Typical AI phone tool
One instruction set
A single prompt tries to cover every caller need at once. The model improvises what happens next.
Unpredictable when the call gets complex
Features
Built for live patient calls, not demos
Named safeguards on every turn, confirm-to-book rules, full call logs, and set-up that can go live in 24 hours after a complete handover.
Call control
How step-by-step call control works
Most AI phone tools let one instruction set run the whole call. Vgentive breaks each call into clear steps. The system decides what happens next, not free-form AI answers.
Vgentive
Clear steps per turn
Each call is broken into defined steps. The system decides what happens next — not free-form AI answers.
Every turn logged with a named outcome
What happens on a live call
Six defined phases. Each one has a handler, a fallback, and a logged outcome — not a guess from the language model.
Step 01
What does the caller need?
The first words decide the path: question, booking, emergency, reminder reply, or handoff to staff. Each path follows its own steps, not an open chat.
- Question
- Booking
- Emergency
- Reminder reply
- Staff handoff
Step 02
Facts from your setup only
Timings, doctors, slots, and answers come from what you have set up. If nothing is found, the agent says so. It does not make up an answer.
Step 03
Caller must say yes
Bookings and sensitive steps need a clear yes from the caller. The agent reads back doctor, time, and department from the booking system only.
Step 04
Call closes with a record
Every ended call is logged on the Facility Desk with transcript and outcome. Built-in checks run throughout the call, not only at the end.
Step 05
Staff can take over anytime
Live transfer paths are defined in your emergency and desk rules. The agent does not hold callers in a loop when staff should answer.
Step 06
Outbound uses the same engine
Reminders, re-engagement, and feedback calls follow scripted flows with the same safeguards as inbound, started only from the Facility Desk.
Booking rules in practice
What the agent is allowed to say
Redacted call excerpt. System results are shown in brackets. The agent does not invent doctors or slots. It does not guess when a lookup fails.
Live lookups only
Doctor names and appointment slots come from your schedule, never invented.
Clear when data is missing
If a lookup fails, the agent says so and offers real alternatives from your roster.
Caller confirmation required
Bookings are read back and confirmed only after an explicit yes from the caller.
System results drive speech
Every factual line maps to a tool result, visible in the transcript as system events.
Illustrative call excerpt
Anonymised sampleMulti-speciality clinic
Redacted sample. System lookups appear between turns. The agent only speaks from what those tools return.
CallerBooking enquiry flowAgent
Caller
Do you have Dr. Sharma available tomorrow morning?
- System result
doctor.lookup → no match for "Sharma"
Agent
I don't have a Dr. Sharma on our schedule. We have Dr. Mehta and Dr. Kapur in general medicine. Would you like an appointment with either of them?
Caller
Dr. Mehta, please.
- System result
slots.list → 10:30 AM, 11:15 AM (tomorrow)
Agent
Dr. Mehta has 10:30 AM or 11:15 AM tomorrow. Which works for you?
Caller
10:30.
- System result
booking.create → pending confirmation
Agent
I'll book Dr. Mehta tomorrow at 10:30 AM. Shall I confirm that booking?
Caller
Yes.
- System result
booking.confirm → confirmed
Agent
Your appointment with Dr. Mehta is confirmed for tomorrow at 10:30 AM. Is there anything else I can help with?
Built-in checks
40+ safeguards, named
Each check is a small rule that runs during a live call. The platform has 40+ of them. Below is a sample grouped by what each protects.
Boundaries
Built with limits, on purpose
A voice agent on a hospital line should know what it isn't. Here's what Vgentive refuses to do, by design.
No medical advice
Symptoms go to a department or doctor. The agent never diagnoses, recommends medication, or suggests treatment.
No invented confirmations
A booking is only confirmed after the booking system succeeds. If it fails, the agent says so. It does not pretend the appointment is set.
No invented facts
If something isn't in your knowledge base, the agent says it doesn't know and offers a callback. It never guesses to sound helpful.
No unsolicited outgoing calls
Reminders, follow-up, and feedback calls only go out when your staff starts them. Nothing dials on its own schedule.
No silent failures
If a call drops, stalls, or hits an error, it's logged and flagged. It is not marked as a clean, successful call.
No pretending it's human
The agent identifies itself as the hospital or clinic voice assistant. It does not claim to be a staff member.
No third-party disclosure
The agent will not share another patient's appointment, diagnosis, or details with a caller, regardless of how the request is phrased.
No cross-facility data
A branch agent never reads another facility's patients, schedule, or knowledge, even under a shared group login.
No off-script promises
The agent will not offer discounts, waivers, or clinical commitments that are not in your configured knowledge.
These aren't gaps. They're the same discipline that makes the rest of the agent's behavior something you can trust on a real patient line.
Set-up model
Every hospital or clinic runs on the same engine
24-hour onboarding works because you change data, not code. The voice engine and built-in checks are shared. Your doctors, hours, and rules are yours alone.
Per facility
What changes per hospital or clinic
Loaded as your data and settings, with no custom code per site.
- Doctors, departments, and bookable services
- Hours, holidays, and slot rules
- Knowledge checklist (parking, insurance, fees, custom facts)
- Greeting, language setting, and clinic voice
- Emergency acceptance and transfer rules
- Incoming and outgoing call types per module
Shared platform
What never changes
The platform every site inherits on day one.
- Voice engine and speech stack
- Step-by-step call control and 40+ built-in checks
- Booking chain: doctor schedule → slots → confirm
- Safety fast-paths and refusal rules
- Facility Desk modules, call records, and audit trail
Testing discipline
Built from real production calls
Every safeguard on this page exists because a real call surfaced the problem it solves. We do not ship call behaviour from a whiteboard alone.
Scenarios from real recordings
70+ test scenarios come from anonymised hospital and clinic calls. They cover booking edge cases, emergency language, Hindi-English switches, and schedule mismatches.
Checks tied to observed failures
When a live call exposes a gap, the fix becomes a named check with a replay scenario. It is not a one-off instruction tweak.
Language parity before release
Hindi sites get Devanagari copy. English sites get English-only paths. Both are tested on the same scenarios before release.
Replay before every release
Named checks are replayed against stored call scenarios before a build goes live. A failed replay blocks release until the fix is verified.
Flags visible in the Facility Desk
When a safeguard fires on a live call, the flag name and outcome appear on the call record. Ops can filter and review without digging through raw logs.
This is ongoing review from production calls and scenario replay. It is not a claim that every path runs in a fully automated test suite today.
Facility Login
What your team controls after go-live
The desk is where staff review calls, manage the doctor schedule, and run outgoing calls, separate from the voice engine that handles the line.
Facility Desk modules
See call quality, full call records, appointments, outgoing call queues, review flags, and credit balance in one login.
Team roles and credit controls
Set who sees call records, spends credits, or starts outgoing calls. Voice pauses when credits run out.
One platform, many sites
Each hospital or clinic gets its own knowledge, doctor schedule, language setting, and emergency rules on shared platform infrastructure.
Recently added
Voices you can hear
Your facility gets a natural clinic voice from an official speaker list. Staff hear a sample before it goes live on the line. Unofficial speakers are not used.
Voice stays out of the long settings form
Changing the speaker is a focused picker on the facility record, so the rest of hospital setup stays short.
Low-credit warning
The Facility Desk shows a warning when a hospital is running low on credits, before calls are affected.
True Hindi on generated copy
Knowledge drafts, campaign lines, reminder wording, and call summaries follow the hospital's language. Hindi facilities get proper Devanagari, not Latin Hinglish. Every scenario is tested in both Hindi and English before release.
Review flags on every call
Calls that need review are flagged automatically with a named reason, visible on the Facility Desk before your team opens the transcript.
Credit history per call
Every minute, transfer, and AI desk action is logged against the call record so finance and ops can audit spend.
Everything the hospital runs from one desk
People, reception, outgoing calls, services list, and access. Scoped to your account.
Home
Overview of calls, outcomes, and quality, including completion and needs-review. Credits: balance and history. View only.
Overview of calls, outcomes, and quality, including completion and needs-review. Credits: balance and history. View only.
People
Callers who phoned the line. Patients the agent and desk share for booking and follow-up.
Callers who phoned the line. Patients the agent and desk share for booking and follow-up.
Reception
Call logs with transcript, intent, outcome, and duration. Open a conversation and generate an AI summary (uses credits). Enquiry view. Em…
Call logs with transcript, intent, outcome, and duration. Open a conversation and generate an AI summary (uses credits). Enquiry view. Emergency when the facility accepts it, with transfer to staff. Appointments: calendar, slots, desk booking, and next available, from the same roster the voice agent uses.
Re-engagement
Campaign lists with agenda, calling window, and desired outcome. People and leads, including CSV upload. The agent records the outcome on…
Campaign lists with agenda, calling window, and desired outcome. People and leads, including CSV upload. The agent records the outcome on the call.
Reminders and scheduled
Appointment reminders with confirm, cancel, callback, or no-answer. Staff-scheduled calls for follow-up, missed visit, feedback, payment …
Appointment reminders with confirm, cancel, callback, or no-answer. Staff-scheduled calls for follow-up, missed visit, feedback, payment reminder, or custom, with Call now, reschedule, or cancel.
Patient Experience
Post-visit patient feedback: ratings, comments, and questions the hospital edits. Alerts on poor scores.
Post-visit patient feedback: ratings, comments, and questions the hospital edits. Alerts on poor scores.
Services list
Departments and doctors with hours and bookable slots. Knowledge: a checklist of hospital facts (hours, parking, insurance, billing, book…
Departments and doctors with hours and bookable slots. Knowledge: a checklist of hospital facts (hours, parking, insurance, billing, booking rules, plus custom facts). Not a document-upload chatbot. Staff can ask for an AI draft in the hospital's language.
