Onepint - OneTruth, Pint Planning & Control Center

Product: OneTruth (AI-powered Unified Inventory Management Tool)
Role: UX Designer (part of cross-functional team with PMs, engineers, and data scientists)
Timeline: Q4 2024 - Q3 2025
Platform: Enterprise Web Application
Team: UX Designer × Director × Tech Lead × Data Science × Product Manager

Introduction

OneTruth - Trusted inventory, every time.

Although we have our robust OMS platform efficiently managing orders and inventory, we noticed planners still relied on manual analysis and spreadsheets for critical decisions. This revealed a gap between operational visibility and intelligent decision-making.

My Role & Approach

As the UX Designer, I conducted a heuristic evaluation to surface key issues in navigation, visual hierarchy, and task flows.

Key Design Improvement

• Experience Modernization

• Usability & Flow Optimization

• Cognitive Load Reduction

• Foundation-First Redesign

Skill demonstrated

• Legacy Product Redesign & UX Modernization

• Workflow Simplification & Efficiency Optimization

• Visual System & Interaction Design

• Conversational UX and voice UX

Outcome

Streamlined workflows and a modernized interface improved task speed and usability while preserving existing user mental models.

Challenge - What was broken

• Redesigning the robust OMS platform into new intuitive modern platform

• Designing for trust — users needed to understand how AI calculated inventory and promises.

• Presenting large, complex data in a clean, scannable interface.

• Maintaining continuity with existing OMS workflows while introducing new intelligence features.

Who the User is ?

Why it mattered to the business

• Forecast Accuracy: Inaccurate guessing that leads to either too much dusty inventory or empty shelves.

• Fulfillment Costs: Expensive labor, high rush-shipping fees to fix errors, and high rent for storing items that aren't selling.

• In-Stock Levels: Constant "Out of Stock" messages and backorders that frustrate and drive away customers.

• Sales Growth: Stagnant revenue caused by missing products and being unable to compete with better-priced competitors.

Walkthroughs of Critical Flows

Search Filtering - Existing Version

Feedbacks

• Cognitive overload for the customer

• The user can't remove the item ID from the main screen.

• Multiple clicks to perform a single task From this screen the user can delete them or add them.

• No data upfront

Search Filtering - Improved Version

Improvements

• Search filters are surfaced upfront, making discovery faster and reducing initial friction.

• Bulk deselection of Item IDs is enabled, eliminating the need to unselect items one by one.

• Fewer interactions required to run a search, reducing overall click count and task time.

• Backend efficiency improved by fetching search results only after the user applies filters, minimizing unnecessary API calls.

• Personalization options added so users can structure and view data in a way that fits their workflow.

• Relevant data is shown upfront instead of an empty “No data found” state, helping users understand the type and structure of available data.

Inventory Details Screen

Existing Screens

Feedbacks

• Layout is not consistent and scattering of data

• User has to do infinite scroll to read all the information that is shown

Improved Version

• Layout has been changed with data segregated

• Following a 80% 20% split

• Reduced infinite scrolling

• Data has been segregated

• Highlighted only the primary data upfront

• Secondary date will show up only on interaction

Hard design Problems!

No Stable data patterns

AI adoption improved when the interface respected existing mental models. Structuring outputs to mirror known data fields reduced cognitive load, increased predictability, and built trust

Probabilistic Output

Free-form AI recommendations raised trust questions. Progressive disclosure sources and confidence on demand, maintained speed while restoring credibility.

Multiple Conflicting Truth

When priorities conflicted, I used design to surface trade-offs and anchor decisions in user impact using options, prototypes, and scenarios to replace opinion with evidence.

User expected enterprise grade control

Because the product supported high-impact enterprise decisions, recommendations alone weren’t sufficient. Users needed visibility and the ability to review, adjust, or override system outputs

Yet needed fast, intuitive decisions

Although the workflow demanded enterprise-grade transparency and control, users operated under time pressure.

