Predictive AI
Designing a GenAI platform for proactive risk management in complex logistics operations
AI PLATFORM DESIGN
SAAS
ML & GENAI
*Certain details in this case study have been modified or omitted to protect confidentiality. All sensitive and proprietary information has been fully anonymized.
[01]
INTRODUCTION
Overview
CLIENT
FlowTrack - B2B SaaS platform (ERP) used by logistics companies to manage end-to-end operations, from compliance to procurement.
PLATFORM
Orion - AI-driven intelligence layer delivering risk prediction, decision support and operational automation on top of the FlowTrack ERP.
MY ROLE
Lead Product Designer on behalf of Kido
TIMELINE
Q3 2024 - Q1 2026
The Problem
High-Stakes Logistics
Logistics operations operate in constant risk conditions. Minor disruptions, such as weather changes, supply shortages, etc., quickly cascade into delays, financial loss, and compliance exposure. Despite years of accumulated operational data, FlowTrack lacked real-time visibility and predictive intelligence.
0%
Real-time predictive risk capabilities
12
Years of operational data unused for foresight
~48–72h
Average delay in detecting operational issues
The Mission
From Reactive Operations to Proactive Risk Management
Extend FlowTrack with Orion, a generative AI platform that transforms logistics operations into a proactive, intelligence-driven system by delivering:
1
Predictive risk detection
Early identification of emerging risks with actionable alerts.
2
AI operations orchestration
Enable teams to act quickly via AI insights, automated workflows, and data-driven decision support.
[02]
RESEARCH
Continuous User Research
From the outset, Orion was guided by continuous user research grounded in day-to-day needs. To gather actionable feedback, we held weekly sessions with customer managers, conducted client interviews, and developed bi-weekly prototypes presented by the CEO to users and prospects, consistently generating insights that shaped the product. Key insights we gathered:
1
Prioritize real-time actions
Many work in high-stress field conditions and prioritize quick decisions over detailed reports.
2
Bridge data gaps across roles
In this hierarchical industry, critical data is limited to higher ranks, leaving frontline roles less visible.
3
Support protocol-driven decision making
Workflows are governed by strict protocols, but users often lack the guidance to follow them all.
4
Ensure workflow integration
AI layer should fit within ERP workflows to maintain established operational patterns.
Journey Mapping
Based on research insights, we mapped two primary journeys:
1
Push flow (system-initiated)
AI proactively detects risks and surfaces insights at the right moment.
2
Pull flow (user-initiated)
Users query the AI, explore scenarios, trigger actions, and configure automations.
In parallel, AI feedback-loop touchpoints were embedded directly into these journeys - allowing the system to learn from confirmations, overrides and outcomes from day one.
AI Trend Monitoring
Staying Ahead of the Rapidly Changing AI Landscape
Given the rapid evolution of AI, Orion was designed as a dynamic, adaptive system. We continuously monitored emerging AI interaction patterns and capabilities to meet evolving user expectations. Orion adapted in real time, ensuring long-term relevance.
Chat-based automation: we aligned Orion with this capability as soon as it launched in the AI Agents field in Jan' 25.
Voice mode feature: we immediately adopted voice-mode capabilities upon their Sep' 24 release.
[03]
DESIGN PROCESS
Shaping the AI Experience
Defining Platform Structure and Behavior
As a separate product in FlowTrack’s ecosystem, Orion required seamless ERP integration to ensure both platforms worked in harmony, letting users switch between them without friction and maintain their familiar workflows.
After exploring several solutions, we implemented an ERP overlay that preserved context and used familiar patterns to keep the experience intuitive and consistent.
Laying the Foundations
Defining Orion’s Identity
As a new product in the FlowTrack ecosystem, Orion required a distinct AI-driven identity. We positioned it as the first dark-mode product, creating a modern, intelligent presence.
We extended the design system with a dedicated color mode, reused core components for speed and consistency, and followed a design-system-first approach to enable fast, aligned iterations.
[04]
FINAL DESIGN
Actionable Risk Units
Orion detects risks and surfaces insights at the right moment. Risks are structured as operational decision units, not alerts. Each follows a three-layer flow that turns detection into action. Users see clear next steps first, then can drill down to understand the reasoning, balancing speed with insight:
1
Recommended actions
Context-aware steps, prioritized by impact and effort.
