Pint Control Center: Gen AI-Powered Inventory Optimization

Product: Pint Control Center: AI-Powered Inventory Optimization
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

Pint Control Center: AI-Powered Inventory Optimization

The goal was to transform complex supply-chain data into an intuitive, proactive, and actionable decision-making platform for planners, operations managers, and business leaders.

This case study captures how I approached the problem from understanding user needs to shaping a system that predicts risks, recommends transfers, and unlocks real-time visibility across warehouses.

Context

Modern retail, D2C, and multi-warehouse businesses struggle with fragmented systems like ERPs, WMS tools, spreadsheets, and siloed teams.
This results in delayed stock movements, over-stock at one location and stock-out at another, manual firefighting, no predictive visibility, and no single source of truth.

Problem

From user research, three key pain points emerged.

• Prevent Stock Disruptions -
  Predict stockouts and overstock before they impact your bottom line.

• Plan With Confidence and Minimize Financial Risk -
  Assess the financial impact of inventory issues in real time and test different strategies with scenario planning before making key inventory decisions.

• Optimize Stock Distribution -
  Ensure the right stock is at the right place at the right place with Al-suggested

  transfers and replenishment plans.

• Boost Productivity and Empower Teams -
  Automate routine inventory decisions, freeing up time for strategic planning and Improve decision-making with a centralized platform that connects supply chain,

  finance, and operations teams.

What does OneTruth Offer

• What-if Simulations
  Predict stockouts and excess inventory before they impact your bottom line.Assess financial risks in real-time and take corrective action before disruptions occur.

• Stockout/Overstock Predictions
  Monitor every event affecting supply, demand, and available-to-promise (ATP) for a complete, up-to-the-minute view. Optimize stock distribution with AI-driven recommendations, ensuring products are available where and when they’re needed.

  
  

• Risk Assessment
  Save time by automating routine inventory decisions, reducing manual intervention.Ensure critical issues get the attention they need, with automatic escalation to planners.

  
  

• Autonomous decisions

  Test different scenarios, such as supplier delays or demand surges, to understand their impact.

  Make informed decisions by visualizing potential risks and identifying the best course of action.
  
  

• Seamless Collaboration & Insights
  Empower teams with real-time, data-driven insights, keeping planners, analysts, and financial leaders aligned.Improve decision-making with a centralized platform that connects supply chain, finance, and operations teams.

Challenge

• Complex Data Structures

  Inventory data spans SKUs, batches, locations, supplier SLAs, and multi-location demand. Simplifying this complexity without losing meaning was a core challenge.

• Multiple User Roles

  Daily planners, monthly supply managers, and quarterly business leaders all needed the system.
  Designing an interface that adapts to each mindset required careful prioritization.

• Translating AI into Human Actions

  AI predictions alone are not useful.The challenge was converting them into understandable, trustworthy, and actionable insights.

• Avoiding Cognitive Overload

  Users needed clarity, not clutter.
  Balancing information depth while avoiding overwhelming screens required multiple iterations.

Design Decisions

1. Incorporate Pinto (AI agent) into the workflow

2. Important information appears upfront. Advanced details open only when the user chooses to dive deeper.

3. MakingRisks and alerts became the starting point because users primarily come to identify and mitigate problems.

4. Introducing Action-Oriented Recommendations which was paired with a clear action like “Move 120 units from Warehouse A to Warehouse B.”

5. Multi-Level Drill-Downs

   Users can move from an overview to SKU-level risk, then to warehouse analytics, and finally to an action workflow.

6. Consistent System Language

   Card structures, colors, and iconography were standardised to create familiarity across all modules.

7. AI Transparency

   A “Why am I seeing this?” explanation was added to help users trust and understand AI logic.

Consumer Personas

Walkthroughs of Critical Flows

Pinto (Conversational AI bot)

• It is a conversational or assistant-type tool integrated into OnePint.ai’s platform. An LLM‐based virtual assistant … bridging the gap between data complexity and operational decision-making with AI supply chain solutions.

• It simplifies user interaction with the analytics/data: instead of requiring users to dig through dashboards or spreadsheets, the assistant can interpret queries, surface insights, guide decision-making.

Things it can do

• User can ask natural questions like which warehouses are at risk of stock-out or What transfers should we initiate

• Suggests actions such as internal transfers or replenishments using AI-driven analytics.

UX design decision:

• Curated some user actions such as copy, like and dislike to track the LLM model and its behaviour towards answering the questions/during conversation

• While being deployed in other screen where there is heavy data the pinto screen which is a 100% ratio can now be minimized to a 30 % ratio and the rest 70% contains the rest of the screen with heavy data about the inventory.

• From the data heavy screen/native screen the user can chat with pinto and pinto responses with solutions for the same.