Hard design Problems

1. User-Initiated → AI Response (Controlled Input Model)

2. AI-Suggested, Human-Decided (Decision Support)

3. Predictive UX (Anticipatory Assistance)

4. Explanation & Transparency Layer

5. Learning by Observation (Adaptive Feedback Loop)

Information Architecture & Content Strategy

How Information Was Grouped

Information was grouped around user intent and decision stages, rather than around system outputs or data sources. The interface was structured into three logical layers:

Decision Layer

Primary actions, AI recommendations and outcomes

Context Layer

Alt suggestions, Supporting insights

Explanation Layer

Data sources, confidence indicators, and historical signals

What Was Elevated

• Primary AI recommendations and next best actions


• Critical signals that directly affected decisions


• High-confidence predictions and learned patterns

What Was Suppressed (but still accessible)

• Secondary metrics and edge-case data


• Model-related complexity (scores, probabilities, internal logic)


• Historical noise that didn’t change the immediate decision

What Was Intentionally Hidden

• Raw model probabilities and technical explanations

• Advanced configuration and governance controls


• System-level diagnostics and logs

Information Architecture & Content Strategy

Why Alerts Were Contextual, Not Global

Global alerts were avoided because they lacked situational relevance and frequently interrupted workflows.
Instead, alerts were designed to appear only at the moment and place where user action was required.

Voice Interaction Design Decisions

What decision the user is making

In moments where users already know what they want to do, voice allows them to express intent faster than navigating menus or typing structured commands.

What friction was intentionally added or removed

• Removed the need for manual input by allowing natural voice commands

• Avoided forcing users to structure commands precisely


• Required explicit confirmation for high-impact actions

Chat Interaction Design Decisions

• The need for user to scroll into infinity loops to view the full information that is provided by the pinto


• Reduced the cognitive load for the user


• The interaction reveals detail on demand, avoiding infinite scrolling and preserving context

Multi Tenant Interaction Design Decisions

• The user is switching the tenant in order to view multiple organisation so checking the inventory and the associated stocks


• Eliminated the need to log out and re-authenticate when switching tenants

• Avoided multi-step flows for frequently accessed tenants

Chatbot Library

• The user can save the commands and reuse them from the saved list

• The pinto also suggest from the pre saved prompts so that the user can select from them

Visual Hierarchy & Data Visualization

Visual Hierarchy & Data Visualization

How Noise Is Suppressed



Noise was actively reduced by filtering out signals that didn’t change user decisions:
- Secondary metrics were visually de-emphasized
- Repetitive alerts were consolidated or throttled
- Historical or low-confidence signals were hidden by default

The system optimized for signal density, not data density.

Earlier Design Iterations

Final Design

What decision is enabled here?

Only to show the the data that is relevant to the user in the first half/fold and then to show the secondary data in the second fold/details of the screen. To avoid cognitive load/showing more details at a time

What decision is enabled here?

To enable the user to take quick action by showing all the filter information upfront without any secondary screens and also letting the user to cancelling the filter if something is not required for them to consider

What decision is enabled here?

To declutter all additional information under a single area for ease of user and also by grouping the data for additional spacing

What decision is enabled here?

To provide the user a detailed/clarity about the system that is being shown upfront and also a personalisation for the user

Design System

What This Reveals About My Design Ability

• Design for decision-making rather than visual aesthetics

• Structure complex information to reduce ambiguity and cognitive load

• Use visual hierarchy to guide attention and user judgment

• Simplify workflows by reducing steps and unnecessary interactions

• Suppress noise by de-emphasizing non–decision-critical data

• Select and apply appropriate mental models for AI-driven interactions

Impact :

These results are the real time results obtained from various clients

Future Enhancements

• As a future enhancement, instead of relying on Pinto for full-page queries, we can introduce data-driven contextual bots that appear when users select specific data points, enabling more focused and relevant interactions.

• List vs cards in the design decision would require a testing so that we can understand about the user workflow

Things I learned

• Architected and launched Al-enhanced Inventory Planning System from 0 -> 1 and 1 -> 10. Took initiative as a lead designer to make significant changes to the UX and owning the design.

• Initiated the GTM activites and created product demo videos
  
  

That's all folks!!!