An agent for every job you give it.

Every IndusLabs agent is built for one business and one job: who it talks to, what it knows, what it may do, and what it writes down afterwards. Pick a job, type your company's name, and watch the agent take shape.

The two-minute tour

How an agent goes from its first conversation to a record in your reports, and how it gets better every week.

Length: 2:00
Languages: 4

Asha

Sunrise's support agent

Your company

Opens with

Build this agent with us

Talk to Asha now, as one of Sunrise's customers. Uses your microphone.

Your help centre and return policy
Each caller's orders and tickets
What changed in the last release

The two-minute tour

How an agent goes from its first conversation to a record in your reports, and how it gets better every week.

Length: 2:00
Languages: 4

Watch a conversation turn into a record.

A re-enacted conversation, played at reading speed. As the other person talks, the agent fills in what you need to know. Pick another one to see a different job in a different language.

1
It listens in any language
Hindi, Tamil, Bengali, English, and every mix of them, mid-sentence.
2
It does the job
Takes the order, teaches the lesson, screens the candidate, books the slot.
3
It writes it down
Each fact lands in a field your team and your systems can use.

Conversations a day across our live deployments

74,300

NBFC32,000
Consumer appliances20,000
Automotive14,000
Healthcare6,500
Food retail1,800

Re-enacted for an appliance distributorHindi and English

Agent
नमस्ते शर्मा जी, Sunrise Distributors से बोल रहा हूँ। आपका last order 34 दिन पहले आया था, आज stock refill करना है?
Retailer
हाँ, 750 watt mixer grinder के 10 piece भेज दो। और 400mm pedestal fan का rate क्या है?
Agent
Fan ₹1,840 per piece है, और इस महीने 12 पर 1 free scheme चल रही है।
Retailer
ठीक है, fan के 12 piece भी add कर दो। Friday तक आ जाएगा?
Agent
जी, order बन गया है, Friday तक delivery होगी। Confirmation WhatsApp पर भेज रहा हूँ।

Agent
வணக்கம் Divya! நேத்து fractions பார்த்தோம். 3/4 + 1/8 எவ்வளவு?
Student
4/12?
Agent
Close! முதல்ல denominators ஒண்ணா ஆக்கணும். 3/4-ஐ எட்டில் எழுதினா என்ன வரும்?
Student
6/8! அப்போ answer 7/8.
Agent
Perfect! நாளைக்கு mixed fractions. அம்மாவுக்கு இன்னைக்கு summary அனுப்புறேன்.

Agent
Hi Rohit, आपने Pune में delivery partner role के लिए apply किया था। पाँच छोटे सवाल पूछ लूँ?
Candidate
हाँ, पूछिए। मेरे पास अपनी bike है और दो साल का delivery experience है।
Agent
बढ़िया। आप कौन-सी shift कर सकते हैं, morning या evening?
Candidate
Evening, 4 बजे से 11 बजे तक।
Agent
आप shortlist हो गए हैं। Thursday 11 बजे interview book कर दिया है, details SMS पर आ रही हैं।

Caller
আমার মায়ের জন্য cardiology appointment লাগবে। কালকে হবে?
Agent
Dr Sen কাল 11:30-এ free আছেন। মায়ের কি এখন বুকে ব্যথা বা শ্বাসকষ্ট হচ্ছে?
Caller
না, routine follow-up, আগের reports দেখাতে চাই।
Agent
Booked, কাল 11:30, OPD 3। পুরোনো reports সঙ্গে আনবেন, SMS-এ details পাঠাচ্ছি।

Record, written by the agent

Outcome: Order placed
Order value: ₹58,480
Delivery: Friday
Follow-up: WhatsApp confirmation sent
Items: Mixer grinder 750 W × 10, Pedestal fan 400 mm × 12

Talk to it yourself.

We can't play you our customers' calls, so talk to the agent instead. Same voice and speech models, any of 29 languages.

Start in English, finish in Hindi: Switch mid-sentence. It follows without being asked.
Interrupt it: Cut in while it talks and change your mind.
Book something: Ask for a service slot and give it a day and time.
Ask for a person: Say it any way you like, and hear how it hands over.

Ready when you are

नमस्ते and 28 more languages

Uses your microphone while the call is on. Leaving the page ends the call.

All of it, in one dashboard.

Every conversation lands in your reports, in the fields you chose. Here, for a reorder agent: whether the retailer ordered, why not, and what was promised. Click any value to see the calls behind it.

Demo workspace

Agent builder

How to build an agent from scratch, how Retailer reorder is doing, and the fields it fills in.

Build an agent, step by step
How well it understands
Paths its calls take
Where reply time goes

Build an agent from scratch

A step-by-step walkthrough, from the first line it says to going live. Video coming soon.

