Grow and run
Every call recorded, scored, and written into the lead form
Gandiv listens to each call after it ends, redacts sensitive details, scores the lead 0-100 and fills the form with an evidence quote for every value. You decide which fields AI may write and who may edit them.
11 languages
Hindi, English, Hinglish and 8 more Indian languages, with speaker split
Hot / Warm / Cold / Junk
Score 0-100 with written reasons for every call
4 edit rules
Editable, locked, editable with reason, or needs approval - per field and role
Call to transcript
Every call becomes searchable text
When a call ends, the recording is picked up from the phone or your telephony provider and transcribed automatically. Hindi, English, Hinglish and eight more Indian languages are handled, with agent and customer separated. Click a sentence and the audio jumps to that moment.
- Language detected per call
- Agent and customer split
- Search across all calls
Language detected per call
Agent and customer split
Search across all calls
Privacy first
Sensitive details removed before AI sees them
Before a transcript is stored or sent to a language model, Gandiv redacts card numbers, CVV, OTP, Aadhaar, PAN and bank account numbers. The list is configurable. Access to the original audio can be limited to chosen roles, so a caller never has to read a customer's OTP back.
- Configurable redaction list
- Redacted before the model
- Audio access by role
Configurable redaction list
Redacted before the model
Audio access by role
Lead scoring
Hot, Warm, Cold or Junk, with reasons
Each call gets a score from 0 to 100 and a label: Hot, Warm, Cold, Junk, wrong number or not interested. The card lists why: stated budget, timeline, urgency, objections, a competitor name. Scores are recomputed from every call and chat on the lead, not only the last one.
- 0-100 score and text label
- Reasons listed, not a black box
- Uses all calls of the lead
0-100 score and text label
Reasons listed, not a black box
Uses all calls of the lead
Auto-filled form
AI fills the form, shows its proof
You map any lead field to an instruction such as budget in rupees lakh, course interested in, or loan amount. After the call, AI returns the value, a confidence percentage and the exact quote with its time stamp. The caller confirms instead of typing the whole form.
- Works with custom fields
- Evidence quote with time stamp
- Confidence shown on every value
Works with custom fields
Evidence quote with time stamp
Confidence shown on every value
Field control
You decide what AI may write and who may edit
For each field, choose auto-fill, suggest or off, plus a minimum confidence below which AI only suggests. Then choose whether employees can edit: freely, locked, editable with a mandatory reason, or sent for manager approval. Rules can differ by role, team or lead source.
- Auto-fill, suggest or off
- Locked or approval-based edits
- Rules enforced on the server
Auto-fill, suggest or off
Locked or approval-based edits
Rules enforced on the server
Trust and cost
Audit badges, accuracy dashboard, AI credits
Every field carries a badge showing whether AI or a person set it, when, and from which call. An accuracy dashboard shows how often AI values are kept or edited. Platform AI credits are included in your plan, and you can bring your own model key if you prefer.
- Full field history drawer
- Kept vs edited per field
- Platform credits or your own key
Full field history drawer
Kept vs edited per field
Platform credits or your own key
How it works
- 1
Map the fields
Pick the lead fields AI should fill, write a one-line instruction for each and test it on a sample call.
- 2
Start in suggest mode
New workspaces begin with suggestions, so callers accept or reject values while you watch accuracy.
- 3
Switch on auto-fill
Move reliable fields to auto-fill, lock the ones you trust, and keep approvals on the sensitive ones.
Frequently asked questions
Does AI really fill the lead form without the employee?
Yes, for the fields you switch to auto-fill and only when confidence is above your threshold. Anything below it becomes a suggestion. New workspaces start in suggest mode so you can check accuracy first.
Can employees change what AI wrote?
Only if you allow it. Per field and role you can make it editable, locked, editable with a logged reason, or editable but sent for approval. The rule is enforced on the server, so imports and API calls follow it too.
Which languages are supported?
Hindi, English and Hinglish, plus eight other Indian languages. Accuracy varies with audio quality and accent, which is why every value shows its evidence quote and the accuracy dashboard exists.
Is my call data used to train models?
Our design is that tenant data is not used to train models, and sensitive details are redacted before any text reaches a model. Admins can view prompt and output logs for their own workspace.
What does AI cost?
Platform AI credits are included in your plan, with a usage meter and a hard cap you control. If you prefer, you can add your own model key. Check the pricing page for the credit quota of each plan.
See Gandiv on your own leads
Start free, or book a short demo and we will set it up with your pipeline.