Co-Pilot for
Data Science
A fragmented Data Analysis tool, users with wildly different skill levels, and an AI assistant that needed to feel like a colleague — not a chatbot. This is how we redesigned Analyttica’s LEAPS into a unified system that works for everyone.
The entire platform focuses on creating
unified, persona-driven experience
that maps analytical features to
user goals
through
data storytelling
and
effortless interaction.
Recurring engagement
to be fostered via
intelligent systems
and
gamified community features
with transparent, tailored frameworks used to measure and
communicate
user growth
and
business impact.
In short, its going to be looong project.
brief
PROJECT TYPE
Retainer
PLATFORM
DURATION
10 months
Domain
Data Analysis
AI
TEAM STRUCTURE
UX Research
Product Design
SCALE
9 Modules
4 User Categories
Three
Challenges
Identified
✦
methodS employed
1
A platform serving two completely different users
Data Citizens who view dashboards and need confidence — and Data Scientists who run transformations and need control.
2
An AI that couldn't feel like a pop-up
Copilot needed to be present throughout the platform without demanding attention. The line between guidance and spoon feeding is narrow — and easy to cross.
3
multiple modules, global & isolated together be one product
From Data Analysis tool to community building had its own complexity.
After mapping pain points across user categories, we looked for the underlying tension that connected all of them:
The platform was built for data scientists, but the majority of its users were data citizens who needed a fundamentally different relationship with the tool.
01/05
Ideating
Solutions
01/05
Building Confidence
The platform's job is to make the user feel capable.
AI as Colleague
Growth has to be felt
Blocker free flows
Quick Access
Intervening by defining
nine surfaces
Constructing a
System
Before designing individual screens, we mapped every module and how they connect. This system view guided every navigation, shortcut, and cross-module flow throughout the project.
✦
methodS employed
01
Workspace
A personalized resource hub. Entry point for every user, adapting to their goals and past activity.
DEEP DIVE ↓
02
Copilot
Context-aware AI assistant. Always on, never intrusive. Multiple modes for different working styles.
DEEP DIVE ↓
03
Growth
Gamified growth tracking. Tetris-mapped tasks, consistency speedometers, maturity titles, performance buckets to back track.
DEEP DIVE ↓
04
iLeaps
Tool to explore data sets. The core workspace. Progressive disclosure for beginners; full power for experts. Streamlined flows avoiding any confusion.
DEEP DIVE ↓
05
Feed
Community-first sharing. Insights posted directly from exploration, comments linked to projects. Increasing visibility significantly.
DEEP DIVE ↓
06
Reports
Templatised report creation with smart recommendations and outlier detection to quickly share user works in presentable format
DEEP DIVE ↓
07
Learn
Adaptive learning hub. Bite-sized content linked with copilot, consistency nudges, skill-linked progress.
DEEP DIVE ↓
08
Profile
Personal growth hub. Published work, skill blocks, engagement meters.
DEEP DIVE ↓
09
Resource Hub
Centralised library for a single project asset, datasets, and curated content.
DEEP DIVE ↓
Four users,
one platform
The modules only tell half the story. Here's how real users move through LEAPS — from first entry to tracking their own growth.
Data Citizen · Viewer
Thomas, 40
Data Citizen · Creator
Praanah, 30
Data Scientist
Ravinder, 37
A familiar place to start
Thomas lands on his personalised Workspace. The dynamic hub surfaces his recent dashboards and prompts him to continue where he left off. He chooses to create new project. A large widget guides him to start with project creation, capturing metadata Copilot needs to assist him.
Workspace
→
copilot
→
learn
→
feed
→
ileaps
The decisions that
shaped
the platform
These are the three modules where the design challenge was hardest. For each, we ran multiple concept directions — and walked away from most of them deliberately. We designed two states for each screen - a zero state screen and retaining user screen across all modules, to make sure the platform caters to all. It was an approach level decision catering a system level execution.
01 Copilot — AI Assistance
Direction A
Unified Assistance Layer
Our first direction treated Copilot as a dedicated onboarding experience. The concept was a unified assistance layer — combining notifications, reminders, growth and a contextual chatbot into a single surface designed to provide deep, proactive support across the entire platform workflow
Rejected
DEEP DIVE
↓
Direction B
Proactive, Always-Visible Sidebar Panel
The second direction surfaced Copilot as a persistent sidebar — always visible, always ready. This solved discovery (users always knew where to find it), consuming significant horizontal real estate.
Rejected
DEEP DIVE
↓
Direction C — Chosen
Multi-State Copilot
Respects user autonomy. Scales across confidence levels. The drag-and-describe interaction means users don't need to articulate their problem — they can just show it.
Chosen
Deep dive ↓
Client Alignment
& Design Advocacy
Client expectation
Client pushed for surfacing the three-engine model directly, as backend was already built and a revamp was off the table. Rebuilding the architecture was not a conversation they were open to having.
Our intervention
We kept the backend intact, reframing the frontend to unify all three engines into one coherent Copilot presence. The existing content was tied to a new frontend logic — no rebuild, significantly higher perceived value.
How we aligned
We walked the client through each direction with explicit before/after comparisons, showing the three-engine model surfaced directly versus mediated through a unified presence. The difference was immediately legible and moved the conversation forward.
Final Screens
The final direction gives Copilot three modes: an always-on chat for active problem-solving, a collapsible nudge panel for passive guidance, and a contextual support mode triggered by specific actions — including drag-and-describe directly into the chat input.
This matches the mental model of a skilled colleague: present when needed, unobtrusive when not.
Why it works
Respects user autonomy. Scales across confidence levels. The drag-and-describe interaction means users don't need to articulate their problem — they can just show it.
key features
contextual states
Recommend Next Step
Always-accessible
bite size learn content
screen grab in chat
kick start flows
proactive experience
Copilot · Copilot states - Nudge, Chatbot & Contextual Tabs
Copilot · Drag and tell
02 ILEAPs - Data Visualization Tool
Direction A
Progress-Led Exploration
A structured learning environment where users explore data through goal-driven activities, with continuous visibility into time spent, roadmap position, and daily KPI progress.
Rejected
DEEP DIVE
↓
Direction B
Tool-First Workspace
A modular, tool-first workspace where users move freely between a coding environment, dataset panel, and visualisation canvas using a persistent bottom action bar.
Rejected
DEEP DIVE
↓
Direction C — Chosen
Context-Driven Tool
A dataset-first interface where users select and filter features to drive all downstream actions, with tools that adapt contextually based on what the user intends to do.
Chosen
Deep dive ↓
Client Alignment
& Design Advocacy
Client expectation
Client's existing product had multiple entry points into the same flows, creating confusion they couldn't clearly articulate themselves. They knew something was wrong but lacked the language to describe it.
Our intervention
Our designs pushed for fewer entry points, cleaner flows, one coherent path throughout the tool. This streamlines user flows to ensure end to end flow without a blocker.
How we aligned
We ran an audit and live demo, comparing current flows against our intervention to make the impact immediately visible. With this, the impact became concrete rather than conceptual.
Final Screens
A context-first workspace where tools adapt to user intent, creating a natural build-refine-share progression without prescribing a rigid flow.
This option puts the user action at the centre of every decision. Users select features using drag-and-drop, filter by data type, and are shown only the tools relevant to their current intent — create a chart, optimise it, or add to the asset library. The workspace responds to the user rather than presenting every option upfront.
The collapsible panel respects canvas space when the user is ready to focus on their output.
Dataset and feature selection as the entry point mirrors how data scientists actually think — the data comes first, the action follows.
Dual Action Copilot that can toggle to User's Growth parameters helps them quickly grasp their progress in real time.
Why it works
Contextual tools that adapt to user intent mean the interface never overwhelms — users only see what's relevant to their next step
Upfront filtering of features by type (categorical, numeric, date) reduces cognitive load by narrowing choices, increasing efficiency
Copilot's Growth and Chat tabs complement the workflow rather than leading it, giving users the option to seek guidance without being dependent on it
key features
DETAILED DATASET PANEL
categorised tools
collapsible panels
contextual menu bar
contextual copilot
asset library
insight generation & uploading
drag & drop feature listing
ILEAPS · Context driven Data Visualisation
ILEAPS · Real time progress understanding
03 Growth - Real Time Progress
Direction A
Performance KPI Dashboard
A data-dense personal dashboard that reflects user performance across six technical KPIs, designed for power users who want direct access to their metrics.
Rejected
DEEP DIVE
↓
Direction B
Metaphoric Visualisation
A narrative-driven dashboard that maps a user's platform journey onto the life cycle of a tree, turning abstract metrics into a living, visual story.
Rejected
DEEP DIVE
↓
Direction C — Chosen
Global, Integrated, Gamified Experience
A globally integrated growth system combining a badge tier framework with a visual activity map that gamifies platform engagement through earned progression.
Chosen
Deep dive ↓
Client Alignment
& Design Advocacy
Client expectation
Client wanted gamification that felt professional and simple — not a layer that would overshadow the core platform experience. Without a shared reference point, every direction risked being evaluated against a standard we couldn't see.
Our intervention
We built 10 concepts, created wireframes and flows for 4, spanning the full spectrum, using them as a landscape to help the client locate their own instincts on depth and complexity. Sprint sessions validated each direction iteratively.
How we aligned
We collaborated across calls and brainstorming sessions until the client had co-authored every decision that led to the final system. The deliverable wasn't a surprise — it was a product of shared process.
