AI-Powered Request Routing
E-Wolf Support is an AI-augmented request-routing platform designed to serve FRCC's 50,000+ user with faster, more accurate support. After analyzing the failures of the previous AI chatbot, which frequently surfaced incorrect deadlines and policy information, I designed a human-in-the-loop model where AI classifies inquiries and drafts responses while staff verify all outbound communication. The system is built around a confidence-gated routing mechanism: every classification is scored 0–100, and an 85% threshold determines whether a ticket auto-routes to the correct department or lands in a staff review queue for human judgment.
This workflow maintains 100% accuracy for institutional information, reduces manual sorting time by 30–40%, and saves staff an estimated 5–6 hours per week that were previously spent triaging misrouted requests. For students, it delivers clearer routing and more trustworthy guidance; for FRCC, it reduces operational overhead, improves compliance, and creates a scalable foundation for future AI-enabled services. In March 2026, FRCC's cabinet approved $200K in Year 1 funding to take E-Wolf Support into production, with Gecko Engage scoped as build partner.
Timeline
Nov'25 - Under Production
AI Prototyping Tool
Loveable, Claude Code
Role
Senior Product Designer
Led product strategy, research, information architecture, and interaction design. Designed and configured the LLM routing logic classification, taxonomy, confidence thresholds, and the human-in-the-loop review flow — that powers the system.
A Stanford-University–Led Research Project Funded by FRCC
The Investment Case That Won Production Funding
Presented to FRCC leadership in March 2026 as the formal investment case for E-Wolf Support. Cabinet approved $200K in Year 1 production funding; Gecko Engage was scoped as production lead. What follows is the full deck — strategic alignment, the confidence-gated architecture in detail, the testing-phase outcomes, and the production roadmap through Q4 2026.
The Challenge
Fragmented Contact System
Students faced a complex contact page with 13+ department options and no clear path. Unable to determine the correct recipient, most students defaulted to sending all requests to E-Wolf, regardless of which department actually owned the issue.
User faced FRCC's contact page with multiple department options.
Unreliable AI Chatbot
FRCC’s new AI chatbot routinely surfaced incorrect or incomplete answers on critical topics like deadlines, payments, and financial aid. With no human verification layer, students couldn’t trust the guidance they received—leading to misinformation risks and confusion.
Because of this, the chatbot did not reduce support workload at all. Students continued calling and emailing for clarification, leaving staff with the same volume of requests the bot was meant to relieve.
Legacy chatbot responses were repetitive and unreliable for sensitive topics like tuition—highlighting the need for a human-verified support system.
Manual Routing
E-Wolf staff spent 5–6 hours per week manually categorizing and routing student requests to appropriate departments. A review of 1,000+ student support submissions revealed frequent misrouting, unclear categories, and heavy manual triage. This dataset became the core training source for E-Wolf's AI-assisted routing model and the empirical foundation for the testing-phase results that earned cabinet approval.
A review of 1,000+ student support submissions revealed frequent misrouting, unclear categories, and heavy manual triage. This dataset became the core training source for E-Wolf’s AI-assisted routing model.
High Level Goal
How might I design an AI-augmented support system that reduces manual categorization workload for E-Wolf staff while improving response clarity and routing accuracy for students — while maintaining the human oversight, professional judgment, and empathy essential to effective student support?
Research & Discovery
Understanding the needs of both students and E-Wolf staff through data analysis and user interviews.
Chatbot Evaluation
Tested FRCC's existing AI chatbot to document accuracy issues with critical information.
Workflow Mapping
Mapped current support workflows to identify inefficiencies
User Interviews
Conducted interviews with E-Wolf staff and students about pain points
Data Analysis
Analyzed Formstack submission data to identify patterns and common request type.
Key Findings
Chatbot Analysis
Existing chatbot provided outdated financial aid deadlines during testing
Gave inconsistent answers to the same question asked differently
Used a confident tone even when wrong—no way for students to verify accuracy
At institutional level, AI errors are unacceptable for critical information
Student Perspective
Students want immediate acknowledgment of their requests
Prefer clear, contextual responses over generic replies
Need to know their request reached the right department
Staff Perspective
Need to maintain oversight over AI-generated responses
Want AI to handle routine categorization while they focus on complex cases
Require ability to refine and personalize AI suggestions
Early Design Work
The solution had to create a collaborative relationship between AI and humans, where AI handles pattern recognition and initial categorization while staff provide the nuance, empathy, and contextual understanding that students need.
Defining Structure
With insights in hand, I sketched an updated information structure and user flows. The structure was organized around two axes: user-facing portals and AI-powered backend processes.
Student-Facing Surfaces Request Submission Form · AI-Generated Response · Status Tracking · Department Routing Display
Staff-Facing Surfaces Request Queue Management · AI Suggestion Review · Response Editing Interface · Performance Analytics
AI Processing Layer
Natural Language Processing: analyzes student request text for intent, keywords, and context
Department Classification: maps requests to FRCC departments with confidence scoring
Response Generation: creates contextual draft responses for staff review
The Mechanism — A Confidence Gate
Every AI classification is scored 0–100. An 85% threshold determines what happens next: tickets above it auto-route to the correct department with a full audit trail; tickets below it land in a staff review queue for one-click reassignment. Every override becomes labeled training data.
