Case studyCAI · Bajaj Finserv2024–26Product designer, DCX · led the early direction

An AI-first platform

CAI, Bajaj's AI-first platform, puts a financial butler in every customer's app. I helped leadership commit to it, and argued for the moments when tapping beats talking.

Ledthe idea and key parts of the project; laid the groundwork
2 shippedBlu, Bajaj's chatbot, upgraded; AI on the live website
400+ peoplebrought in over 64 days, 47 meetings, 4 workshops
18 principleswritten into 67 guidelines
POC v1 · Dec 2025An early proof of concept: the homepage in dark mode, with an insurance renewal card, quick actions and an AI bar.
Delivered library · Nov 2026Phone discovery in the delivered library: the customer asked for phones under ₹25,000 with a good camera, and drags one phone card onto another to compare them.
POC v5 · Mar 2026A later proof of concept: the command bar inside a partner store, offering to repeat the last order.
01

The idea

Late 2024 → Jun 2025

A financial butler, not a chatbot.

In workshops with leadership we settled the pitch: anyone who opens the app gets a butler for their money. It remembers, reminds and explains across loans, cards, insurance and savings. That makes it a platform: one AI layer over every Bajaj product.

It went into the long-range plan as "Pioneering AI" and became a project of its own.

From today to AI-first

  1. Rigid, templated page structuresFluid yet structured
  2. Fixed ecosystem mapsPatterns-based output
  3. For the next generationReal, now and for everyone
  4. Built from the company outBuilt from the consumer in
  5. Refresh every 4 yearsAlways in beta
The five shifts we asked leadership to make.
02

The room

Sep → Oct 2025

Leadership had to learn to think like designers.

A lending company plans from the business case. An AI-first platform can't be planned that way, so the first thing I designed was a way of deciding: DTPI, a design-thinking framework that later won Bajaj's HEROES award. We brought in IDEO; I sat on the panel that chose them. Their workshops with leadership turned the pitch into four opportunity areas.

  1. 01ContextWhere we stand
  2. 02SignalsKey themes emerge
  3. 03OpportunitiesThemes become opportunities
  4. 04PropositionFeatures and functions
  5. 05PrototypesCustomer-tested concepts
  6. 06Go / no-goA brief to build from
03

The tension

Throughout

Not everything should be a conversation.

The hype pulled every step into chat. But typing or saying an answer the customer could tap is slower, and less certain. I kept arguing for direct manipulation, controls you touch inside the conversation, and wrote it into the system's rules.

  1. Talk when it's faster to say."Phones under ₹25,000 with a good camera" beats six filters.
  2. Tap when there's a choice.One option is a question. Short answers are chips to tap. When a search returns two or three products, each gets a card with its picture and monthly EMI.
  3. See it before you commit.Money gets live controls: the EMI changes as you pick a plan or move the down payment.
  4. The customer owns the controls.A tap is the customer's answer. Blu suggests, and never moves a control for them.
Choosing a no-cost EMI plan inside the chat: plans as chips, a down-payment slider and the monthly amount updating live, with Blu's suggestion above.
Tapped, not typed. Blu explains the plans; the customer picks.
04

The proofs

Oct 2025 → Mar 2026

Every meeting ended with something leadership could touch.

We brought screens, motion, a coded prototype and videos of the principles. Research with 23 customers and 7 partners across 4 cities gave us the line we built on: guide the customer, and let them decide.

400+

people brought into the work in 64 days, across 47 meetings, 4 workshops and 5 countries.

Version 5, the chat-led home screen79 / 90 · 88%
  • Layered depth8
  • Glow9
  • Soft shadows8
  • Depth shows priority8
  • One thing in focus9
  • Visual feedback for touch8
  • Glow on what's tappable9
  • Transition effects10
  • How much fits on screen10
Five versions, one rubric. Each scored out of 10 on nine visual qualities we defined, from depth and glow to how much fits on screen.

"If it pays automatically, I'll be scared."

A customer in the concept research. It changed what the butler was allowed to do.
05

What shipped

2025 → now

Two pieces went live before the platform existed.

Blu, Bajaj's chatbot, got its upgrade: a loan journey that runs inside the chat, from eligibility to the offer. It's live on every product.

The live website got AI answers. 97% of testers found the summary helpful, but only 40% noticed it. So it moved to the top, and it writes itself in word by word, which marks it as AI.