Final Screens
The left panel anchors users in their current badge level (in screen seen as explorer)— and tracks performance across four behavioral buckets: Engagement, Explore, Explain, and Experiment. Each action user performs on platform is tied to these buckets. Yes, all platform actions are mapped to tie them to each buckets.
A calendar with monthly and yearly view heatmap shows activity cadence at a glance.
The central panel surfaces relevant tasks tied to these buckets with sub actions in horizontal scroll. A Tetris-style level map that reveals new challenges as users advance, rewarding progress with unlockable content, echoing game mechanics.
With tasks listed upfront from the landing workspace, users are able to directly get into their workflow, whether they are just starting or continuing their work. This takes away the confusion on where to begin as user lands on platform.
Why it works
The badge system gives users a named identity they can invest in, which creates psychological ownership over their progress.
By tying tasks and gauges to the same four buckets, the design creates a coherent loop — users always know what to do next and why it matters.
The Tetris-style unlocking mechanic adds genuine anticipation, transforming the platform from a tool users open into one they return to.
key features
consistency metrics
tier progress system
globally integrated
quanitfied behaviour
heatmap for tracking progress
tetris game visualisation
mapped tasks and goals
workspace to growth
tetris view
GROWTH · Central Panel
ILEAPS · Left Panel - Consistency Metrics
Six more surfaces,
same thinking
04 — Workspace
Personalised hub with dynamic states
A dynamic landing page that adapts to the diverse user goals - surfacing recent work, quick learning actions, and project shortcuts. Different states for different users, without the user needing to configure anything.
Key decision
We prioritised inclusive design for any kind of user. We wanted any user to land on this page and get started with an action they want, without thinking much. They could be a data visualiser, content consumer, a learner or even someone who creates datasets. This made us build dynamic workspace to help any user kickstart their workspace.
The Growth section on top enables users to easily get back to any work they were doing or to get started. This reduces the friction users face when they are excited to start something but is bombarded with other actions.
✦
view page anatomy
05 — Feed
Community-first, clutter-free sharing
This hub enables sharing of dashboards and insights directly from dashboard. Project-based chats and platform for feedback foster a continuous, community-driven knowledge exchange for all teams, providing user with visibility within organisation and access to other colleagues.
Key decision
We prioritized instant and contextual sharing to remove technical friction, ensuring data insights are immediately searchable and actionable without workflow delays.
✦
view page anatomy
06 — Profile
Growth made visible
A personal hub that surfaces role tags, engagement meters, published work, and Tetris-style skill blocks - making growth legible at a glance for both the user and their manager.
Key decision
We chose Tetris-style skill blocks over radar charts. Radar charts show relative performance - which creates comparison anxiety. Tetris blocks show your own shape filling in - which creates satisfaction.
✦
view page anatomy
07 — projects: Resource Hub
Centralised, context-aware library
A single place for all project-related content - datasets, templates, reports, assets - organised with metadata at a glance. Access control and role-based visibility built in from the start.
Key decision
The Hub is structured around projects, not file types. Searching by project surfaces everything related - datasets, reports, and templates together - rather than forcing users to know which folder type they're looking for.
✦
view page anatomy
08 —Learn
Adaptive learning that builds habits
A dynamic hub that adapts to the user's goals - continuing courses, previewing next actions, tracking consistency. Bite-sized content is surfaced contextually across the platform.
Key decision
The Learn Consistency meter lives on the Workspace, not only inside the Learn module. Making progress visible from the entry point turned learning from an intentional destination into an ambient habit.
✦
view page anatomy
09 — reports
Templatised insight generation
Users generate structured reports from pre-built templates matched to their data type. Smart recommendations surface the right starting point; outlier detection highlights anomalies automatically.
Key decision
We built outlier detection directly into the chart view - not as a separate analytics step. Users discover anomalies while building the report, not after publishing it.
✦
view page anatomy
Impact &
Reflection
Identified
What the Redesign achieved - and what it honestly cost us to reach there.
1
Designing for data literacy is not the same as designing for data tools
The most technically capable interfaces we explored failed the moment a data citizen sat down with them — not because they were badly built, but because they assumed a baseline of confidence that most users don't have yet.
2
A strong IA makes every surface easier — a weak one makes every surface harder
When the IA is right, adapting to a new surface is a layout problem. When it's wrong, every surface requires a redesign from scratch. Getting the system architecture right in the first month shaped the quality of work for the remaining nine.
3
The best AI features do less, better
The temptation with AI-assisted features is to make them do more. The better instinct, confirmed through this project, is restraint. An AI that speaks when spoken to and nudges without interrupting is more trusted than one that announces itself at every opportunity.
4
Gamification in professional contexts requires a different model
Points and leaderboards work when users want to compete. In a data literacy platform, most users want to grow without being ranked. The final system rewards consistency and progress over performance.