Appendix

​

11. Appendix

​

A. Methodology & Data Sources

​

Market Sizing Approach

Global annual calls: 270 billion (IBM/TCS 2018 baseline, 3% CAGR)
Average Handle Time: 6 min 10 sec (Sprinklr cross-industry benchmark)
Total minutes: 270B × 6.17 min = 1,665 billion minutes (rounded to 1,647B)

• Gartner: 1.6% baseline (2022), 10% projection (2026)
• NICE: 4% midpoint (2024)
• EY/Quatrro (India): 30-50% (2025)

​

India-Specific Methodology

• Global contact center revenue: $352B (Grand View Research 2024)
• India contact center revenue: $33B (Ken Research, IBEF)
• India share: 9.4%
• Assumed similar share for voice minutes

• EY: 30-50% AI containment in voice+chat (2025)
• Rediff/Economic Times: Indian BPOs leading in AI adoption
• Washington Post: Labor cost compression driving faster shift

​

Competitive Intelligence Sources

• Public companies: SEC filings (10-K, 10-Q)
• Private companies: Funding press releases, GetLatka estimates, media reports
• India companies: MCA filings (Ministry of Corporate Affairs)

• Public pricing pages (as of October 2025)
• G2/Gartner Peer Insights reviews (mentioning pricing)
• Sales conversations (anonymized)

​

B. Key Company Profiles

​

Twilio (Public - TWLO)

• Market leader in CPaaS, but growth slowing (from 50%+ in 2020 to single digits)
• Facing commoditization in voice/SMS; investing in Segment (customer data platform) for differentiation
• Limited AI-native offerings (mostly third-party LLM integrations)

​

Bandwidth (Public - BAND)

• BYOC leader: Integrates with Genesys, Five9, Zoom Phone
• On-net advantage: National fiber network reduces wholesale costs
• Less exposed to AI disruption (infrastructure play, not application)

​

Replicant (Private)

• AI-native: Built for GPT-4-o latency from day one
• Vertical playbooks: Pre-configured for telecom, insurance, healthcare
• Competition: NICE, Cognigy (incumbents); PolyAI, Observe.ai (startups)

​

Uniphore (Private - India)

• India’s AI unicorn: Largest pure-play voice AI company in APAC
• Emotion detection: Differentiator for compliance/quality monitoring
• Expanding globally: 60% revenue now outside India

​

Exotel (Private - India)

• India’s leading cloud telephony provider (pre-dates voice AI wave)
• Ameyo acquisition (2021): Added enterprise CCaaS
• Transitioning to AI: “House of AI” launched 2023 (bots, analytics)
• Competition: Ozonetel, Knowlarity (now Gupshup), Airtel IQ, Tata Comms

​

LiveKit (Private)

• OSS core: 11k+ GitHub stars, community-driven adoption
• AI-native: “Agents” framework purpose-built for LLM latency (<200ms)
• Competition: Daily (similar model), Agora (larger but legacy), Twilio (commoditized)

​

C. Use Case Deep-Dives

​

Use Case 1: Telecom Customer Support

• Volume: 500M+ calls/year (major US carrier)
• Current cost: $7.50/call (human agent)
• Target: Automate 40% of routine inquiries (billing, tech support tier 1)

Intent Classification (first 10 seconds)
  ├─ Routine (75% of calls)
  │   ├─ Billing inquiry → AI bot (85% containment)
  │   ├─ Plan change → AI bot (70% containment)
  │   └─ Tech support tier 1 → AI bot (60% containment)
  └─ Complex (25% of calls)
      ├─ Escalation to human
      └─ AI agent assists human (co-pilot mode)

• Platform: Replicant + Twilio (SIP trunking)
• Deployment time: 6 months (pilot) + 12 months (full rollout)
• Training data: 10M historical call transcripts
• Languages: English only (Year 1), Spanish added (Year 2)

​

Use Case 2: BFSI Fraud Alerts (India)

• Bank: Top-5 Indian private sector bank
• Volume: 80M fraud alert calls/year (SMS + call)
• Current approach: IVR with keypad (DTMF) navigation
• Challenge: 40% customers don’t complete IVR (press wrong button, drop off)

Fraud alert triggered (suspicious transaction)
  ↓
AI places outbound call to customer
  ↓
Natural language:
  "Hi, this is [Bank Name] calling about a suspicious transaction.
   We noticed a ₹25,000 charge at [Merchant]. Did you make this purchase?"
  ↓
Customer responds (yes/no/unclear)
  ├─ Yes → "Thank you, transaction approved."
  ├─ No → "We've blocked the transaction and your card. A new card will arrive in 3 days."
  └─ Unclear → Transfer to human agent

• Platform: Skit.ai (India-based, Hindi + English + regional languages)
• Telephony: Airtel IQ (bank’s existing telco partner)
• Compliance: RBI guidelines on customer communication, voice biometrics for authentication
• Deployment: 4 months (pilot in one city) + 8 months (national rollout)

​

Use Case 3: Healthcare Appointment Reminders

• Healthcare system: Large US hospital network (20 hospitals, 300 clinics)
• Volume: 5M appointments/year
• No-show rate: 18% (industry average)
• Cost of no-show: $200/appointment (lost revenue + operational inefficiency)

48 hours before appointment:
  AI calls patient (or texts if no answer after 3 attempts)
  ↓
Natural language:
  "Hi [Patient Name], this is [Hospital Name] calling to confirm your appointment
   with Dr. [Name] on [Date] at [Time]. Can you confirm you'll be there?"
  ↓
Patient responds:
  ├─ Confirm → "Great, see you then. Do you need directions or parking info?"
  ├─ Reschedule → "Let me find a new time for you..." [Check calendar API]
  └─ Cancel → "I've canceled your appointment. Would you like to reschedule?"

• Platform: PolyAI (40-language support, accent-agnostic)
• Telephony: Bandwidth (HIPAA-compliant SIP trunking)
• Integration: Epic EHR (90% of appointments), Cerner (10%)
• Compliance: HIPAA, patient consent for automated calls
• Deployment: 9 months (pilot 2 hospitals) + 15 months (full network)

​

D. Financial Model Template

Customer Segments:
├─ SMB (50% of customers, 20% of revenue)
│   • ACV: $24k
│   • CAC: $3k (inbound, PLG)
│   • Gross margin: 70%
│   • Payback: 2.1 months
│   • Churn: 22%/year
│
├─ Mid-Market (35% of customers, 40% of revenue)
│   • ACV: $180k
│   • CAC: $22k (outbound + AE)
│   • Gross margin: 68%
│   • Payback: 3.6 months
│   • Churn: 12%/year
│
└─ Enterprise (15% of customers, 40% of revenue)
    • ACV: $800k
    • CAC: $120k (field sales + SE)
    • Gross margin: 65%
    • Payback: 5.5 months
    • Churn: 6%/year