Playbooks, not blank canvases.
Voice is not generic. A collections call in Chennai and a site-visit booking in Dubai share almost nothing beyond a telephone line. Every Nivākya engagement starts from a shipped playbook — a proven rule set, a set of agents and the metrics your buyer already tracks.
The four numbers our playbooks are actually judged on.
Five verticals where the phone still decides the quarter.
Each block below is the shape of a real deployment: the pain that paid for it, what the agent actually does, the numbers the sponsor reported, and one line lifted verbatim from the starter rule set.
Real estate — site visits, not gatekeeping
A national developer with fourteen live projects was losing roughly a fifth of its qualified enquiries to voicemail between 6pm and 9pm, when buyers actually browse. The leads were real, the intent was high, and nobody was home.
What the agent does
- ▪ Answers the project line on the first ring, in English, Hindi or Arabic, around the clock.
- ▪ Qualifies purpose, budget band, preferred possession date and financing status in one pass.
- ▪ Books the site visit into the right sales rep's calendar and texts directions plus a map pin.
- ▪ Writes the qualified lead, transcript and audio into Salesforce with the source project tagged.
“If the caller mentions a budget below ₹60 lakh for a 3BHK, do not offer Tower C inventory; offer the Phase-2 launch and note the constraint on the record.”
Healthcare — reminders that actually reduce no-shows
A multi-specialty clinic group was burning two full-time coordinators on confirmation calls and still running a 21% no-show rate across diagnostic appointments. Reschedules were being made in a browser that half the patients never opened.
What the agent does
- ▪ Confirms the appointment 48 hours ahead and reschedules in-call if the slot no longer works.
- ▪ Delivers prep instructions — fasting windows, medication holds, arrival times — in the patient's language.
- ▪ Triages by symptom to the correct department, capturing duration and severity as structured fields.
- ▪ Writes everything into HIPAA-scoped storage with no PHI leaving the pinned region.
“Never advise on medication. If the caller describes chest pain, breathlessness or bleeding, end the script and transfer to the on-call nurse line within one turn, and record their exact words.”
Lending & insurance — verification without friction
An NBFC's collections floor could reach about 1,900 accounts a day across three shifts, and the cost of a regulatory misstep was higher than the cost of the missed payment. Verification and early-bucket collections were both high-volume and unforgiving.
What the agent does
- ▪ Steps through KYC and income confirmation, capturing every answer as a discrete field.
- ▪ Offers EMI restructure or a part-payment option inside the policy limits you set.
- ▪ Records a promise-to-pay with date, amount and channel, then sets the follow-up.
- ▪ Enforces calling windows, DNC lists and per-state consent before any number is dialled.
“Third-party callers never receive balance information. Ask for the registered mobile number and the last four digits of the loan ID; if either fails, log identity_unverified and close politely.”
E-commerce — fewer failed deliveries
A D2C apparel brand shipping 12,000 cash-on-delivery orders a month was losing 9% of them to returns-to-origin — pure margin, gone to a courier truck. Bad addresses and unreachable buyers were the two causes, and both were fixable before dispatch.
What the agent does
- ▪ Confirms the COD order and delivery window before it leaves the warehouse.
- ▪ Repairs incomplete or ambiguous addresses, verifying landmark and pincode in-call.
- ▪ Offers a one-time prepaid discount to convert COD to prepaid where that improves margin.
- ▪ Handles returns triage — reason, pickup slot, replacement intent — in the same conversation.
“Offer the prepaid discount once, never twice. If the caller declines, confirm the address and dispatch; do not re-open the offer on the same call.”
BPO & call centres — absorb the peaks, keep the people
A 900-seat BPO was paying overtime to cover a nightly spike between 7pm and 11pm and losing a fifth of its tier-1 agents every quarter. The work driving both problems was repetitive, scripted and perfectly delegable.
What the agent does
- ▪ Absorbs the repetitive 68% — order status, balances, hours, repeat questions — without queueing.
- ▪ Warms-transfers anything complex with a three-line summary already on the human's screen.
- ▪ Holds overnight and holiday capacity flat, so the night shift stops being a staffing gamble.
