Reimagining how product teams discover product problems with the use of AI.
Embedded Files
Rather than asking Product teams to manually search through thousands of behavioural events, I designed an AI-native product intelligence system that automatically detects signals, explains what changed, and helps teams investigate customer workflows through AI.
Staff Product Designer • AI-native Workflow • Behaviour Analytics • Product Intelligence • Internal Platform
Overview
Products are how organisations build. Workflows are how customers get work done.
Over the past three and half years, Intropic, a fintech data startup, has grown from a single spreadsheet-based application into a suite of enterprise financial products.
Every day, thousands of user sessions, performance events, errors and behavioural signals are generated across multiple products. Although the data existed, understanding what actually happened still required someone to manually review session recordings, navigate multiple dashboards and piece together evidence from different systems.
The problem wasn’t a lack of data.
The problem was turning behavioural data into product decisions. Rather than asking teams to search for problems, I wanted AI to continuously surface the most important product signals, allowing teams to spend less time analysing data and more time improving the product.
The Problem
Searching doesn’t scale.
Searching doesn’t scale.
Every morning followed the same pattern.
Open session recording.
Review recordings.
Compare analytics.
Investigate performance regressions.
Try to understand what user were doing.
What went wrong ?
This process worked while the product was small.
As more applications were introduced, it became increasingly difficult to connect customer behaviour across products.
Traditional analytics answered questions like:
-Which page was visited?
-Which feature was used?
-How many users logged in?
But they couldn’t answer the question that mattered most.
What was the customer actually trying to accomplish?
Discovery
Customers don’t just use products. They complete workflows.
While analysing raw behavioural data, I noticed something unexpected. Customers rarely stayed inside a single application. Instead, they naturally moved across multiple products while completing investment research.
Understanding individual product metrics no longer reflected the actual customer experience. The most valuable insights weren’t inside a product. They existed between products.
This became the turning point. Rather than measuring individual products, I started mapping how customers naturally moved across the ecosystem.
Instead of seeing six separate applications, behavioural data revealed a much bigger picture.
Customers weren’t using products.
They were completing investment workflows. The product was a tool in their workflow.
For example:
SHI → Portfolio Forecast
Quantitative Analytics → Portfolio Forecast → ECM
ECM → Insights
EQLS → Portfolio Forecast
This completely changed how I thought about product analytics. The most valuable insight wasn’t inside a product. It existed between products.
My Role
0→1 Product Builder
I independently conceived, designed, built and shipped the entire platform from end to end.
Unlike a typical design project where responsibilities are shared across Product, Design and Engineering, this project was entirely self-driven.
I identified the opportunity through behavioural analysis, defined the product strategy, designed the user experience, built the AI workflow, and implemented the working prototype that the team used daily.
Designing the system
From Behaviour Data to Product Decisions
Rather than building another dashboard, I designed a layered AI-native product intelligence system.
Each layer solves a different problem.
Instead of overwhelming teams with data, the platform progressively transforms information into product decisions.
Behaviour Layer
Collecting behavioural evidence
The platform continuously collects behavioural events from multiple sources vai MCP.
- Recording Sessions
-Errors
-Performance Metrics
-User Journeys
-Cross-app Navigation
This layer captures what happened.
Signal Layer
Slack Summary
Every morning, the system automatically generates a product health summary and posts it into Slack.
Instead of reviewing dozens of recordings, teams immediately understand:
Product usage
Active users
Performance regressions
Network failures
Cross-product journeys
Engineering priorities
The summary provide the signal to the team, became the team’s starting point each morning.
Slack Summary
Reasoning Layer
AI Product Analyst
One interesting behaviour emerged after introducing the daily summaries.
People rarely stopped after reading the report. Instead, they immediately asked follow-up questions.
- Why did usage drop yesterday?
Which users were affected?
Has this happened before?
Which workflow experienced the most friction?
Show me the session replay.
This observation led me to design and build an Chat Product Assistant directly into the platform.
Instead of navigating dashboards, anyone can ask natural language questions and receive contextual answers grounded in behavioural data.
Examples include:
- Why did ECM usage drop yesterday?
- Which JavaScript exception affected the most users?
Show users who moved from QA to ECM but never reached Portfolio Forecast.
Compare today’s workflow with last Monday.
The goal isn’t replacing dashboards. It’s making product intelligence conversational.
Ask questions like:
“What do users mostly do in this app?”
The system automatically identifies the most common user workflows and analyses how users navigate through them.
Evidence Layer
Dashboard
AI provides context.
Evidence provides confidence.
Every AI explanation links directly into the supporting behavioural evidence.
Each insight links directly into the dashboard where teams can explore:
Product health
Performance
Session Replay
User Behaviour
Feature Usage
Error Analysis
Cross-product Navigation
Rather than presenting raw analytics, the dashboard highlights meaningful behavioural signals that help Product and Engineering prioritise work.
a internal dashboard as evidence layer
Decision Layer
from insight to action
The final outcome isn’t another dashboard.
It’s better product decisions.
Instead of spending hours discovering problems, Product, Design and Engineering can immediately discuss:
What changed?
Why did it happen?
Who was affected?
What should we prioritise?
The conversation shifts from analysing data to deciding what to do next.
Impact
Changing how product teams discover problems
The biggest success wasn’t the dashboard.
It wasn’t even the AI assistant.
It was changing how product discussions started.
Before building the platform, investigating product issues typically meant manually reviewing sessions recordings, switching between multiple dashboards, and piecing together evidence from different systems.
After introducing the AI Product Intelligence workflow, the process became fundamentally different.
Instead of starting with raw behavioural data, teams now begin with AI-generated signals, investigate only when necessary, and focus discussions around customer workflows rather than isolated product metrics.
From Reactive Investigation to Proactive Intelligence
The platform introduced a new way of working across Product, Design and Engineering.
Instead of asking:
Which dashboard should I open?
Which recordings should I review?
Which product has the problem?
Teams naturally began asking:
Why did this happen?
Which customers were affected?
Which workflow broke?
What should we fix first?
The conversation shifted from searching for data to making decisions.
Reflection
This project fundamentally changed how I think about AI-native products.
The goal isn’t replacing dashboards with chat.
It’s designing the right layer of intelligence at the right moment.
Behavioural data tells us what happened.
AI explains why.
Humans decide what to do next.
Designing AI products isn’t about adding another interface.
It’s about transforming complex information into actionable signals that help people make faster, better decisions.
Products are how organisations build software. Workflows are how customers get work done. AI helps bridge the gap between the two.
Other interesting project
From a 200-row spreadsheet to a SaaS platform driving $10M ARR
Transforming a spreadsheet workflow into a scalable analytics platform
Staff Role • Fintech • UX & Data Visualization • Cross-Platform • User Research • B2B
Years: 2022-2025
Read the case study->
Report abuse