Access
Users and roles for this facility only.
Users and roles for this facility only.
Facility settings
Hours, emergency, online, offline, or both, and incoming or outgoing. Turn reception, re-engagement, reminders, and feedback on or off. D…
Hours, emergency, online, offline, or both, and incoming or outgoing. Turn reception, re-engagement, reminders, and feedback on or off. Desk menus hide what is off. Language for the hospital. The clinic voice is set up for staff, not picked in a long form.
Support
Tickets to Vgentive when the hospital needs help.
Tickets to Vgentive when the hospital needs help.
Branches
Switch facility if the admin is on more than one branch.
Switch facility if the admin is on more than one branch.
Blocked states
Suspended facility or zero credits: clear desk screens, not a silent failure.
Suspended facility or zero credits: clear desk screens, not a silent failure.
How calls actually work
Not one AI prompt. A state machine with 40+ named checks per call.
Every turn has one decision owner. The agent never guesses what to do next.
Generic voice AI runs on a single instruction set: one prompt that has to handle enquiry, booking, emergency, decline, and hang-up simultaneously. When it fails, you don't know why.
Vgentive runs a deterministic call engine: each turn has a defined handler, a fallback path, and a named outcome. Booking turns bypass free-form AI entirely. Slot lookup, read-back, and confirmation are code, not language model output.
Every defense flag has a name: unknown_doctor blocked, fake_booking_confirm refused, repeated_intent_menu detected. Each one is logged, visible in the Facility Desk, and tested against real hospital recordings before each release.
Knowledge base
25+ hospital fact topics. Not a document upload. Not a chatbot.
The agent only speaks from what you configure, structured by topic, not scraped from a PDF.
Scoped by topic
Parking, fees, insurance, booking rules, teleconsult. Each fact type has its own field. The agent cannot answer what you haven't added.
Visibility by facility type
An online-only clinic's agent won't invent ambulance or parking answers. Knowledge scope follows your operating mode.
Drafted in your language
Knowledge entries are drafted by AI in your facility's language: Hindi for Hindi facilities, English for English. Your team reviews before it goes live.
Re-engagement
Design outbound call scripts in a visual flow editor. No code.
Branch on caller responses, add delays, loop on no-answer. Built visually, run reliably.
Most outbound call tools give you a script field. Vgentive gives you a flow builder: branches, delays, loop conditions, and LLM-enhanced openers, built visually, executed without free-form AI inference on scripted calls.
Facebook leads land directly in your calling list. Meta Lead Ads integration auto-ingests new leads without a CSV upload.
Visual branch editor
If caller says yes, if no-answer, if declined.
Step delays and loop conditions
Delays and loop conditions between attempts.
Meta Lead Ads → calling list
Automatically. No CSV upload required.
Scripted calls bypass LLM inference
Faster, cheaper, more predictable.
Patient Experience
More than feedback calls. A full post-visit ops module.
Ratings, trends, volume audits, and queue-eligible bulk calling, not just a list of call outcomes.
Trends over time
Rating trends and volume audit across any date range. See if patient satisfaction is improving week on week.
Poor-score alerts
Low ratings are flagged automatically so your ops team can follow up before a bad review goes public.
Queue-eligible bulk actions
Select completed visits in bulk and queue feedback calls in one action. No manual list management.
Outcome per call
Every feedback call is logged with rating, comment, and call recording, tied back to the original appointment.
Department-level trends
Filter satisfaction and volume by department or doctor so ops can spot weak points without exporting spreadsheets.
Recording archive
Every feedback call keeps its recording and transcript in the Facility Desk, tied to the visit and rating for audit.
Multi-branch
Hospital chains: one group, separate lines, one admin login.
Each branch keeps its own phone line, credit balance, and Facility Desk. Your admin sees all of them.
One parent group
Branches are linked under a parent organisation. Group-level admin can switch between facilities without logging out.
Per-branch credits
Each branch has its own credit balance and transaction history: shared admin, separate spend.
Per-branch branded desk
Each branch gets its own subdomain (branch.vgentive.com) and login. Patients and staff never see another facility's data.
Reliability
Built for peak OPD hours, not quiet demo rooms.
Stall detection, barge-in control, and conflict warnings, because real clinics are noisy and busy.
Stall detection
If the agent goes silent for ~9 seconds, it plays a check-in line instead of dead air. No caller ever sits in silence wondering if the call dropped.
Barge-in control
Callers can interrupt the agent while it's speaking, but not during processing. No accidental cutoffs, no lost words.
Outbound conflict detection
The desk warns your team before a reminder call, confirmation call, and scheduled follow-up all queue for the same patient on the same visit.
Get a facility live
Book a demo. Bring your facility phone line and doctor list. See inbound and outbound healthcare use cases on the Facility Desk.