Step 1: Pick the job and the languages

Who the agent calls, why, and which languages it may switch to.

Chapters

What the walkthrough covers

Understood on first try: 94%; Replies that needed no repeat
Reply time: 0.1 s; Median, end of speech to first word
Handed to a person: 2.1%; With the transcript attached
Stayed on script: 98%; Checked against the opening and rules

How conversations went

The paths all 8,412 connected calls took through the agent

Hover a path to see how many calls followed it.

Where the reply time goes

The 0.1 s from the retailer stopping to the agent speaking

Hearing the end of speech22 ms
Deciding what to say53 ms
Starting to speak25 ms

Report fields filled

Share of connected calls where the agent recorded a usable value

Order placed100%
Why no order97.8%
Language the retailer chose100%

Report fields

The fields Retailer reorder fills in at the end of every call. These become the reports. Click a value to see it there.

Order placed
Did the retailer place an order on this call?
Why no order
If no order, the main reason the retailer gave.
Language the retailer chose
The language the call ended in.

Calls

Every dial this week: what connected, what didn't, and when to call.

Where every dial ended up
Why calls missed, and the retry
When each retailer picks up
How the calls sounded

Calls dialled: 11,960; Every attempt, including retries
Connected: 8,412; 70.3% of dialled
Attempts per connect: 1.42; Down from 1.61 last week
Callbacks kept: 96%; Called at the time asked

Where every dial went

All 11,960 calls this week, from the dial to how each one ended

Hover a flow to see how many calls took that path.

Connect rate, day by day

This week against last week, and what the next three days should look like

8 Oct: 73% this week, 68% last week

234567891011

This week
Last week
Forecast range

Why calls didn't connect

3,548 calls. Pick a reason to see what the agent did.

What the agent does: Retried after 2 hours, up to 3 times

Best time to call

Share of dialled calls that connected, last 4 weeks

Thu, 5–7: 86% · best slot

9–1111–11–33–55–77–9

Mon

Tue

Wed

Thu

Fri

Sat

Sun

6 pm: 84% connect · peak hour

9 am12 pm3 pm6 pm8 pm

Calling each retailer at their predicted best time

Fixed time for everyone61%

Best time per retailer74%

+13 points more calls connected

Next best time, per retailer

Each retry is timed for the person, not the list

Retailer 4194Thu, 5:40 pm · Picked up 4 of the last 5 calls after 5 pm88% likely
Retailer 4231Thu, 6:10 pm · Shop is busiest before 5; answers after closing rush81% likely
Retailer 4268Fri, 11:30 am · Asked for a morning call last time74% likely
Retailer 4305Sat, 12:15 pm · Missed two evening calls; trying a new window63% likely

Attempts needed

Which try reached the 8,412 connected calls

1st attempt5,720 68%
2nd attempt1,884 22.4%
3rd attempt808 9.6%

Caller ID health

Outgoing numbers are rotated and rested so they never get marked as spam

Numbers in rotation: 12
Marked as spam: 0
Rested this week: 2

How the calls sounded

Voice analytics across the 7,236 calls that ran past 30 seconds

Agent 38%62% Retailer

Talk time. The retailer does most of the talking, as it should be.

34% of calls ended warmer than they started.

Retailer mood through the call

End: Positive 41% · Neutral 51% · Negative 8%

Start of callMiddleEnd of call

Positive
Neutral
Negative

Interruptions handled: 1,204; Agent stopped talking within 300 ms
Dead air: 1.1 s; Average silence per call
Language switches: 41%; Calls that mixed Hindi and English
Words understood: 96.2%; Speech recognition confidence

Call length

Connected calls by how long they ran

Most common: 1–2 m, 2,520 calls

Reports

How Retailer reorder's calls went, last 7 days

Calls: 8,412; +18% vs last week
Average call length: 2:12; +6% vs last week
Total talk time: 308 h; +21% vs last week
Calls with a recording: 100%; 8,412 of 8,412 calls

Calls per day

Busiest 8 Oct with 1,385

After the call

Sent automatically when the call ends. Click one to see it.

WhatsApp to Retailer 4120

Your order is confirmed: 10 × Ceiling fan 1200mm. Delivery by Friday. Reply here to change anything.

Call outcomes

What the agent recorded at the end of each call

Order placed

Did the retailer place an order on this call?

Why no order

If no order, the main reason the retailer gave.

Language the retailer chose

The language the call ended in.

Call logs filtered to Stock not moving, 2,416 calls

It gets better every week.

Every week the records show why calls fail. The biggest reason gets fixed, and the new version has to beat the old one before it stays. In two months, that took this agent from 7.9% to 13.0% of calls ending in an order.

