AI Deflection

Track contacts resolved by chatbots, IVR, and virtual agents before they reach live agents, and see the impact on your staffing requirements.

AI Deflection Analytics is available on the Pro and Enterprise subscription tiers.

Overview

AI deflection measures the percentage of inbound contacts that are resolved automatically — by a chatbot, IVR self-service menu, virtual agent, or customer self-service portal — without reaching a live agent. Tracking deflection helps you:

Right-size staffing: Reduce forecast demand to reflect only live-agent contacts
Measure AI ROI: See exactly how many contacts your automation handles
Spot trends: Identify channels and queues where deflection is growing or declining
Forecast accurately: Apply deflection rates to staffing calculations

Key Formula

Deflection Rate = Contacts Deflected / Total Contacts Offered × 100

Getting Started

1. Import Deflection Data

Deflection data is imported via the External Integration API. Your chatbot or IVR platform pushes deflection counts per interval alongside your regular ACD data.

bash

curl -X POST https://ezywfm.com/api/v1/import/deflection \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "data": [{
      "date": "2026-03-15",
      "interval": "08:00",
      "contactsDeflected": 12,
      "totalContactsOffered": 45,
      "queue": "support",
      "channel": "chat",
      "deflectionSource": "chatbot"
    }]
  }'

Import deflection data on the same interval cadence as your historical contact data (typically every 15 or 30 minutes) for the most accurate analytics.

2. View Deflection Analytics

Go to Analytics in the sidebar
Scroll to the AI Deflection section
Select your time period (7d, 30d, 90d)

Analytics Dashboard

KPI Cards

The deflection summary shows four key metrics:

| Metric | Description | |--------|-------------| | Deflection Rate | Percentage of contacts resolved by AI | | Contacts Deflected | Total count of deflected contacts in the period | | Est. Agents Saved | Estimated agents no longer needed per interval (via Erlang C) | | Top Source | The deflection source handling the most contacts |

Deflection Trend Chart

The trend chart shows daily data with:

Stacked bars: Live-agent contacts vs. deflected contacts
Line overlay: Deflection rate percentage over time

Use this to identify whether your deflection rate is trending up (AI is handling more) or down (contacts may be leaking through to agents).

Breakdown by Source

See which automation handles the most contacts:

| Source | Description | |--------|-------------| | chatbot | Web or messaging chatbot | | ivr | Interactive voice response self-service | | virtual_agent | AI-powered voice or chat agent | | self_service | Customer self-service portal |

Breakdown by Channel

View deflection rates across communication channels (voice, chat, email, SMS). Chat and voice typically have the highest deflection potential.

Using Deflection in Forecasting

The forecast generator includes a built-in deflection rate adjustment.

Applying a Deflection Rate

Go to Forecasting → Generate Forecast
In the configuration panel, set AI Deflection (%) to your expected rate
Click Generate Forecast

The forecast will show:

Total Volume: Predicted total contacts (before deflection)
Deflected: Contacts expected to be handled by AI
Live Agent: Remaining contacts requiring live agents
Peak Agents / Scheduled: Staffing based on live-agent demand only

Use your 30-day average deflection rate from the Analytics dashboard as a starting point. Adjust upward if you're rolling out new automation.

How It Works

The deflection adjustment is applied after the forecast is generated but before staffing is calculated:

Live Agent Volume = Predicted Volume × (1 - Deflection Rate)
Required Agents = Erlang C(Live Agent Volume, AHT, SLA Target)

This means confidence intervals and staffing requirements all reflect the reduced demand.

Exporting Deflection Data

Export deflection data as CSV:

Go to Analytics
Click Export
Select AI Deflection

The CSV includes: Date, Queue, Channel, Source, Contacts Offered, Contacts Deflected, and Deflection Rate %.

You can also export via the API: