A semantic layer first
Before any AI, we write down your metrics with your finance and business teams: what counts as an active customer, how returns hit revenue, which date a collection belongs to. The agent can only use these agreed definitions.
Build
Ask your data a question. See the working.
Leaders ask questions of their own sales, collections, inventory and operations data in plain English or Hindi. The agent answers with the number, the chart, and the exact query it ran, using the metric definitions your finance team has signed off.
- Typical length
- 6 to 10 weeks
- Indicative price
- ₹8 lakh to ₹30 lakh, excluding GST
- Engagement path
- Discover→Prove→Build→Run
Sound familiar?
Typical situations, not quotes from clients.
- Every question from the MD becomes a ticket to the MIS team, and the answer comes back on Thursday.
- We have four dashboards that show four different revenue numbers.
- I don’t want a chatbot that invents numbers. I want the right number, fast.
Dashboards answer the questions someone thought of last year. New questions wait in a queue. And a tool that writes its own queries without knowing what “net revenue” means at your company will give confident, wrong answers.
What we build
Questions in plain language
“Which ten distributors fell furthest behind target in Q2, and why?” “Collections efficiency by branch for September, compared with August.” English, Hindi or Hinglish.
Answers you can check
Every answer shows the query, the tables used and the filters applied. Analysts can open it, edit it and save it.
Charts and short explanations
The right chart for the question, and a plain summary of what changed. No made-up narrative.
Row-level permissions
A regional head sees their region. Permissions follow your existing data access rules.
Where leaders already are
A web app, Microsoft Teams, Slack, or a scheduled WhatsApp or email digest.
How it fits together
Where it fits
- PostgreSQL
- MySQL
- Microsoft SQL Server
- Oracle
- Snowflake
- Google BigQuery
- Databricks
- Amazon Redshift
- SAP data exports
- Tally data through connectors
- Zoho Analytics
- Power BI and Tableau datasets
- Excel and Google Sheets
If your data is spread across systems with no warehouse, we will tell you that first. Sometimes the right first step is a simple warehouse, not an agent.
How the engagement runs
Typical length: 6 to 10 weeks from pilot start to production.
Discover
1 to 2 weeksWe collect the 100 questions leaders actually ask, map where the data lives, and agree the first set of metric definitions.
Prove
4 to 6 weeksA fixed-price pilot for one domain, such as sales or collections, with a small leadership group. Scored against an evaluation set of real questions with answers checked by your analysts.
Build
2 to 4 weeksMore domains and users, permissions, the chosen channels, security review and training.
Run
monthlyWeekly review of wrong or unanswered questions, new metrics added, evaluation re-run on every change.
How we measure it
Before we build, we agree with you in writing:
Answer accuracy
Numbers match what your analysts produce for the same question.
Correct refusals
Questions outside the defined metrics are declined or routed to an analyst, not guessed.
Permission leaks
Zero tolerance, tested on every release.
Time to answer
From question to trusted number, compared with today.
Adoption
How many of the target users ask questions each week.
What you get
- A production analytics agent for the agreed domains
- A documented semantic layer of your metrics, useful far beyond this project
- Connections to your data sources with read-only access
- Evaluation set and test reports, yours to keep
- Admin screen for analysts to review and correct queries
- Runbook and training for leaders and analysts
- Source code and configuration in your repository
What we need from you
- Read-only access to the data sources, in a test or replica environment
- An analyst or MIS lead who knows the data, for about a day a week
- A finance owner to sign off metric definitions
- Five to ten leaders willing to use the pilot and give feedback
- Your data access rules
Indicative price
₹8 lakh to ₹30 lakh
excluding GST, from pilot to production.
The range depends on the number of data sources, the state of your data, the number of metrics and the permission model. The pilot is a fixed price quoted after Discover. 40% advance, the rest on milestones.
Model, API and cloud costs are passed through at cost plus a 10% handling margin when we hold the accounts (no margin if you pay the providers directly). We estimate the monthly running cost before you sign.
Frequently asked questions
Can the agent change our data?
No. It has read-only access. It cannot insert, update or delete anything.
What if it gets a number wrong?
The query is always visible, so analysts can spot errors. Wrong answers go into the weekly review and become test cases.
Does this replace our BI tool?
No. Dashboards remain the right tool for numbers you track every day. The agent handles the questions dashboards do not cover.
Our data is messy. Can you still do this?
Within limits. We will show you in Discover which questions your current data can answer reliably and which it cannot.
Is our data sent to a model provider?
Table structures and metric definitions are used to write queries. Row-level data can stay in your database. Where results are summarised by a model, we can host that model in your environment.
Bring the five questions your MIS team answers every week.
A 30-minute call with a founder. We will tell you whether your data can support an agent today.