Demo data

Calls that ended in an order

Week by week, 4 Aug to 13 Oct

Since v1: +5.1pts
More orders, same calls: 65%

Each step up is a new version. Hover a week or click a version.

4 Aug18 Aug1 Sep15 Sep29 Sep13 Oct

How v2 got there · 25 Aug

Every call's outcome, ranked by the reasons calls did not convert.

What the records showed

"Busy, call later" was the top reason, at 41% of lost calls: the shop's morning rush.

Busy, call later41%
Stock not moving24%
Asked for price list14%
Ordered from salesman12%
Other9%

What it could be worth to you.

Pick your industry, the agent's job and your call volume. Every number is yours to change. No sign-up.

Your industry

What should the agent do? · pick one or more

How many calls a day? · every dial, answered or not

Reorder calls to retailers

Reorder calls to retailers: 52,000 calls × 45% reached × 3.0% = 702 orders (₹1.3 Cr of sales). 30% extra × 8.0% margin = ₹3.0 L profit.

Calling cost: 11 people × ₹22,000 = ₹2.4 L. One person makes about 183 of these calls a day in 6.5 productive hours, with 1.5 minutes of notes after each answered call.

Change any number · use your own

Working days a month

Reorder calls to retailers

Reach a person (%)End in orders (%)Revenue each (₹)Your margin (%)Extra, not happening anyway (%)

For Distribution / FMCG, these calls could be worth about

₹5.5 La month

In calling cost saved and extra profit, before IndusLabs fees.

Calling cost saved₹2.4 L
Extra profit from extra orders₹3.0 L

₹1.3 Cr of sales would be taken on these calls. Only your margin on the share that is extra is counted.

Orders a month: 702
Talk time handled: 975 h
People these calls take by hand: 11
What that team costs a month: ₹2.4 L

Book a demo with these numbers

Where it already works.

Nine live deployments. Our customers stay unnamed, but the results are theirs, measured in production.

Consumer appliances · retail distribution

A leading Indian consumer appliances brand

₹48.6 L

order value captured per day

Before

Retailer coverage depended on salesmen physically showing up. A salesman on leave meant his whole beat went unvisited, retailers who stopped ordering went unnoticed for weeks, and new launches only reached whoever someone happened to visit.

What the agents do

Four outbound agents work the retailer base daily: beat cover for absent salesmen, reactivation after 30 days without an order, reward-point prompts, and new-product introductions. Each takes the order on the call, straight into the order system.

Derived: ~9,000 connected calls a day convert at ~3%, at an ₹18,000 average order value — retailers often trial a new line with a one- or two-item order, which keeps the average modest.

20,000

retailer calls placed per day

45%

calls connected

₹48.6 L

order value captured per day

Automotive · dealer network

A global automotive manufacturer

31%

of service reminders convert to a booking

Before

Showrooms answered the phone only in business hours, so evening and weekend enquiries — when most buyers call — went nowhere. Service dates, insurance renewals and exchange eligibility sat in the DMS, acted on only when a dealer had spare capacity.

What the agents do

One inbound agent covers the line around the clock for bookings, queries and lead routing. Four outbound campaigns run alongside: service-due reminders, insurance renewals, exchange and upgrade offers, and new-launch announcements.

One 24×7 inbound agent plus four outbound campaigns across the owner base.

14,000

calls handled per day

27%

of enquiries arrive outside showroom hours

31%

of service reminders convert to a booking

NBFC / lending

An Indian NBFC lender

23%

reduction in 30+ day delinquency

Before

Servicing, collections and credit verification competed for one small telecalling team. Pre-due reminders were first to be dropped at month-end, pushing accounts into later buckets that cost far more to recover, while field verification added days to every sanction.

What the agents do

Six agents split the work. Three inbound: existing loans, new applications, general queries. Three outbound: pre-due EMI reminders, post-due collections with compliant scripting, and a verification agent running structured PD checks before and after credit assessment, escalating anything inconsistent.

Verification agent runs pre- and post-assessment PD checks on every application.

6

agents in production, inbound and outbound

32,000

borrower conversations per day

23%

reduction in 30+ day delinquency

Food retail · gifting

A regional sweets and snacks chain

62%

of customers complete the call

Before

High-value gifting and bulk orders were the ones most worth getting right, and the ones nobody had time to follow up. Complaints surfaced days later on social media, and festive planning ran on last year's totals and instinct.

What the agents do

A feedback agent calls every customer whose order crossed ₹5,000 about freshness, delivery and packaging, flagging complaints to the store manager same-day. Responses feed an analytics agent that forecasts demand by line ahead of each festive peak.

Feedback outcomes feed the forecasting agent that drives festive line planning.