- ▪ Feeds the same QA scorecards your supervisors already use, on every AI and human call alike.
“When a human takeover is requested, summarise in three lines — intent, what was tried, what the customer wants — before transferring, and never transfer on the first request without a settle attempt.”
Four jobs, whichever vertical you sit in.
Inbound automation
Every call answered on the first ring, at any hour, in the caller's language. Overflow, after-hours and holiday routing are the same agent wearing a different policy — not a separate queue nobody staffs.
- ▪ Zero queue time, no hold music, no abandonment
- ▪ Context fetched before the greeting finishes
- ▪ Escalation to a human with the summary already written
Outbound campaigns
Pipeline at machine speed, with the politeness enforced by policy rather than by a training slide. Calling windows, retry ladders, DNC scrubbing and per-region consent are checked before a number is dialled.
- ▪ Dialler-free pacing limited only by your trunk capacity
- ▪ Retry ladders that respect local time and holidays
- ▪ Every attempt logged with outcome and disposition
Lead qualification
Scoring you can defend in a revenue meeting. Every lead arrives with the transcript, the score and the exact sentences that produced it — so a rep can argue with the number and the number can answer back.
- ▪ Budget, authority, need and timeline captured as fields
- ▪ Disqualification reasons recorded, not discarded
- ▪ Instant push to the CRM with routing rules applied
Appointment setting
Calendar-native booking that reads real availability and writes real events. The caller never leaves the call to open a link, and the rep never finds out about a booking by email.
- ▪ Round-robin and skill-based assignment across teams
- ▪ Reschedule and cancel handled without a callback
- ▪ Reminders on WhatsApp, SMS or email two days out
Eight rules that ship on day one.
This is a slice of the real estate playbook. Rules are written as plain statements, mapped to your policy references, and versioned like code — so a change is reviewed, not improvised on a live call.
| Rule | Trigger | Action | Escalation |
|---|---|---|---|
| RE-QUAL-014 Budget floor | Caller names a budget below the tower's floor price | Offer the alternate inventory band; note the constraint | None — log only |
| RE-BOOK-002 Double confirmation | A site visit is booked | Repeat day, time and project name; ask the caller to confirm | Rep SMS if the caller hesitates |
| RE-CONSENT-001 Recording notice | Call connects | State the recording notice in full before any qualification | Hard stop if not spoken |
| RE-FIN-009 Financing referral | Caller asks about loan eligibility or EMI | Capture income band and hand to the mortgage desk | Warm transfer, 20s ceiling |
| RE-COMP-004 Competitor mention | Caller names a competing project | Acknowledge, do not disparage, recollect requirements | None — flag for marketing |
| RE-DNC-001 Do-not-call | Caller asks not to be contacted again | Confirm the request, suppress the number immediately | Alert compliance within 60s |
| RE-ID-005 Third-party caller | Caller is not the named buyer | Take a message only; disclose nothing about the file | None — log only |
| RE-HAND-001 Human request | Caller asks for a person | One settle attempt, then transfer with a three-line summary | Immediate warm transfer |
Four weeks from kick-off to a live number.
We run this alongside your team, with a named architect on our side and a single sponsor on yours. If a week slips, it slips visibly — the plan and the blocker are both in the shared workspace.
Week 1 — Discovery
We listen to fifty of your real calls, read your existing scripts and policy documents, and agree the metrics that decide success. Output: a scored playbook selection and a written rule set draft.
Week 2 — Build
Agents are configured in the Studio, tools and integrations are wired, and the voice and language mix is chosen. Output: a working agent on a test number, plus the CRM mappings signed off.
Week 3 — Rehearsal
Up to 10,000 synthetic callers run the agent, including adversarial personas drawn from your own edge cases. Output: a failure list, the fixes, and a signed rehearsal report your risk team can file.
Week 4 — Live
The number goes live on a limited slice of traffic with human oversight on every call for the first 48 hours. Output: a live workspace, dashboards, and a weekly review cadence.
Tell us your vertical. We'll show you the playbook.
Bring three real call recordings and one metric you are embarrassed about. We will map a playbook to it in the demo and show you the rule set we would start from.
No vertical fit? We will say so rather than sell you a pilot that cannot succeed.