Agentic Companion AI for PEA Vendor Restrictions

Overview

Pro associates building large fulfillment orders often encounter complex vendor restrictions that are difficult to interpret and easy to miss — minimum order quantities, delivery thresholds, truck-type limitations, stair carry fees, split delivery requirements, and item-specific service constraints.

The existing experience relied heavily on static banners, tooltips, and fulfillment documentation buried throughout the Product Detail Page (PDP). Associates frequently overlooked important information, resulting in invalid carts, fulfillment rework, delivery delays, and reduced confidence during customer interactions.

I explored an Agentic Companion AI experience embedded directly within the PEA fulfillment workflow — a conversational assistant capable of proactively identifying fulfillment constraints, explaining them in plain language, and helping associates configure valid vendor delivery solutions in real time.

Problem

Vendor fulfillment logic is highly nuanced and varies by item, vendor, market, and delivery method. Associates needed to understand:

• Which delivery methods were eligible
• Quantity minimums and order multiples
• Service fees and thresholds
• Truck restrictions
• Jobsite delivery capabilities
• When split delivery was required
• Additional service configurations like Boom & Scatter

The existing experience created several usability issues:

• Important fulfillment information suffered from banner blindness
• Associates had to mentally synthesize rules across multiple UI regions
• Restrictions were explained in system language rather than conversational language
• Errors were often discovered late in the workflow
• Associates lacked confidence when explaining fulfillment options to Pro customers

The challenge was not simply displaying information — it was helping associates understand and act on it correctly.

Goal

Design an AI-assisted fulfillment companion that could:

• Translate complex vendor restrictions into simple, actionable guidance
• Help associates build valid carts faster
• Reduce fulfillment configuration errors
• Increase confidence during customer conversations
• Surface relevant information at the right moment in the workflow
• Maintain efficiency for high-volume Pro ordering scenarios

Understanding the user

The primary user was the Pro Specialist (PSS) working in-store with contractors and Pro customers — high-pressure environments where speed and accuracy are critical.

Through prior fulfillment research, associates preferred:

• Quick recommendations over documentation
• Plain-language explanations
• Guided decision-making
• Minimal interruption to workflow

Associates struggled with:

• Hidden restrictions
• Vendor terminology
• Understanding pricing implications
• Remembering rules across multiple items

This revealed an opportunity for AI to act less like a chatbot and more like a contextual fulfillment partner.

Design principles

• Guidance over automation Assist associates without fully taking control of the workflow.
• Progressive disclosure Only surface information when it becomes relevant.
• Conversational clarity Translate operational rules into customer-friendly language.
• Workflow efficiency Avoid additional clicks or modal interruptions.
• Context awareness Adapt to cart state, item eligibility, and fulfillment selections.

Exploration

Early concept: Reactive assistant

First explorations focused on a traditional support model where associates manually opened an AI assistant to ask questions about delivery restrictions.

While functional, this approach had problems:

• It relied on associates knowing when they needed help
• It introduced extra interaction cost
• It behaved more like a help center than an intelligent companion

This led to a strategic shift toward a more proactive model.

Final direction: Agentic Companion AI

The evolved concept positioned AI as a contextual fulfillment companion embedded directly within the workflow. Instead of waiting for user prompts, the AI would:

• Detect eligible fulfillment scenarios
• Recommend delivery methods
• Explain why options mattered
• Identify invalid configurations
• Suggest corrections
• Reuse fulfillment configurations across similar items

The interaction model blended inline recommendations, lightweight conversational prompts, contextual confirmations, and smart automation opportunities.

Key experience moments

1. Intelligent fulfillment recommendation

When an associate selected Vendor Delivery, the AI proactively explained why it was the correct path — e.g., “Vendor Delivery includes jobsite delivery services like Boom & Scatter that may better support this customer’s needs.” Associates often knew what to select but not why; the AI reinforced confidence while educating users naturally.

2. Explaining complex restrictions in plain language

Instead of technical messaging like “Minimum quantity threshold not met for vendor truck type,” the AI translated constraints into actionable guidance — e.g., “This item requires a larger delivery quantity before Boom delivery becomes available.”

3. Recommending valid cart configurations

As associates added products, the AI monitored eligibility and surfaced recommendations such as switching truck types, splitting deliveries, adjusting quantities, or applying matching fulfillment setups — shifting from reactive error handling to proactive cart guidance.

4. Cross-item fulfillment reuse

When a newly viewed item qualified for the same fulfillment setup as items already in the cart, the AI offered to apply the configuration automatically — an important efficiency unlock for large Pro orders.

UX challenges

Balancing visibility with disruption

Too passive and guidance would still be missed; too aggressive and the experience became noisy. The solution was layered: high-priority issues surfaced inline, recommendations appeared contextually, and secondary information stayed accessible but unobtrusive.

Maintaining user trust

Associates needed to understand why recommendations were made. Rather than “Use Boom Delivery,” the AI explained reasoning — e.g., “Boom Delivery supports rooftop placement for this quantity and item type” — reducing skepticism toward automation.

Outcome

The concept demonstrated how conversational AI could improve fulfillment comprehension without requiring associates to leave their workflow.

• Reduced cognitive load — complex restrictions easier to understand and act on
• Faster cart building — reuse fulfillment configurations across eligible items
• Increased confidence — AI explanations support customer conversations in real time
• Better error prevention — issues identified before fulfillment submission
• Scalable guidance model — framework extendable to fees, scheduling, installation, and inventory

Reflection

This project challenged me to think beyond static interface patterns and design for collaborative decision-making between humans and AI. Successful AI experiences are not about replacing user judgment — they reduce ambiguity and help users move forward confidently.

Rather than designing a chatbot, I focused on a fulfillment companion that understood workflow context, surfaced guidance at the right moment, and helped associates make better decisions with less effort.

My role

Lead UX Designer. Responsibilities included:

• UX strategy and concept development
• Conversational interaction design
• Information architecture
• AI workflow exploration
• PDP integration concepts
• Error prevention patterns
• Mobile and desktop behavior considerations
• Cross-functional collaboration with Product and Engineering

Skills demonstrated

• Conversational UX
• AI-assisted workflows
• Enterprise UX strategy
• Information architecture
• Contextual guidance systems
• Fulfillment & logistics UX
• Progressive disclosure
• Human–AI interaction
• Systems thinking
• Proactive error prevention