The design was deliberate. Higher-ed AI tools fail in a recognizable pattern — confident on general questions, wrong on specific ones, with no way for users to verify the answer. The confidence gate inverts that pattern. Rather than hiding uncertainty, the system surfaces it — which is what makes the model safe to deploy in a FERPA-governed environment.
In testing, 91% of tickets cleared the gate on first pass. The remaining 9% were exactly the cases where staff judgment was needed.
The Institutional Unlock — Cross-Departmental Data Integration
The hardest problem at FRCC wasn't routing speed. It was data fragmentation. Wolf Central, Advising, Financial Aid, and the Registrar each held their own ticket data in their own systems, with no shared queue, no shared metrics, and no visibility into where a request stalled once it crossed a team boundary.
E-Wolf Support is the first system at FRCC to integrate ticket data across all four departments into a single routing layer. The same data that powers AI classification feeds an analytics view that surfaces routing accuracy by team, bottlenecks by department, cross-team friction patterns, and evidence-based staffing recommendations.
This was the strongest argument to cabinet leadership. Comparable cross-departmental integrations at peer institutions run $290K–$760K per year through Salesforce or EAB. E-Wolf delivers the same institutional capability on FRCC's existing infrastructure, with Gecko Engage as production lead.
The Solution
Design Philosophy: Augmentation, Not Automation
Unlike traditional automation that replaces human workers, this system is designed to augment staff capabilities. The AI handles the repetitive pattern recognition and initial categorization, freeing staff to focus on what humans do best: providing empathy, nuanced judgment, and personalized support
This creates a partnership model where AI and humans learn from each other—AI improves through staff corrections, while staff benefit from AI's ability to process patterns at scale.
A two-part system that combines AI intelligence with human oversight to create better outcomes for everyone.
Smart Request Submission (Student View)
On the student submission side, AI analyzes the user’s issue description and suggests relevant departments (e.g., Academic Support, Financial Aid, Registration). This lowers cognitive load for students who may not know which office to contact, while giving staff clean, pre-structured tickets for faster handling. Students can still override suggestions, ensuring transparency and control.
Staff Dashboard & Core Features (Staff View)
The staff-facing dashboard centralizes all student requests and layers AI assistance on top of existing workflows. The interface provides real-time visibility into workload, routing precision, and departmental trends—transforming a previously fragmented support process into a unified, data-informed operations hub.
Key capabilities include:
1. Unified Request Management
Staff can view all pending, in-progress, and completed requests in one place, supported by search, filtering, and clear status indicators. This consolidates email-based workflows and provides a single operational view for all student support activity.
A dedicated AI Accuracy metric on the dashboard tracks model performance in real time, allowing staff to understand how reliably the system classifies requests and where manual oversight may be needed.
2. AI-Powered Categorization & Routing and AI-Assisted Response Drafting
Every incoming request is analyzed by the model and assigned a predicted department with a confidence score. This dramatically reduces manual triage time and improves routing consistency across 13+ service areas.
For each request, the system generates a draft response aligned with institutional policies and tone. Staff can edit, refine, or regenerate the message before sending—maintaining 100% human oversight while accelerating turnaround times.
3. Departmental Analytics & Operational Insights
The analytics view highlights request load, distribution, and throughput across all departments. Donut charts and stacked bar graphs make it easy to identify bottlenecks, spike periods, and under-resourced areas. These insights enable data-driven staffing decisions and help leadership anticipate student needs more proactively.
Testing Results & Projected Impact
Usability Testing Metrics
91%
AI Categorization Accuracy
Validated across 300+ test scenarios
5-6h
Projected Weekly Time Savings
Based on current manual routing workload
100%
Human Oversight Integration
All AI suggestions require staff approval
8
Department Coverage
Comprehensive routing across FRCC services
Testing Methodology
Sample Size: Prototype tested with 300+ authentic student request scenarios sourced from historical E-Wolf support data
Evaluation Criteria: AI routing accuracy, response completeness, confidence score reliability, and staff workflow integration
Stakeholder Validation: System design reviewed and approved by E-Wolf staff members who will utilize the platform upon implementation
Benefits for Students
Instant Initial Response: AI provides immediate acknowledgment and relevant information while staff prepare personalized follow-up
Accurate Department Routing: 91% accuracy means requests reach the right expert the first time
No More Request Ping-Pong: Students don't get bounced between departments trying to find help
24/7 Submission: Submit requests anytime; AI categorizes and queues for staff review
Benefits for E-Wolf Staff
5-6 Hours Saved Weekly: AI handles initial categorization, freeing staff for complex cases
AI-Generated Draft Responses: Start with a smart draft, then add personal touch and department-specific details
Transparency & Control: See AI confidence scores, override suggestions, and maintain full oversight
Focus on High-Impact Work: Spend time on empathy, nuance, and complex student situations
Personal Takeaway
Scaling Impact by Simplifying Complexity
The most meaningful insight from this work was seeing how much cognitive load and operational inefficiency came from unclear routing and inconsistent messaging. By simplifying the flow, tightening classifications, and designing AI-assisted tools that actually support staff, I saw how UX can unlock organizational efficiency—not by adding more technology, but by ensuring it’s used thoughtfully. This project strengthened my conviction that the best UX outcomes happen when technology, process, and human judgment work together, not in isolation.