed data or cited pages. If the answer isn't there, Marut says “not found”.
Permission-aware
Only quotes documents and data the person asking is allowed to see.
Prompt log and consent gate
Every AI question is logged. A data-consent gate and redaction mode protect sensitive fields.
Bring your own model, or none
Choose the AI provider per workspace: OpenAI, Azure OpenAI, Anthropic, Gemini, DeepSeek, OpenRouter or a local model on your own servers. Commands, OCR and document search need no model at all.
Coming next: agentic AI
AI agents that carry out multi-step work across the platform, with a human approving every important step.
See Marut AI on your own data.
Bring your questions and your documents. See it on your own data in 2 to 3 days.