BeforeBelow the fold
Before: the personal loan page with the AI answer in a 'FinAI' block below the fold.
AfterOn top, written live
After: an AI Overview card at the top of the page with a Listen chip and follow-up questions.
06

The library

2026 → now

Tapping and talking now live in one library.

The project began as Terranova and was renamed CAI. The CAI library, 204 components and 902 variants, follows the rules I set for the everyday library behind the live website, and gives every chat journey both kinds of part: things to say and things to tap. The library team delivered it in November 2026.

CAI library page for the Table component: references, in-use variants and notes on the left; five chat screens using it on the right.
One component, its rules, and the chat screens it builds.

When ideas piled up, I gave the room a scoring rule.

The Critical Checklist, built in Figma Make, scores every idea against our principles. One critical "no" and it's out.

Live · try it
Pick an idea

CAI principles91%

Critical7 checks5 yes · 0 no
Protect Privacy by Default

Does the idea use only necessary data and make that visible to users at each step?

Cite with Evidence

Can every claim or recommendation be traced to a source without overpromising?

Freedom from Syntax Accuracy

Can users interact naturally, even with imperfect language, without learning commands or formats?

Handle Mistakes Gracefully

If the AI or system fails, does the idea provide a useful fallback and clear next steps?

Fair and Safe for Everyone

Could this affect user groups differently, and is there a way to detect and fix that?

Follow Regulatory Boundaries

In regulated flows, does the idea apply consistent rules, keep an audit trail, and handle consent clearly?

Easy and Actionable

Does it always show what happened, what is next, and how to continue without dead ends?

High6 checks4 yes · 1 no
Multimodality

Does it support multiple input/output modes and choose the right one for the user's context?

Let People Set Autonomy

Can users choose the automation level and undo or review automated actions?

Explain with Proof

Does it explain why it made a suggestion with reasoning users can verify?

Manages Memory and Data

Can users view, edit, and clear memory, with session vs persistent memory clearly explained?

Recall What Matters

Does it use prior context to keep the experience continuous without making users repeat themselves?

Goal Driven

Does it guide users to goals through intent-aware flows instead of fixed menus?

Medium5 checks2 yes · 0 no
Respond with Emotional Tone

Can users influence response tone, and does it stay consistent across contexts?

Personalised

Does it adapt to user context and behavior in a visible, non-intrusive way?

Generative

Does it generate tailored outputs that users can refine?

Ambient

Does it stay in the background and let users take over without losing context?

Orchestrate Across Journeys

Does it coordinate multiple tools into one coherent flow without exposing unnecessary complexity?

Usability checks89%

Critical5 checks4 yes · 1 no
Clear, Simple, and First-Time Usable

Can a first-time user with no specialized knowledge complete the core task without support?

Show Consequences Before Commitment

Before submission, can users see the future impact of this decision?

Preserve User Agency While Using AI

Does AI reduce effort without taking away user control?

Build Trust Through Safe Recovery Paths

If something goes wrong, can users recover quickly without losing progress or context?

Compliance and Consent Are In-Flow

Are compliance, consent, and disclosures handled clearly in the flow?

High4 checks4 yes · 0 no
Accessible to Target Demographics

Is this usable for the target demographic across literacy, language comfort, ability, and device limits?

Solve a Real Need with Clear Value

Does this feature address a real and immediate user problem, and is the value clear within seconds?

Explain Recommendations with Evidence

Can users understand and verify why a recommendation was made?

Keep Cross-Product Context Continuous

Does it carry context across products to reduce financial fragmentation?

Medium1 check1 yes · 0 no
Appears Naturally in Existing Workflow

Is this encountered naturally in existing financial workflows?

07

Also in this work

DTPI

A design-thinking framework for all 65,000 people at Bajaj Finserv. It won a HEROES award.

Award · Oct 2025
The 18 principles

Each tied to the feeling it should create, from "I trust this advice" to "I am the boss".

18 principles · 67 guidelines
The coded prototype

A design agency's app for the project, rebuilt as a working prototype: the AI answers with screens, not text, choosing from 87 component types.

Expo + React Native
Blu research

171 chatbot users: 89% had felt the chat was scripted, 41% said their need didn't fit any preset button.

Survey
FigmaFigma MakeExpo + React NativeClaude CodeGemini