We Replaced L2/L3 Support with Origon AI

Customer support at scale has a structural problem: meaningful diagnosis requires system-level access. When issues involve deep technical dive, or billing disputes, resolution demands production data and multi-step diagnostics—capabilities that traditionally required senior resources.

At Samespace, this meant engineers were constantly pulled into support operations to access call logs, run network diagnostics, and validate configurations. We solved this by building a multi-agent support system on Origon with direct access to the same operational tools our engineers use.

The System

We built this on Origon, our agentic AI platform designed for production environments. Origon handles multi-agent orchestration, tool integration, and context management—eliminating the complexity of building custom coordination logic between agents, APIs, and data sources.

Three agents with defined boundaries:

Root agent manages customer interaction and delegates to specialists based on issue type.

Technical agent executes diagnostics: call quality analysis, network troubleshooting, configuration validation across our CX and VoiceCloud products.

Billing agent handles payment processing, account queries, and dispute resolution.

We built a custom MCP server over Hub, our infrastructure layer, giving agents authenticated access to:

• Customer databases and account systems
• Call detail records and failed call logs
• Network diagnostics and routing data
• Billing systems and payment APIs
• Product documentation and troubleshooting runbooks
• FreshDesk ticketing

What They Actually Do

These agents execute L2 and L3 support functions:

• Correlate failed calls across system logs to identify network path failures
• Analyze CDR patterns to diagnose audio quality issues
• Execute billing reconciliation across payment processors
• Validate SIP configurations and diagnose routing failures
• Run diagnostic queries across distributed systems

Technical Example: Customer reports intermittent call drops. The agent correlates CDRs with network logs, identifies a carrier gateway timeout pattern, validates against known issues, and either resolves through configuration change or escalates with complete diagnostic context.

Billing Example: Billing discrepancy on prorated charges. The agent retrieves usage data, validates pricing logic, recalculates amounts, cross-references payment records, and either confirms accuracy or initiates correction with full audit trail.

When escalation is needed, engineers receive actionable information: relevant logs, test results, timeline correlation, and failure hypothesis. No ambiguous tickets.

Results

We deployed a working system in one day. Within one week, it was handling production traffic.

Current performance:

• 70-80 support interactions resolved daily without human intervention
• Resolution time: days to minutes for diagnostic-heavy issues
• 24/7 global coverage without additional staffing
• Engineers focused on product development, not support operations

Why It Works

Origon's agentic architecture: Purpose-built for production environments. Handles agent coordination, tool routing, and state management without custom orchestration code. Agents operate autonomously within defined boundaries while maintaining context across interactions.

System-level access: Agents operate with the same data and tools engineers use. No surface-level chatbot responses.

Technical depth: Handles L2/L3 diagnostics that previously required senior resources. This isn't triage—it's resolution.

Precise escalation: When humans are needed, they get complete context. No re-explaining, no lost state.

Production-ready deployment: Built on Origon's unified platform. One week from prototype to production, including security review.

What's Next

We're adding capabilities for deeper diagnostics and critical escalation:

Real-time packet-level trace: Agents will perform network packet analysis to diagnose complex routing and quality issues at the protocol level.

Business continuity handoff: When critical issues impact operations, agents will initiate phone calls to relevant personnel—even in the middle of the night—ensuring immediate human intervention when it matters.

The system now handles the majority of our support operations autonomously, executing complex technical diagnostics that previously consumed engineering time.

Why This Matters for Every Enterprise

Here's what we learned using our own system in production: the barriers to enterprise AI automation have collapsed. Long development cycles, integration nightmares, security concerns—they're no longer roadblocks.

We deployed a production-ready, multi-channel, intelligent support system in under a week. It handles our actual customers, protects their data, and works better than we honestly expected. The ROI was immediate: reduced costs, improved satisfaction, 24/7 availability, and freed-up engineers to do actual work.

The future of customer service isn't coming. We're living it, and we built it in less time than most companies spend planning their quarterly roadmap. That's why we're excited to share Origon with the world.

Learn more about Origon at origon.ai