01 / THE REALITY
One messy sheet.
One important decision.
No analyst in the room.
Hisaab
An AI-powered financial companion that turns everyday sales data into calm, practical business guidance.
- ROLE
- Product Designer
- SCOPE
- Product thinking · UX · UI
- PLATFORM
- Web app
- TEAM
- Independent
CONTEXT / 02
Small retailers make big decisionswith very little decision support.
Pricing, delivery fees, promotions and retention choices can directly affect orders — yet many small retailers are still working from instinct, scattered spreadsheets, and no analyst support.
The opportunity wasn’t to build another dashboard. It was to help a shopkeeper ask a real business question, use the data they already have, and get an answer they can actually trust.
02 / THE CHALLENGE
The challenge was not generating an answer.
It was generating one worth trusting.
APPROACH / 03
One system understands the question.
Another computes the answer.
explain it.
don’t invent it.
“If I increase the delivery fee by ₹20,
what will happen to repeat orders?”
01 / LANGUAGE BRAIN
Understands the question
- Reads messy columns
- Interprets the business question
- Explains the verified result
- Responds in the user’s language
02 / MATH BRAIN
Computes the answer
- Aggregation
- Regression
- Projection
- Range and confidence
LIKELY IMPACT
+₹22.7K / month- CONFIDENCE
- Medium
- RANGE
- ₹14K — ₹30K
The AI can explain the number.
It cannot change it.
HOW IT WORKS / 04
From one messy question
to one decision-ready answer.
Hisaab takes the data a retailer already has, understands its structure, computes the likely impact, and turns it into an answer they can act on.
01 / CONNECT
Use the data you already have.
Google Sheet · CSVNo technical setup
02 / SCAN
The system reads what’s there.
Columns · formats · gapsBefore any answer is attempted
03 / ASK
Ask in your own words.
Text or voiceEnglish · Hindi · Hinglish
04 / COMPUTE
Math calculates the impact.
Projection · range · confidenceINPUT SIGNALS
ordersrepeat_ordersavg_order_valuedelivery_fee9 months
REGRESSION / COMPUTATION
projectionrangeconfidence
LIKELY RESULT+₹227
LOSS-₹712
LIKELY+₹227
GAIN+₹1,166
High confidence · 90%
Calculated from your history.
Not generated by AI.
05 / EXPLAIN
Gemini explains the verified result.
Hisaab showing a clear next step while honestly flagging missing promotion data
TRUST GUARDRAIL / 05
Sometimes the most trustworthy
answer is “not enough data.”
A weak signal shouldn’t become a confident recommendation.
When Hisaab can’t support a question reliably, it explains what’s missing and points the retailer toward questions the data can actually answer.
A DESIGN PRINCIPLE
No fake precision.
No confident guessing.
uncertainty is part
of the answer.
LEARNINGS / 06
Designing for trustchanged how I thoughtabout AI products.
Hisaab made one thing clear: useful AI isn’t about sounding confident.
It’s about knowing what the data can support, showing uncertainty honestly, and giving people enough clarity to make a decision themselves.
TRUST BEFORE INTELLIGENCE
The product became more useful when it stopped trying to answer everything.
A clear limitation was often more valuable than a confident guess.
SEPARATE EXPLANATION FROM CALCULATION
AI was useful for understanding messy inputs and explaining results.
The actual numbers still needed a deterministic source of truth.
DESIGN FOR IMPERFECT DATA
Real small-business data is messy, incomplete and inconsistent.
The experience had to work with that reality instead of assuming perfect spreadsheets.
good AI knows
when not to answer.