1,800

feedback calls per day on orders above ₹5,000

62%

of customers complete the call

4.4/5

average satisfaction captured

Interiors & décor

An interiors and décor retailer

under 2 min

median time to first contact

Before

Enquiries arrived from ads, web and walk-ins with no consistent qualification, so designers spent consultation slots on prospects with no budget or timeline. After-hours leads sat until morning, by which point most had called someone else.

What the agents do

A qualification agent calls every new enquiry within minutes to establish budget, timeline, property stage and scope, then books the consultation into the right designer's calendar. Unqualified leads are nurtured rather than taking a slot.

Every enquiry is called within minutes, day or night; unqualified leads are nurtured instead of taking a designer's slot.

4,200

leads qualified per month

under 2 min

median time to first contact

38%

pass through as sales-ready

Healthcare · patient access

A multi-speciality hospital group

82%

resolved without a human

Before

The reception line handled everything at once — bookings, report status, ambulance requests, emergencies — and was busiest exactly when staff were most stretched. Someone in genuine distress waited in the same queue as someone asking about visiting hours.

What the agents do

One inbound agent answers every call around the clock: bookings against live doctor availability, report status, ambulance requests and visiting queries, answered in full. It triages as it talks, and any sign of an emergency or distress goes straight to an on-duty supervisor, transcript and caller record already on screen.

The agent triages but never advises clinically — anything critical is escalated to a human, not answered.

6,500

inbound calls handled per day

82%

resolved without a human

under 10s

to reach a supervisor on a critical call

Insurance · renewals and leads

A private general insurer

22%

of renewal calls end in a payment

Before

Renewal reminders went out by SMS and email that most customers ignored, and the call centre could only phone the largest policies before they lapsed. Web leads waited a day or more for a first call, by which time many had bought elsewhere.

What the agents do

A renewal agent calls every policy 30, 15 and 3 days before expiry, answers premium and cover questions, and sends the insurer's payment link on WhatsApp during the call. A second agent calls each new web lead within minutes and books a slot with an advisor.

Payments go through the insurer's own link; the agent never takes card details on the call.

12,000

policyholders called per day

22%

of renewal calls end in a payment

under 5 min

from web lead to first call

E-commerce / D2C · order confirmation

A D2C personal care brand

31%

fewer parcels returned to origin

Before

Most orders were cash on delivery, and too many were refused at the door, paying shipping both ways. Failed deliveries were followed up only when the courier escalated, often after the parcel was already on its way back.

What the agents do

A confirmation agent calls every cash-on-delivery order within the hour to confirm the address and intent, and offers a prepaid discount by WhatsApp link. A delivery agent calls after each failed attempt, takes a new slot or address, and updates the courier the same day.

Returns measured against the three months before the agents went live.

8,000

orders confirmed per day

60%

of customers reached on the first attempt

31%

fewer parcels returned to origin

Education · admissions

A higher-education institute group

15%

book a counselling session

Before

Admission enquiries peak for a few weeks each year, and counsellors could not call them all back. The earliest enquiries got attention and the rest went cold, while fee reminders pulled counsellors away from new students.

What the agents do

An admissions agent calls every enquiry the same day in the student's language, answers course, fee and eligibility questions, and books a counselling session. A fee agent reminds enrolled students before each due date and records the date they promise to pay.

Calls run in the student's own language, and every fee promise is logged against the enrolment.

2,500

enquiries called per day in peak season

15%

book a counselling session

100%

of enquiries called the same day

Built in-house, connected to what you run.

We make the voice and speech models ourselves, so we can tune them for Indian accents, names and code-switching instead of waiting on someone else's roadmap.

Connected to 9 systems you already run

Writes to the systems you already use.

Orders, payment promises and bookings go straight into your ERP or CRM when the call ends. Confirmations and owner alerts go out on WhatsApp.

Sales order SO-48213 posted to SAP

Point at a system to see what a call sends it.

29 languages, one voice

Our own voice and speech models.

नमस्ते
வணக்கம்
নমস্কার
Hello

Text-to-speech that keeps one voice across 29 languages, and speech recognition tuned for Hindi-English mixing. 14 Indian languages, from Hindi and Tamil to Bhojpuri and Chhattisgarhi.

Read the API docs

Policies and processes are public

Accountable by design.

DPDP Act notice and consent
Recording and transcript on every call
Escalation to a person, with the transcript
Published governance documents

Every outcome in your reports links back to the call that produced it.

Visit the Trust Center

Book a demo.

Tell us about your calls: how many, in which languages, and which systems they touch. We'll come back with the agent we'd build and what we expect it to return.

Or reach us directly

Conversations in.
Value out.

An agent for every job you give it.

Watch a conversation turn into a record.

Talk to it yourself.

All of it, in one dashboard.

It gets better every week.

What it could be worth to you.

Where it already works.