Legion Technologies
How a workforce management platform built "Logan," a custom AI support agent with audience-aware knowledge base controls, to scale support without scaling headcount.
Industry: Workforce Management Software
Client Profile: Legion Technologies builds AI-powered workforce management software for enterprise clients. Their platform handles demand forecasting, automated scheduling, and labor optimization for large-scale operations.
Key Results
- 20–30 hours/month of agent time recovered from automated responses
- Zero internal content leaks to external customers since launch
- Email support pilot validated for second AI-assisted channel
About the Client
Legion Technologies builds AI-powered workforce management software for enterprise clients. Their platform handles demand forecasting, automated scheduling, and labor optimization for large-scale operations. As their customer base grew, so did the pressure on their support team. They needed a way to handle more inquiries without hiring proportionally more agents.
Client Voice
Custom AI agents like Logan handle the routine questions so your team can focus on the problems that actually need a human. Built inside Zendesk, grounded in your knowledge base, and scoped to respect audience boundaries.
Scaling support but can't keep hiring agents at the same rate?
The Challenge
Legion's support team fielded questions from two very different audiences: internal teams and external customers. Both groups needed fast answers, but they needed different answers.
The Scoping Problem
Internal staff required access to the full knowledge base, including operational procedures and internal documentation. External customers needed help articles, product guides, and troubleshooting steps—but nothing internal-facing. A single AI agent answering from one pool of articles could easily surface internal-only documentation to a customer. That's a compliance risk and a trust issue.
Manual Triage Bottleneck
Manual triage kept things safe but slow. The team spent time on repetitive questions that a well-configured AI agent could handle in seconds—but only if that agent understood the boundary between internal and external content.
The Solution
The team built "Logan," a custom AI agent inside Legion's Zendesk environment. Logan does three things: answers common support questions automatically, pulls responses from Legion's Help Center documentation, and respects strict audience boundaries.
AI Agent Design & Deployment
Logan was designed, tested, and launched inside Legion's production Zendesk instance. It handles routine support inquiries without human intervention, pulling answers directly from official Help Center documentation. As articles are updated, Logan's responses update too.
Audience Scoping Architecture
Rather than building a custom permissions layer from scratch, the team used Zendesk's native constructs in a non-obvious way:
- An integration user configured as an End User with restricted Help Center access creates a natural boundary
- When Logan answers an external customer, it queries through this restricted user's permissions and only sees external articles
- Internal queries bypass this restriction
- Article labels and URL-based rules add a second layer of control
Email Support Pilot
A separate pilot validated automated email triage and response for Legion's support workflows, opening the door to a second AI-assisted channel using the same knowledge base architecture.
The Results
By the Numbers
- 150–200 routine tickets/month now handled automatically by Logan
- 8–12 minutes saved per ticket on common inquiries (previously required human lookup and response)
- ~20–30 hours/month of agent capacity recovered
- Zero incidents of internal documentation surfacing to external customers
Operational Impact
Logan handles routine inquiries automatically in production. The support team spends less time on repetitive questions and more on complex, high-value issues.
The email support proof-of-concept validated a second channel for AI-assisted responses. Legion can expand automated support to email without rebuilding the knowledge base architecture.
Perhaps most importantly, the internal/external content boundary has held. Zero incidents of internal documentation surfacing to external customers since launch. That's not just an efficiency win—it's a trust and compliance safeguard that scales with every new article added to the knowledge base.
Custom AI agents like Logan handle the routine questions so your team can focus on the problems that actually need a human. Built inside Zendesk, grounded in your knowledge base, and scoped to respect audience boundaries.
Scaling support but can't keep hiring agents at the same rate?