• User can save the conversation and can later continue them. They can also start a new conversation without being lost

• Incorporated (Prompt Library), Audio conversation (mic) for the pinto interaction which is played to be developed (Planned),

Filtering Options

Before

Feedbacks

• Previous in this filter the ueser needs to click on many action to perform a search/filter function.

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

• One has to go the advance filter -> click on the green color bubblle with number -> Screen with multiple numbers would be displayed.

• From this screen the user can delete them or add them.

• This includes too many clicks and also it was confusing for the user while handling this large number of data

After

Efficiency - Filter options

• EfficiencyIn this the search/filter option can be just completed in two clicks. Thus by reducing the Number of clicks. This way the user can act proactively and thus by increasing the efficiency

• I've taken a decision that upon clicking on the search button it will hit the backend and then only the result will be displayed thus by reducing the number of hits

• Introduced a horizontal scroll bar with some fields as fixed thus by prodiving the user a better visibility over large number of datas

• The number bubble which shows the number of filter that are currently applied will be shown upfront thus by providing the user a better visibility to take better decisions

Personalisation

• Introduced a column filter enabling the user to rearrange the columns for their preferences

• Ability to change the number fields in the pagination so that it would give a better view of the data

Create Simulation

Lemme drive you through a example
There are 4 most common scenarios

• Demand Surge Simulation
  Demand pace unexpected demand increase / decrease across channel

• Supply Disruption
  Simulate supplier cancellations or delays

• PO Disruption

  Simulate purchase order cancellations or delays

• Channel Shift

  Demand moving between online/offline channels

Scenario
“What if the supplier shipment for SKU A101 gets delayed by 7 days?”

System Output

• Stock-out predicted in Warehouse B on Day 5

• Safety stock breach in Store C

• Potential impact: 140 missed orders

• Recommended actions:

  • Move 60 units from Warehouse A to Warehouse B

  • Increase replenishment frequency for next 2 weeks

UX Outcome
The user can instantly see how a supplier delay cascades across locations and act before disruptions occur.

What Pint Control Center (Pinto) Does

Pinto acts as the central intelligence layer for inventory and supply chain operations.
Instead of teams manually checking reports, switching between systems, or reacting late, Pinto continuously analyzes real-time data and proactively tells users what is going wrong and what to do next.

• Creates a Single Source of Truth
  Pinto connects data from warehouses, stores, suppliers, and sales channels to give one unified view of inventory.
  No more spreadsheets or cross-checking multiple dashboards.

• Predicts Future Problems

  Pinto uses AI to detect upcoming risks such as:

  • Stock-outs

  • Over-stock

  • Supplier delays

  • Demand surges

  • Misaligned inventory across locations

  This lets teams act before the disruption happens.

• Recommends the Best Actions

  Instead of just showing data, Pinto guides users by saying:

  • “Move 120 units from Warehouse A to Warehouse B.”

  • “Expedite PO #234.”

  • “Increase buffer stock for SKU S91.”

  It gives ready-to-execute recommendations.

• Optimizes Internal Transfers

  Pinto identifies imbalances across warehouses and suggests:

  • Where stock should come from

  • Where it should go

  • How much to move

  • When to move it

  This reduces both excess inventory and shortages.

• Runs What-If Simulations

  Users can test scenarios like shipment delays, demand spikes, or warehouse closures.Pinto shows the impact and generates the best mitigation plan.

• Aligns Teams Across Supply Chain, Finance, and Ops

  Everyone sees the same risks, predictions, and actions.This improves coordination and reduces decision lag.

• Recommendation
  Pinto provides recommendation for the user with multiple options and also with the details. it will provide upto 3 options so that the user can select from them.

Investigating the alerts

Recommendation alerts

Where in this the pinto AI bot conversation, the details that you see is of heavy data and therefore too much of scrolling would be a pain for the customer and also the data that is presented

Active Simulation

Active Simulations are real-time, continuously running scenario models inside Pinto that monitor live data and automatically evaluate “what-if” situations without the user needing to trigger them manually. List of options that are available right now

• Demand Surge Simulation
  Demand pace unexpected demand increase / decrease across channel

• Supply Disruption
  Simulate supplier cancellations or delays

• PO Disruption

  Simulate purchase order cancellations or delays

• Channel Shift

  Demand moving between online/offline channels

The below screen displayes the -

• High Impact Stockout Timeline

• Items at risk, Potential Loss, Minimum days to stockout, Well stocked

• It will display the list with the actions to monitor them at which it can give the recommendations for the user to work on the items

Impact :

These results are the real time results obtained from various clients

Milestones

• Architected and launched Al-enhanced Inventory Planning System,achieving rapid market validation with 3 signed clients post-launch

• GO-LIVE for one of our clients on November

Outcome

The final design of Pint Control Center helped teams move from reactive firefighting to proactive planning.Users now detect risks early, take guided actions, optimise stock movements, improve service levels, and reduce excess inventory.