Overview
eigi.ai is built for enterprise scale. Whether you’re handling hundreds or thousands of concurrent calls, the platform delivers consistent performance and reliability.Capacity & Scale
Handle Any Volume
Concurrent Calls
Handle thousands of simultaneous conversations
Auto-Scaling
Capacity automatically adjusts to demand
Global Infrastructure
Low latency from anywhere in the world
No Code Changes
Scale without modifying your setup
Concurrency Management
How Concurrency Works
| Metric | Description |
|---|---|
| Concurrent Calls | Number of active calls at the same time |
| Peak Capacity | Maximum concurrent calls your plan supports |
| Queue Management | How overflow calls are handled |
| Burst Handling | Temporary capacity for traffic spikes |
Concurrency Tiers
| Tier | Concurrent Calls | Use Case |
|---|---|---|
| Starter | Up to 10 | Small businesses, testing |
| Growth | Up to 50 | Growing operations |
| Business | Up to 200 | Medium enterprises |
| Enterprise | Unlimited | Large-scale deployments |
Queue Management
When Capacity is Full
Configure how to handle calls when at capacity:Hold & Wait
Caller waits with music/message until agent available
Callback Queue
Collect number and call back automatically
Voicemail
Record message for later follow-up
Overflow Routing
Route to backup agent or human team
Queue Settings
| Setting | Description |
|---|---|
| Max Wait Time | How long callers wait before alternative |
| Position Announcements | Tell callers their queue position |
| Estimated Wait | Provide estimated wait time |
| Music/Message | What plays while waiting |
Performance Optimization
Real-Time Processing
Voice AI requires fast processing. eigi.ai is optimized for:| Component | Optimization |
|---|---|
| Speech Recognition | Stream processing for instant transcription |
| LLM Processing | Optimized inference with low latency |
| Voice Synthesis | Word-by-word streaming for natural flow |
| Tool Execution | Parallel processing when possible |
Latency Targets
| Metric | Target |
|---|---|
| First Response | < 800ms after user stops speaking |
| Turn-Taking | Natural conversation pace |
| Tool Calls | < 200ms overhead |
| Audio Quality | HD voice with < 50ms jitter |
Reliability
Uptime & Availability
99.9% Uptime
Enterprise SLA available
Redundancy
No single point of failure
Auto Recovery
Automatic failover
What Happens During Outages
| Scenario | Handling |
|---|---|
| Provider Outage | Automatic failover to backup |
| Partial Degradation | Non-critical features disabled |
| Maintenance | Zero-downtime deployments |
| Network Issues | Automatic reconnection |
Provider Redundancy
Multi-Provider Fallback
Configure backups for critical services:LLM Fallback
LLM Fallback
If primary LLM is slow or unavailable: - Primary: GPT-4o - Fallback 1:
Claude 3.5 Sonnet - Fallback 2: Gemini 2.0 Flash
TTS Fallback
TTS Fallback
If voice synthesis fails: - Primary: ElevenLabs - Fallback: Cartesia -
Emergency: OpenAI TTS
STT Fallback
STT Fallback
If speech recognition fails: - Primary: Deepgram - Fallback: Azure Speech
Load Balancing
Traffic Distribution
Calls are intelligently distributed:| Strategy | Description |
|---|---|
| Geographic | Route to nearest server |
| Load-Based | Route to least busy server |
| Provider Health | Avoid degraded providers |
| Cost Optimization | Balance performance and cost |
Monitoring
Real-Time Dashboards
Monitor your system health:Active Calls
Current concurrent calls and capacity
Queue Status
Callers waiting and wait times
Provider Status
Health of all connected services
Error Rates
Track and alert on issues
Key Metrics
| Metric | Description | Alert Threshold |
|---|---|---|
| Concurrency % | Current vs. max capacity | > 80% |
| Avg. Latency | Response time | > 1000ms |
| Error Rate | Failed calls | > 1% |
| Queue Length | Callers waiting | > 10 |
Analytics at Scale
Handling Large Data Volumes
Real-Time Analytics
Live metrics even during high volume
Historical Data
Access all historical call data
Data Retention
Configurable retention policies
Export & Backup
Export data for external analysis
Multi-Agent Scaling
Managing Multiple Agents
Scale across many agents:| Feature | Description |
|---|---|
| Shared Resources | Pool capacity across agents |
| Agent Prioritization | Allocate more capacity to critical agents |
| Load Spreading | Distribute traffic across agents |
| Independent Scaling | Scale specific agents as needed |
Enterprise Features
Advanced Capabilities
Dedicated Infrastructure
Isolated resources for your organization
Custom SLAs
Tailored uptime and support agreements
Priority Support
Direct access to engineering team
Compliance
HIPAA, SOC2, GDPR compliance options
Cost Optimization
Scale Efficiently
Right-Size Your Plan
Right-Size Your Plan
Choose concurrency limits that match your actual usage patterns.
Optimize Call Duration
Optimize Call Duration
Well-designed prompts lead to shorter, more efficient calls.
Use Appropriate Providers
Use Appropriate Providers
Match provider capabilities to task complexity.
Monitor and Adjust
Monitor and Adjust
Regularly review usage patterns and optimize.
Traffic Patterns
Handling Variable Load
Configure for your traffic patterns:| Pattern | Strategy |
|---|---|
| Steady | Consistent capacity allocation |
| Peak Hours | Scale up during busy times |
| Campaign Bursts | Reserve capacity for campaigns |
| Seasonal | Adjust for seasonal variations |
Testing at Scale
Load Testing
Before launching high-volume campaigns:1
Baseline Test
Measure performance at current volume
2
Gradual Increase
Slowly increase load to find limits
3
Stress Test
Test beyond expected maximum
4
Failure Testing
Verify graceful degradation
Best Practices
Troubleshooting
Calls being queued unexpectedly
Calls being queued unexpectedly
- Check current concurrency vs. limit - Review call duration (longer calls use capacity longer) - Consider upgrading your plan - Optimize agent prompts for efficiency
High latency during peak times
High latency during peak times
- Verify you’re not hitting capacity limits - Check provider status for degradation - Review if specific tools are causing delays - Consider geographic distribution
Inconsistent call quality
Inconsistent call quality
- Check network connectivity - Verify provider health status - Review audio settings - Test with different providers