AKAkhil Kumar
← All work

01 — Overview · AI product design · Conversational commerce

YARA AI — Intelligent Retail & Shopping Assistant

An AI shopping assistant that helps customers of a large UAE home & furniture retailer find, compare and choose products by text or voice, across retail stores, kiosks, web and mobile.

Company
Home & Furniture Retail, UAE
Role
Lead Product Designer

02 — Problem

The retailer sells a large furniture and home-improvement range across its website, app and showrooms. Finding a product meant knowing its category and then narrowing with filters. That's hard when a customer starts from a need, like “minimalist living-room furniture in white and natural wood under AED 5,000”, rather than a product name.

The goal was an assistant that understands that intent, keeps the customer moving toward a decision, and behaves the same way in showrooms, on kiosks, on the web and in the app.

03 — Users

  • Online shoppers

    Start from a need or a style, not a product name, and want relevant options fast.

  • Showroom & kiosk customers

    Want product details, availability and price on the spot, from the item in front of them.

  • Customers furnishing a room

    Need to see how pieces will look together before committing to a purchase.

  • Comparison shoppers

    Weighing several similar products and want the differences laid out clearly.

04 — My role

Lead Product Designer

  • AI UX strategy
  • Conversational design
  • Customer journey
  • Information architecture
  • Wireframes
  • Prototyping
  • UI
  • Design system
  • Front-end collaboration

Team · Worked with product managers, AI engineers, developers and business stakeholders. My part was the experience: strategy, flows, interaction design, UI and the design system.

What I owned

  • Led the UI/UX design and experience strategy for the assistant
  • Owned UX research, user flows, wireframes, high-fidelity UI and interactive prototypes
  • Designed conversational AI interactions for text and voice
  • Designed AI-powered discovery, recommendations and product comparison
  • Designed the QR-enabled showroom experience linking customers to product details
  • Designed “Imagine with YARA” for visualising products in realistic interior spaces
  • Built the scalable omnichannel design system across web, mobile and kiosks

05 — Challenge

  • P1

    Filters assume catalogue knowledge

    Category trees and filter panels work for customers who already know what a product is called. People shopping by need had to translate it into the catalogue's vocabulary first.

  • P2

    Comparing meant opening page after page

    Weighing up options required visiting each product page in turn and holding the details in memory.

  • P3

    Out-of-stock was a dead end

    A strong match that was unavailable stopped the journey instead of steering it toward a close alternative.

  • P4

    Store and online were separate journeys

    Help in a showroom and help on the website didn't share context, so the experience reset at each touchpoint.

  • P5

    Hard to picture furniture at home

    Product photos show an item on its own, not in the customer's style or room.

06 — Journey / workflow

From customer intent to purchase

  1. 01Customer intentThe customer types or says what they need, in their own words.
  2. 02AI understandingThe assistant picks out category, style, colour, material and budget.
  3. 03Search & recommendationsMatching products appear as a result set rather than a filter panel.
  4. 04Product comparisonSide by side on specifications, pricing, ratings, reviews, features and availability.
  5. 05Product discoveryRich content (videos, A+ Content, specs, reviews) with live availability and pricing.
  6. 06PurchaseAdd to cart, or order in bulk, without leaving the result set.

07 — UX approach

Customer Journey → Information Architecture → User Flow → Wireframes → Component Strategy → Design System → UI → Prototype → Development

  1. Customer JourneyMapped intent → purchase across web, app and store, including the store-to-digital handover.
  2. Information ArchitectureSeparated two modes up front: talking to YARA (help and search) and imagining with YARA (visualisation).
  3. User FlowConversational flows for text and voice, including refinement, comparison and recovery when nothing matches.
  4. WireframesResult sets, the comparison tray and one-tap actions explored at low fidelity before visual design.
  5. Component StrategyDefined conversational components that sit alongside the existing commerce components rather than replacing them.
  6. Design SystemAdded the conversational components to the shared design system so web and app stay consistent.
  7. UIA distinct assistant identity, gradient accent for AI moments, and standard product cards for results.
  8. PrototypeInteractive prototypes for text and voice interactions.
  9. DevelopmentWorked with AI engineers and developers to ship across web, mobile and kiosks.

08 — AI interactions

  • Conversational UI

    UX decision · A named assistant with a clear identity, so customers know when they're talking to AI rather than browsing.

  • Text interaction

    UX decision · Natural-language queries replace filter-building: the customer describes intent once.

  • Voice interaction

    UX decision · “Talk to Yara” is a primary entry point, not a hidden mic icon, for customers who'd rather speak.

  • AI search

    UX decision · Intelligent search returns a curated result set with a plain-language title of what was understood.

  • Recommendations

    UX decision · Personalised suggestions based on intent, preferences and shopping behaviour.

  • Product comparison

    UX decision · Select products to compare specifications, pricing, ratings, reviews, features and availability side by side.

  • QR / showroom

    UX decision · QR codes in the showroom link customers straight to product details on their own phone.

  • Rich product content

    UX decision · Videos, A+ Content, specifications and reviews answer pre-purchase questions without a sales associate.

  • Live availability & pricing

    UX decision · Real-time stock and price shown with every result, so a recommendation is always something you can buy.

  • Imagine with YARA

    UX decision · A separate visual mode for seeing products in realistic interior spaces before purchase.

09 — Solution

Entry point — two modes, chosen up front
  1. 1Two intents, two buttonsGetting help (talk) and visualising (imagine) are different jobs, so they're separate entry points rather than one chat box that has to guess.
  2. 2Voice is visibleThe waveform icon signals voice before the customer taps, so speaking feels like a supported path, not an accessibility afterthought.
  3. 3A face for the assistantA named, illustrated character makes it clear the customer is talking to an assistant, and gives the AI a consistent identity across channels.
AI product search — describe intent instead of building filters
  1. 1Intent in, results outOne natural-language query replaces several filter steps: style, colour, material and budget are read from the sentence.
  2. 2Dynamic multi-select traySelecting products enables compare, add to cart or order in bulk, so the result set becomes a decision space.
  3. 3One-click answersDelivery timeline, return policy and specs are one action away, answering the usual pre-purchase questions in place.
  4. 4Smart refinementWhen items are out of stock, the assistant offers refinement variations instead of an empty state.

10 — Design system

One omnichannel design system across web, mobile and kiosks. Conversational patterns sit alongside the commerce components, so AI results look and behave like the rest of the store.

  • Entry points

    Primary pill buttons for each assistant mode, with icon cues for voice and generation.

  • Query input

    Large natural-language field with a gradient focus state reserved for AI moments.

  • Result set

    Titled result group that restates what the assistant understood.

  • Product card

    Standard commerce card (image, name, price, was-price, add to cart) plus a select state.

  • Multi-select tray

    Contextual actions that appear once one or more products are selected.

  • States

    Out-of-stock and no-match states that suggest refinements instead of ending the flow.

11 — Design → code

  1. 01FigmaFlows, conversational components and prototypes.
  2. 02Design systemConversational components added alongside commerce components.
  3. 03Web · mobile · kioskBuilt with developers and AI engineers for each touchpoint.
  4. 04ProductionLive as YARA AI search on the retailer's website.

12 — Outcome

  • Live on the retailer's website as YARA AI search.
  • Simplified product discovery and comparison.
  • Bridged showroom and online shopping with QR-enabled journeys.
  • More engagement and purchase confidence through visualisation and personalised recommendations.
  • A scalable design system supporting multiple retail touchpoints.