2
Risk overview & supporting data
Framing the situation with real-time data, trends, and anomalies.
3
Risk analysis
AI-generated insights explaining contributing factors.
AI Workflow Touchpoints
We treated user trust and continuous machine learning as core features, not just small additions. Integrated into daily operations, the AI becomes a reliable, evolving operational partner rather than merely a tool.
Building User Trust
Users see supporting data, explanations, and predictable reasoning behind every AI recommendation, allowing validation and confident action.
Continuous Machine Learning
User feedback is incorporated into every critical interaction, allowing the system to learn from real operational decisions without disrupting workflows while continuously improving.
[05]
OUTCOME & IMPACT
What We Achieved
Orion became an AI-driven operational partner, enabling faster, more confident decisions and delivering measurable improvements across workflows. The estimated impact:
~0h
Reduced risk detection time from 2–3 days to near real-time.
X2
Decision speed, with reduced error rates.
90%
Untapped operational data now leveraged.
100%
Clear ownership through role-based actions and real-time status.
Reflection
Calibrating Trust
With Orion integrated into decision-making, after significant effort to make it trusted and reliable, the next challenge is calibrating user trust - helping people understand when to rely on AI outputs, how they were generated, and their limitations. This NN/g article on explainable AI underscores the importance of responsible design as the platform scales.
Responsible AI is now as crucial as intelligence itself, ensuring Orion remains a trusted partner rather than an unreliable or harmful.
[06]
MORE PROJECTS
Operational Command Chat Layer
Streamlining Operations in the Field
Orion’s chat layer acts as a command surface, generating pre-defined components such as text, visuals, timelines, and voice input to deliver information quickly and clearly.
Built for fast-paced field environments, it supports all operational workflows and streamlines decision-making, enabling teams at all levels to access internal and external data without relying on senior staff.
Designing a GenAI platform for proactive risk management in complex logistics operations
AI PLATFORM DESIGN
SAAS
ML & GENAI
Designing a GenAI platform for proactive risk management in complex logistics operations
AI PLATFORM DESIGN
SAAS
ML & GENAI
*Certain details in this case study have been modified or omitted to protect confidentiality. All sensitive and proprietary information has been fully anonymized.
[01]
INTRODUCTION
Overview
Overview
CLIENT
FlowTrack - B2B SaaS platform (ERP) used by logistics companies to manage end-to-end operations, from compliance to procurement.
PLATFORM
Orion - AI-driven intelligence layer delivering risk prediction, decision support and operational automation on top of the FlowTrack ERP.
MY ROLE
Lead Product Designer on behalf of Kido
TIMELINE
Q3 2024 - Q1 2026
The Problem
The Problem
High-Stakes Logistics
Logistics operations operate in constant risk conditions. Minor disruptions, such as weather changes, supply shortages, etc., quickly cascade into delays, financial loss, and compliance exposure. Despite years of accumulated operational data, FlowTrack lacked real-time visibility and predictive intelligence.
~48–72h
Average delay in detecting operational issues
12
Years of operational data unused for foresight
0%
Real-time predictive risk capabilities
The Mission
The Mission
From Reactive Operations to Proactive Risk Management
Extend FlowTrack with Orion, a generative AI platform that transforms logistics operations into a proactive, intelligence-driven system by delivering:
1
Predictive risk detection
Early identification of emerging risks with actionable alerts.
2
AI operations orchestration
Enable teams to act quickly via AI insights, automated workflows, and data-driven decision support.
[02]
RESEARCH
Continuous User Research
Continuous User Research
Orion was shaped through continuous research, including weekly sessions with customer managers, client interviews, and bi-weekly prototypes presented to users and prospects - consistently generating insights that shaped the product. Key insights we gathered:
1
Real-time actions
Many work in high-stress field conditions and prioritize quick decisions over detailed reports.
2
Data gaps across roles
In this hierarchical industry, critical data is limited to higher ranks, leaving frontline roles less visible.
3
Protocol-driven decision making
Workflows are governed by strict protocols, but users often lack the guidance to follow them all.
