The glance
Total · Month-over-month · Meaning
Minimal effort
malMAL · PRODUCT DESIGN · OSAMA ALI · 2026
MAL · PRODUCT DESIGN TAKE-HOME
Designing Monthly
Spending Insights
as a Conversation
A progressive disclosure model for Mal’s mobile AI experience.Platform · Mobile app · iOS-first
Designed & presented byOsama Aliwww.os3li.com ↗
Use → to move through the story Some slides reveal in steps
Your June spending overview is ready.
THE REAL DESIGN PROBLEM
Financial data is multi-dimensional.
Chat is linear.
Scope · The assistant already understands the request. This concept focuses on how the answer is presented and explored.
Everything, all at once
→
How was my spending last month?Here’s what matters first.
The answer is not to remove useful data.
The answer is to reveal it at the right moment.
What happened?Why did it happen?What can I do next?
DESIGN PRINCIPLE
One answer. Three layers.
Concise detail
Categories · Values · Tap targets
Fast comparisonDeep dive
Patterns · Merchants · Transactions
Only when requestedControl when each layer appears — not how much data exists.
EXPERIENCE FLOW
A conversation with two useful paths.
ENTRYConversation startsUser asks, or opens the monthly overview
→SHARED ANSWERPrimary responseTotal · trend · key insight · tappable categories
USER-LED PATHCategory drill-downA tap becomes an implicit follow-upNew assistant message · context preserved
ASSISTANT-LED PATHProactive insightOne relevant observation after value is deliveredOptional Round-Up · easy to dismiss
Two possible paths. Neither is required to complete the other.
User-led exploration Assistant-led, earned moment
LAYER 2 · CONCISE DETAIL
Layer 2 makes the data scannable.
Ordered by spend
Exact values remain visible
Exact values remain visible
Rows earn a tap
Color is not the only signal.
Color is not the only signal.
malMal assistant
Here’s your June spending overview.
Total spentAED 8,450↓ 12% less than May
LAYER 3 · DEEP DIVE
Layer 3 appears only when requested.
Context preserved
Detail stays reversible
Detail stays reversible
The conversation creates
a natural audit trail.
a natural audit trail.
malMal assistant
Dining
Dining increased by 38% in June.
AED 2,150↑ 38% vs May
Most of the change happened on Friday evenings after 8 PM.
SALTAED 450 · 4 visits
- Jun 27 · SALT AED 180
- Jun 20 · Common Grounds AED 120
- Jun 13 · SALT AED 110
EXPLORATION
Choosing the visualization for the task.
Donut chart
Composition
Compact with few categories
Harder label placement
Ordered bar list
Comparison · Precise values
Native to vertical chat · Clear tap targets
The donut communicates composition.
The bar list better supports comparison and exploration.
A CROSS-SELL THAT EARNS ITS PLACE
Value first. Product second.
Based on real behavior
Clearly estimated
Clearly estimated
Easy to dismiss
Frequency capped
Frequency capped
malMal assistant
Your June overview is ready.
A SMALL SAVING OPPORTUNITY
You made 47 eligible card purchases in June.
Round-Up could have moved an estimated AED 24.60 into savings.
Round-Up moves spare change from eligible purchases into savings. You stay in control and can pause at any time.
TRUST AND EDGE CASES
A financial assistant must also know
when not to overstate.
UPDATING JUNE
Some recent transactions are still processing.Your totals may change when they complete.
LIMITED HISTORYNew userYou’ve used Mal for 10 days, so a monthly comparison isn’t available yet.EMPTY STATENo activityNo completed transactions were found for June.RECOVERYCouldn’t refreshTry again · View transactionsPRIVACY CONTROLAmounts hiddenAED ••••Show amountsMONTHLY RETURNTimely, not repetitive“Your June spending overview is ready.” View overview · Later
VALIDATION PLAN
Validate clarity, usefulness
and trust.
Did the answer land?
Ask users to explain what changed after a five-second glance.
Recall · Task success · Explanation confidenceDid the next step feel natural?
Observe whether category taps answer follow-ups without losing context.
Discoverability · Completion · BacktrackDid Mal stay credible?
Test estimates, privacy, dismissals and the optional Round-Up moment.
Perceived trust · Relevance · Perceived pressureClickable prototype→5 moderated sessions→Iterate→Instrumented beta
FINAL SUMMARY
One answer.
Three layers.
Two paths.
One trusted conversation.
The experience moves from understanding, to exploration, to action — without turning chat into a dashboard.
Answer firstReveal progressivelyPreserve conversationEarn the action