"There is a difference between an AI system that informs a decision and a system that makes one. In partnership with SquadStack, we are building the latter."
"We have been able to integrate our product stack, telephony, lead routing, and escalation workflows well with SquadStack's technology. The objective is to move towards an AI-enabled marketplace where every buyer interaction, every channel, every language, is intelligent, contextual, and autonomous by default, and works to deliver a seamless customer experience."
"What we built together was shaped entirely by what IndiaMART's buyers actually need. IndiaMART serves buyers across a lakh of product categories, in multiple languages, with every possible variation of intent and context. Meeting that required going feature by feature, use case by use case, real-time product pivoting, Tamil-language conversations, visual intelligence on live calls, memory across buyer journeys, until there were no significant gaps left. That thoroughness, combined with the speed at which both teams moved, is what produced the outcome."
"At IndiaMART, there's no concept of a 'vendor.' We look for partners who can match our velocity. SquadStack matched us neck-to-neck on every front."
The Scale of the Problem
IndiaMART is India's largest online B2B marketplace. For 30 years, it has connected buyers and sellers across lakhs of product categories, languages, and geographies, bringing the ease of doing business to millions of Indian entrepreneurs.
A garment wholesaler in Pune finds a manufacturer of cotton T-shirts in Tirupur. A pump factory in Ludhiana gets its first inquiry from a trader in Delhi. Industrial machinery, surgical equipment, fashion apparel, food ingredients, packaging materials. Lakhs of categories. Millions of transactions.
Behind each of those transactions is a conversation. A buyer explaining what they need, negotiating terms, comparing options. IndiaMART processes millions of these buyer-seller interactions daily. At that scale, the lead journey faced friction points that a human-only contact center could not solve.
No human team can build expertise across lakhs of product categories, in dozens of languages, around the clock. The scale of IndiaMART's marketplace demanded something different.
What Buyers Actually Do on IndiaMART
The challenge goes deeper than volume. No human agent can build expertise across lakhs of product categories. And IndiaMART's buyers behave in ways that break most AI systems.
A buyer in Ludhiana calls from a factory floor with machines running in the background. He sends a photo of a machine part mid-call on a 3G connection instead of describing what he needs. VANI has to identify the part, find the right category out of lakhs on IndiaMART's catalogue, and respond before he loses patience.
Same second, a conversation in Chennai is happening in Tamil-English. Another in Varanasi in Hindi. A fourth buyer is calling back about steel pipes he discussed three days ago and expects VANI to pick up where they left off.
A buyer asks about one product, then pivots to another mid-call. Another calls with no product name, no city, no name of their own. Someone sends only a photo of what they want.
Multiply that by a hundred thousand. Every day.
How VANI Works
SquadStack deployed a full-stack agentic Voice AI system as the first interaction layer across IndiaMART's buyer-seller journey. Built inside IndiaMART's stack rather than bolted on top, the system reached full production scale within three months of pilot.
Three process lanes run simultaneously:
- Outbound lead engagement: When a buyer searches for a product, VANI calls back, confirms what they need, and captures quantity and specs in real time.
- Inbound missed-call recovery: If a buyer calls a seller who doesn't pick up, VANI steps in immediately, captures the requirement, and ensures no high-intent lead is lost.
- Outbound priority (hot-lead follow-up): Dropped inbound calls are flagged as hot leads and re-engaged instantly.
VANI handles calls in fluent Hindi, English, and Tamil, tuned to IndiaMART's category tree and seller workflows. The system replaces fragmented human outreach with a single AI-led layer plugged into IndiaMART's telephony, lead routing, and escalation systems.
Capabilities Built for IndiaMART
Every capability was engineered, tested, and shipped for IndiaMART's actual buyer behavior.
- Immediate Redial: If a call drops within 10 seconds, VANI redials automatically. Connection drop-offs that would otherwise be lost get recovered.
- Real-Time Product Pivot: When a buyer changes intent mid-call, VANI hits IndiaMART's Search API live and pivots to the new product with the right qualification questions on quantity and specs. No restart, no dead air.
- No Name, No Product, No City: When the buyer hasn't shared anything to start with, VANI runs structured discovery instead of failing or dropping the call.
- Tamil Language Fluency: VANI converses fluently in Tamil at production scale on the IndiaMART platform. Regional language coverage is expanding.
- Visual Intelligence: VANI analyzes product images shared mid-call, understanding design, material, and structure to qualify the buyer with physical-store precision.
- Real-Time Human Handoff: When VANI's confidence dips, the conversation moves to a human agent with full context. The buyer never repeats themselves. The handoff is invisible.
- Persistent Memory: VANI stores and retrieves call-level memory across multiple calls for the same lead, enabling multi-touch journeys, follow-ups, and resumable conversations.
- International Calls: AI infrastructure extends to Nepal, Bangladesh, and beyond, with language models tuned for cross-border commerce.
- WhatsApp Calling (planned): Full voice call capability on WhatsApp. Buyers and sellers connecting via AI-handled calls on India's most-used channel.
%20(1).png)
The Results
VANI beat the human-agent baseline on the same lead pool, across connectivity, conversion, quality, and cost.
Headline outcomes:
- 1 lakh+ autonomous buyer-seller conversations every day, run by AI from first connect to final handoff
- 20% higher conversions than manual calls on the same lead pool
- ~15% lower cost per confirmed lead, translating into better unit economics for sellers
- 75%+ AI connectivity rate, against 50-60% for human agents
- 95% outcome accuracy across the conversation flow
- Three months from pilot to full production, with no drop in buyer experience through the ramp
Operational outcomes:
- Quality uniformity at scale: A consistent qualification standard now holds across lakhs of categories and multiple languages.
- 24/7 coverage: Buyer traffic gets served at full capacity at every hour, without cost scaling linearly with volume.
- Invisible handoffs: Only the calls that exceed VANI's confidence reach a human, and the context travels with the call. The buyer never starts over.
%20(1).png)
What Comes Next
What started as a pilot on buyer-seller connections is now the largest agentic AI deployment in live commerce in India. Vernacular coverage is moving past Tamil into more Indian languages. The roadmap extends the system across memory, channels, and geographies.
VANI deployed as the first interaction layer, not a point tool, compounds. Every new capability builds on the infrastructure, data, and workflow integration of the one before it. For a marketplace at IndiaMART's scale and complexity, that compounding is what makes it work.
