The Floor Index is a benchmark for how a Human + AI contact-centre floor behaves in India, published with its method and refreshed every six months. First edition, November 2026: the AI finishes up to 40% of calls by month 3, 50% by month 4 and 70 to 80% by month 6, settling at 60 to 80% at maturity. A traditional agent seat costs Rs 28,000 to 35,000 a month; AI minutes run from Rs 2.50 at a pilot rate to Rs 1.75 on committed volume. Production covers 14 Indian languages plus English and Hinglish. All figures are counted on the Oriserve platform across 40+ enterprises and about a million calls a day, and labelled as such.
1. The ramp
The share of calls finished by the AI, by month, on a typical programme. The marks are the figures we commit to in writing. The curve starts low because the AI learns the client's scripts, objections and edge cases on live calls while a full human bench takes the rest. The band at maturity, 60 to 80%, is the number to compare any vendor's claim against: a figure above it in week one is a bot's containment number and says nothing about the floor.
2. The cost shape
| Unit | Range, India, November 2026 | Basis |
|---|---|---|
| Traditional agent seat, English or Hindi, loaded | Rs 28,000 to 35,000 per month | Market range; senior and specialised agents higher |
| AI minute, pilot rate | Rs 2.50 | Published pilot rate, no commitment |
| AI minute, largest committed slab | Rs 1.75 | Published slab floor; middle slabs on quote |
| Agent console seat (Vox Agent Desktop) | Rs 2,000 per agent per month | Published; transcripts and wrap-up included |
| Handover between AI and person | Rs 0 | Published |
| Setup | Once per integration, on quote | SIP, CRM read and write, agent-desk webhook |
People are a fixed cost per seat and the AI a variable cost per minute; the handover between them should cost nothing. That shape matters more than the ranges. A floor that moves from the first to the second changes its cost curve from flat to volume-linked, and month one and month six look different for that reason.
3. The language set
Fourteen Indian languages in production: Assamese, Bengali, Gujarati, Hindi, Kannada, Malayalam, Marathi, Nepali, Odia, Punjabi, Sindhi, Tamil, Telugu and Urdu, with English and Hinglish alongside, and switching mid-call. We include this in the index because the share of calls an AI can finish in India depends first on whether it can hold the call in the customer's language at all.
Method
- Counted on the Oriserve platform across the programmes running on it; not a market survey. Figures from other vendors' floors are not included and are not estimated.
- The ramp is the typical shape across programmes; individual programmes sit above and below it. Figures are stated as ranges where programmes vary.
- Cost figures are the published Oriserve rates and a market range for seats; they are not an average of other vendors' prices.
- Not in this edition: per-turn latency (the public figure has not completed our claims process); a same-call handover rate (we have not yet agreed a definition we would stand behind across programmes); client-level results (two are on the proof page with their windows).
- Next edition: May 2027. Readers who run floors and want their own programme's ramp drawn against this one can ask; the comparison is free and private.
Quoting the index
Quote it as: Human + AI Floor Index, India 2027, first edition, Oriserve, November 2026. Link to this page. If you quote the ramp, quote the band at maturity with it.
Questions people ask
What is the Human + AI Floor Index?
A dated, open benchmark for how a Human + AI contact-centre floor behaves in India: the AI-share ramp by month, the cost shape, the language set. Counted on the Oriserve platform and labelled as such. Refreshed every six months.
What share of calls does AI finish on a Human + AI floor?
Up to 40% by month 3, 50% by month 4, 70 to 80% by month 6, settling at 60 to 80% at maturity, on a typical programme.
What does a contact-centre seat cost in India?
Rs 28,000 to 35,000 per month for a standard English or Hindi agent, loaded; more for senior or specialised agents.
Is the index based on Oriserve data only?
Yes, and it says so. It is counted across 40+ enterprises and about a million calls a day on the platform; it is not a market survey.