4
Workflow integration
AI layer should fit within ERP workflows to maintain established operational patterns.
Journey Mapping
Journey Mapping
Based on research insights, we mapped two primary journeys:
1
Push flow (system-initiated)
AI proactively detects risks and surfaces insights at the right moment.
2
Pull flow (user-initiated)
Users query the AI, explore scenarios, trigger actions, and configure automations.
In parallel, AI feedback-loop touchpoints were embedded directly into these journeys - allowing the system to learn from confirmations, overrides and outcomes from day one.
Double click to explore
AI Trend Monitoring
AI Trend Monitoring
Staying Ahead of the Rapidly Changing AI Landscape
Given the rapid evolution of AI, Orion was designed as a dynamic, adaptive system. We continuously monitored emerging AI interaction patterns and capabilities to meet evolving user expectations. Orion adapted in real time, ensuring long-term relevance.
Chat-based automation & scheduled Actions: we aligned Orion with this capability as soon as it launched in the AI Agents field in Jan' 25.
Voice chat mode feature: we immediately adopted voice-mode capabilities upon their Sep' 24 release.
[03]
DESIGN PROCESS
Shaping the AI Experience
Shaping the AI Experience
Defining Platform Structure and Behavior
As a separate product in FlowTrack’s ecosystem, Orion required seamless ERP integration to ensure both platforms worked in harmony, letting users switch between them without friction and maintain their familiar workflows.
After exploring several solutions, we implemented an ERP overlay that preserved context and used familiar patterns to keep the experience intuitive and consistent.
Laying the Foundations
Laying the Foundations
Defining Orion’s Identity
As a new product in the FlowTrack ecosystem, Orion required a distinct AI-driven identity. We positioned it as the first dark-mode product, creating a modern, intelligent presence.
We extended the design system with a dedicated color mode, reused core components for speed and consistency, and followed a design-system-first approach to enable fast, aligned iterations.
[04]
FINAL DESIGN
Actionable Risk Units
Actionable Risk Units
Orion detects risks and surfaces insights at the right moment. Risks are structured as operational decision units, not alerts. Each follows a three-layer flow that turns detection into action. Users see clear next steps first, then can drill down to understand the reasoning, balancing speed with insight:
1
Recommended actions
Context-aware steps, prioritized by impact and effort.
2
Risk overview & supporting data
Framing the situation with real-time data, trends, and anomalies.
3
Risk analysis
AI-generated insights explaining contributing factors.
Operational Command Chat Layer
Operational Command Chat Layer
Streamlining Operations in the Field
Orion’s chat layer acts as a command surface, generating pre-defined components such as text, visuals, timelines, and voice input to deliver information quickly and clearly.
Built for fast-paced field environments, it supports all operational workflows and streamlines decision-making, enabling teams at all levels to access internal and external data without relying on senior staff.
AI Workflow Touchpoints
AI Workflow Touchpoints
We treated user trust and continuous machine learning as core features, not just small additions. Integrated into daily operations, the AI becomes a reliable, evolving operational partner rather than merely a tool.
Building User Trust
Users see supporting data, explanations, and predictable reasoning behind every AI recommendation, allowing validation and confident action.
Continuous Machine Learning
User feedback is incorporated into every critical interaction, allowing the system to learn from real operational decisions without disrupting workflows while continuously improving.
[05]
OUTCOME & IMPACT
What We Achieved
What We Achieved
Orion became an AI-driven operational partner, enabling faster, more confident decisions and delivering measurable improvements across workflows. The estimated impact:
~0h
Reduced risk detection time from 2–3 days to near real-time.
X2
Decision speed, with reduced error rates.
90%
Untapped operational data now leveraged.
100%
Clear ownership through role-based actions and real-time status.
Reflection
Reflection
Calibrating Trust
With Orion integrated into decision-making, after significant effort to make it trusted and reliable, the next challenge is calibrating user trust - helping people understand when to rely on AI outputs, how they were generated, and their limitations. This NN/g article on explainable AI underscores the importance of responsible design as the platform scales.
Responsible AI is now as crucial as intelligence itself, ensuring Orion remains a trusted partner rather than an unreliable or harmful.
[06]
MORE PROJECTS
